Long D A The Raman Effect (Wiley, 2002) (610S)

The Raman Effect: A Unified Treatment of the Theory of Raman Scattering by Molecules. Derek A.
Long
Copyright 2002 John Wiley & Sons Ltd
ISBNs: 0-471-49028-8 (Hardback); 0-470-84576-7 (Electronic)
The Raman Effect
The Raman Effect

A Unified Treatment of the Theory of Raman Scattering by
Molecules
Derek A. Long
Emeritus Professor of Structural Chemistry
University of Bradford
Bradford, UK
Copyright 2002 by John Wiley & Sons Ltd,

Baffins Lane, Chichester,
West Sussex PO19 1UD, England
National
01243 779777
International (+44) 1243 779777
e-mail (for orders and customer service enquiries): [email protected]
Visit our Home Page on http://www.wileyeurope.com
or http://www.wiley.com
All Rights Reserved. No part of this publication may be reproduced, stored in a retrieval
system, or transmitted, in any form or by any means, electronic, mechanical, photocopying,
recording, scanning or otherwise, except under the terms of the Copyright, Designs and Patents Act
1988 or under the terms of a licence issued by the Copyright Licensing Agency, 90
Tottenham Court Road, London, W1P 9HE, without the permission in writing of the
Publisher.
Other Wiley Editorial Offices
John Wiley & Sons, Inc., 605 Third Avenue,
New York, NY 10158-0012, USA
Wiley-VCH Verlag GmbH, Pappelallee 3,
D-69469 Weinheim, Germany
John Wiley Australia, 33 Park Road, Milton,
Queensland 4064, Australia
John Wiley & Sons (Asia) Pte Ltd, 2 Clementi Loop #02-01,
Jin Xing Distripark, Singapore 129809
John Wiley & Sons (Canada) Ltd, 22 Worcester Road,
Rexdale, Ontario, M9W 1L1, Canada
Library of Congress Cataloguing in Publication Data

Long, D. A. (Derek Albert)
The Raman effect : a unified treatment of the theory of Raman scattering by molecules /
Derek A. Long.
p. cm.
Includes bibliographical references and index.
ISBN 0-471-49028-8 (acid-free paper)
1. Raman spectroscopy. I. Title.
QD96.R34 L66 2001
535.80 46dc21
2001046767
British Library Cataloguing in Publication Data
A catalogue record for this book is available from the British Library
ISBN 0 471 49028 8
Typeset in 11/13pt Times by Laserwords Private Limited, Chennai, India
Printed and bound in Great Britain by Biddles Ltd, Guildford and Kings Lynn
This book is printed on acid-free paper responsibly manufactured from sustainable forestry,
in which at least two trees are planted for each one used for paper production.
Dedicated to
Edward and William Long,
grandsons.
vii
Contents
Preface
xix
Acknowledgements
xxiii
Part One Theory
Survey
1.1
1.2
1.3
1.4
1.5
1.6
1.7
1.8
1.9
1.10
of Light-scattering Phenomena
Introduction
Some Basic Definitions
Rayleigh and Raman Scattering
1.3.1
Description
1.3.2
Energy transfer model
Hyper-Rayleigh and Hyper-Raman Scattering
1.4.1
Description
1.4.2
Second Hyper-Rayleigh and Second Hyper-Raman Scattering
1.5.1
Description
1.5.2
Coherent anti-Stokes and Coherent Stokes Raman Scattering
Stimulated Raman Gain and Loss Spectroscopy
Typical Spectra
Bases for the Theoretical Treatment of Rayleigh and
Raman Scattering
Historical Perspective
3
3
4
5
5
7
10
10
10
11
11
11
11
13
14
16
16
viii
Contents
1.11
Caveat
References
17
17
Introduction to Theoretical Treatments of Incoherent Light Scattering

2.1
General Considerations
2.2
Induced Oscillating Electric Dipoles as Sources of
Scattered Radiation
2.3
Basis of the Classical Theory of Light Scattering
2.4
Basis of the Quantum Mechanical Treatment
of Incoherent Light-Scattering Phenomena: Electric Dipole Case
2.5
Extension of Quantum Mechanical Treatment of Incoherent
Light Scattering to Include Magnetic Dipole and Electric
Quadrupole Cases
2.6
Comparison of the Classical and Quantum Mechanical
Treatments of Light Scattering
2.7
The Way Ahead
19
19
Classical Theory of Rayleigh and Raman Scattering

3.1
Introduction
3.2
First-order Induced Electric Dipole
3.3
Frequency Dependence of the First-order Induced Electric
Dipole
3.4
Classical Scattering Tensors aRay and aRam
k
3.5
Selection Rules for Fundamental Vibrations
3.5.1
General considerations
3.5.2
Diatomic molecules
3.5.3
Polyatomic molecules
3.6
Selection Rules for Overtones and Combinations
3.7
Coherence Properties of Rayleigh and Raman Scattering
3.8
Limitations of the Classical Theory
3.9
Example of Rayleigh and Raman Scattering
3.10
Presentation of Raman Spectra
References
31
31
31
Quantum Mechanical Theory of Rayleigh and Raman Scattering

4.1
Introduction
4.2
Time-dependent Perturbation Theory and afi
4.3
Qualitative Discussion of fi
4.3.1
Frequency denominators
4.3.2
Transition electric dipole numerators
4.3.3
Selection rules
4.4
Tensorial Nature of the Transition Polarizability and its Symmetry
4.5
BornOppenheimer Approximation and the Transition
Polarizability Tensor
21
22
24
27
28
29
34
35
36
36
36
38
43
44
45
45
47
48
49
49
50
54
55
56
58
58
61
Contents
4.6
4.7
4.8
4.9
4.10
4.11
4.12
4.13
ix
Simplification of ef vf :eg vi : General Considerations
Simplification by Radical Approximation: the Placzek
Transition Polarizability
Simplification of efvf:eivi by Stages
4.8.1
Introduction of HerzbergTeller vibronic coupling
4.8.2
Identification of non-resonance and resonance
situations
Normal Electronic (and Vibronic) Raman Scattering
Normal Pure Vibrational Raman Scattering
Electronic (and Vibronic) Resonance Raman Scattering
Vibrational Resonance Raman Scattering
Units and Orders of Magnitude
References
Vibrational Raman Scattering

5.1
Introduction
5.2
The Placzek Vibrational Transition Polarizability:
Recapitulation
5.2.1
Cartesian basis
5.2.2
The spherical basis
5.3
Definition of IlluminationObservation Geometry
5.4
Intensity of Scattered Radiation: Some General Considerations
5.4.1
Development of a symbol for scattered intensity
5.4.2
Scattering cross-section
5.5
Intensity Formulae and Polarization Characteristics for a General
Vibrational Transition in Various IlluminationObservation
Geometries
5.5.1
5.5.2
Linearly polarized incident radiation
5.5.3
Natural incident radiation
5.5.4
Angular dependence of scattered intensity
5.5.5
Circularly polarized incident radiation
5.5.6
Symmetry and depolarization ratios, reversal
coefficients and degrees of circularity
5.6
Stokes Parameters for Scattered Radiation
5.7
Specific Vibrational Transitions
5.8
Vibrational Selection Rules
5.9
Patterns of Vibrational Spectra
5.10
Orders of Magnitude
5.11
Epilogue
References
Reference Tables for Chapter 5
Reference Table 5.1: Definitions for I; ps , pi
64
65
68
68
75
77
78
81
83
83
84
85
85
86
86
88
89
94
94
95
97
97
98
102
103
106
109
113
116
120
123
126
127
131
132
132
Contents
Reference Table 5.2(a) to 5.2(g): Intensities, Polarization
Properties and Stokes Parameters for Vibrational Raman
(and Rayleigh) Scattering
Reference Table 5.3: Symmetry classes for x, y, z, the rotations
Rx , Ry and Rz , and the components of the cartesian basis
tensor c a.
132
145
Rotational and VibrationRotation Raman Scattering

6.1
Introduction
6.2
Irreducible Transition Polarizability Components
6.3
Symmetric Top
6.3.1
Selection rules
6.3.2
Placzek invariants G j fi
6.3.3
Intensities
6.3.4
Subsequent development
6.4
Rotational and Vibrational Terms
6.5
Statistical Distribution of Molecular Population
6.6
Diatomic Molecule
6.6.1
Introduction
6.6.2
Heteronuclear diatomic molecule: pure rotation
6.6.3
Heteronuclear diatomic molecule: vibrationrotation
6.6.4
Homonuclear diatomic molecule: nuclear spin degeneracy
6.6.5
Intensity distribution
6.7
Symmetric Top Molecule
6.7.1
Introduction
6.7.2
Symmetric top: pure rotation
6.7.3
Symmetric top: vibrationrotation
6.7.4
Intensities
6.8
Linear Molecules
6.8.1
Rotation and vibration-rotation Raman spectra
6.8.2
Intensities
6.9
Contributions from Electronic Orbital and Spin Angular Momenta
6.10
Spherical Top Molecules
6.11
Asymmetric Top Molecules
6.12
Epilogue
References
Introduction
Reference Tables 6.1 to 6.4
153
153
154
156
156
157
167
169
169
171
173
173
174
175
179
180
186
186
187
191
203
204
204
207
208
210
211
211
213
214
214
216

7.1
Introduction
7.2
Vibrational Transition Polarizability Tensor Components in the
Resonance Case, Based on Perturbation Theory
221
221
222
Contents
7.3
7.4
7.5
7.6
7.7
7.8
7.9
xi
Comparison of the AVI , BVI , CVI and DVI Terms
7.3.1
The AVI term
7.3.2
The BVI term
7.3.3
The CVI term
7.3.4
The DVI term
7.3.5
Subsequent developments
AVI Term Raman Scattering from Molecules with Totally
Symmetric Modes
7.4.1
AVI term Raman scattering from molecules with one
totally symmetric mode
7.4.2
AVI term Raman scattering from molecules
with more than one totally symmetric mode: general
considerations
7.4.3
AVI term Raman scattering from totally symmetric modes
when k is very small
VI
A Term Raman Scattering Involving Non-Totally Symmetric
Modes
7.5.1
7.5.2
AVI term scattering involving a change of molecular
symmetry of the resonant excited state
7.5.3
AVI term scattering involving excited state JahnTeller
coupling
7.5.4
Summary of excited state JahnTeller effects in
resonance Raman scattering
BVI Term Scattering Involving Vibronic Coupling of the Resonant
Excited State to a Second Excited State
7.6.1
Introduction
7.6.2
BVI term scattering from molecules with non-totally
symmetric modes
7.6.3
BVI term scattering from molecules with totally
symmetric modes
Symmetry, Raman Activity and Depolarization Ratios
7.7.1
General symmetry considerations
7.7.2
The AVI term
7.7.3
The BVI term
Time-Dependent Formulation of Resonance Raman Scattering
7.8.1
Introduction
7.8.2
Transformation of the AVI term to a time-dependent
expression
7.8.3
The time-dependent interpretation of resonance Raman
scattering
Continuum Resonance Raman Scattering
References
224
224
227
229
229
230
231
231
237
238
239
239
239
240
240
241
241
241
244
246
246
247
250
262
262
263
264
266
270
xii
Contents
Rotational and VibrationRotation Resonance Raman Scattering

8.1
Introduction
8.2
General Expression for fi for a Symmetric Top Molecule
j
8.3
General Expression for m fi
j
8.4
Contraction of General Expression for m fi
8.5
The Quadratic Term
8.6
Selection Rules
j
8.7
Evaluation of jm fi j2
8.8
Intensities and Depolarization Ratios
8.9
An Illustrative Example
8.10
Concluding Remarks
Reference
271
271
272
274
274
275
276
277
279
283
287
287
Normal and Resonance Electronic and Vibronic Raman Scattering

9.1
Introduction
9.2
Normal Electronic and Vibronic Raman Scattering
9.2.1
9.2.2
AIII -term scattering
9.2.3
BIII C CIII -term scattering
9.2.4
DIII -term scattering
9.2.5
Transition tensor symmetry
9.3
Resonant Electronic and Vibronic Raman Scattering
9.3.1
9.3.2
AV -term scattering
9.3.3
BV -term scattering
9.3.4
CV -term scattering
9.3.5
DV -term scattering
9.4
Selection Rules in Electronic Raman Spectra
9.4.1
General symmetry considerations
9.5
Intensities and Polarization Properties of Electronic Raman
Scattering
9.5.1
Intensities: general considerations
9.5.2
Excitation profiles
9.5.3
Depolarization ratios
289
289
289
289
290
291
292
292
292
292
293
296
297
297
297
297
Rayleigh and Raman Scattering by Chiral Systems

10.1
Introduction
10.2
Outline of the Theoretical Treatment
10.3
Intensities of Optically Active Rayleigh Scattering
10.3.1 General considerations
10.3.2 Intensity formulae
10.3.3 Stokes parameters
303
303
305
310
310
314
317
10
301
301
301
302
Contents
10.4
10.5
10.6
xiii
Intensities of Optically Active Raman Scattering
10.4.2 Discussion of intensities and isotropic invariants
10.4.3 Placzek polarizability theory and optically
active scattering
Symmetry Considerations
Concluding Remarks
Reference
321
321
323
324
326
327
327
328
Part Two Appendices
337
Introduction
339
A1 The Right-handed Cartesian Axis System and Related Coordinate

Systems
A1.1 Introduction
A1.2 The Right-handed Cartesian Axis System
A1.3 Cartesian Coordinate System
A1.4 Cylindrical Coordinate System
A1.5 Polar Coordinate System
A1.6 Complex Coordinate Systems
341
341
342
344
344
345
347
A2 The Summation Convention

A2.1 General Definitions
349
349
A3 Direction Cosines
A3.1 Introduction
A3.2 Definitions and Properties of Direction Cosines
A3.3 Definitions of Direction Cosines in Other Coordinate Systems
351
351
351
354
A4 Isotropic Averages of Products of Direction Cosines

A4.1 Introduction
A4.2 Specific Isotropic Averages of Products of Direction Cosines
A4.3 General Formulae
References
355
355
356
358
358
A5 The Euler Angles and the Rotation Operator

A5.1 Introduction
A5.2 Definitions of the Euler Angles and the Rotation Operator
A5.3 The Relationship of the Euler Angles to the Polar Coordinates
A5.4 Direction Cosines and Euler Angles
References
359
359
359
362
362
363
xiv
Contents
A6 Complex Numbers and Quantities

A6.1
Introduction
A6.2
Definitions and Operations
A6.3
Graphical Representation of Complex Numbers
A6.4
Complex Numbers and Polar Coordinates
A6.5
Complex Quantities and Physical Phenomena
A6.6
Spherical Coordinates
Reference
365
365
365
367
369
370
371
372
A7 Some Properties of Matrices

A7.1
Introduction
A7.2
Nomenclature
A7.3
Some Special Matrices
A7.4
Matrix Representation of Simultaneous Linear Equations
A7.5
Eigenvalues and Eigenvectors
A7.6
Example of Diagonalization of a Matrix
373
373
373
374
377
377
379
A8 Vectors, I
A8.1
Introduction: Scalars, Vectors and Tensors
A8.2
Basic Definition of a Vector
A8.3
Unit Vectors
A8.4
Vector Addition, Subtraction and Multiplication by a Scalar
A8.5
Multiplication of Two Vectors
A8.6
Triple Products of Vectors
A8.6.1 AB C
A8.6.2 A B C
A8.6.3 A B C
A8.6.4 ABC
A8.7
Formal Definition of a Vector in Terms of its Transformation
upon Rotation of Axes
A8.8
Polar and Axial Vectors: Time Even and Time Odd Vectors
A8.9
Vector Differentiation
A8.9.1 The operator
A8.9.2 The gradient
A8.9.3 The divergence
A8.9.4 The curl, V
A8.9.5 The divergence and the curl illustrated and compared
A8.9.6 Composite functions involving
A8.9.7 Successive applications of
A8.9.8 Time derivative of a vector
A8.9.9 Caveat
A8.10 Change of Basis Vectors and Effect Upon Coordinates
of a Fixed Vector
A8.11 The Effect of a Symmetry Operation on Vectors and Basis Vectors
381
381
382
383
385
385
390
391
391
392
392
393
394
395
395
396
397
397
398
398
400
401
401
401
404
Contents
xv
A9 Vectors, II
A9.1 Introduction
A9.2 Cylindrical Coordinates and Basis Vectors
A9.3 Polar Coordinates and Polar Basis Vectors
A9.4 Spherical Components and Spherical Basis Vectors and Direction
Cosines
A9.5 Rotation of Vectors using Spherical Coordinates
A9.6 Vectors in n-Dimensional Space
References
407
407
407
408
A10 Tensors
A10.1 General Definitions
A10.2 Representation or Specification of a Tensor
A10.3 Transformation of Tensors upon Rotation of Axes
A10.4 Some Properties of Tensors
A10.4.1 General
A10.4.2 Tensors of rank two
A10.4.3 Tensors of rank three: the alternating or
LeviCività tensor
A10.4.4 Isotropic tensors of various ranks
A10.4.5 Tensor contraction
A10.5 Irreducible Tensorial Sets
417
417
419
422
425
425
425
426
427
427
428
A11 Electrostatics
A11.1 Introduction
A11.2 Force Between Charges
A11.3 Electric Field Strength
A11.4 Electrostatic Potential
A11.5 Gausss Law
A11.6 The Equations of Poisson and Laplace
433
433
433
435
436
437
438
A12 Magnetostatics
A12.1 Introduction
A12.2 Magnetic Forces
A12.3 The Magnetic Induction
A12.4 The Lorentz Force on a Point Charge Moving in a
Magnetic Field
A12.5 The Divergence of the Magnetic Induction B
A12.6 The Vector Potential A
439
439
441
442
445
446
446
A13 The Interaction of a System of Electric Charges with Electric

and Magnetic Fields
A13.1 Introduction
A13.2 Point Charges in Molecular Systems
449
449
450
409
413
415
416
xvi
Contents
A13.3
A13.4
Electric Dipole in a Molecular System

Basic Treatment of the Energy of a Distribution of Point Charges
at Rest in a Uniform Electric Field
A13.5 Basic Treatment of Permanent and Induced Molecular Dipoles
in a Uniform Static Electric Field
A13.6 Basic Treatment of Macroscopic Polarization and Electric
Susceptibilities
A13.7 Basic Treatment of the Electric Displacement for a Uniform
Static Electric Field
A13.8 The Implications of Using Dynamic Electric Fields
A13.9 More General Treatment of Energy of Interaction of Point Charges
at Rest in a Static Electric Field
A13.10 Interaction of Charges in Motion with a Static Magnetic Field
Reference
A14 The Polarizability Tensor

A14.1 Introduction
A14.2 The Polarizability Tensor in the Cartesian Basis
A14.2.1 General considerations
A14.2.2 Reduction of the tensor
A14.2.3 The polarizability ellipsoid
A14.2.4 Transformation of c a under rotation of axes
A14.3 The Polarizability Tensor in the Spherical Basis
A14.3.1 General definitions
A14.3.2 Reduction of the tensor s a
A14.3.3 Transformation of s a under rotation of axes
A14.4 The Relation Between the and the
A14.5 Irreducible Polarizability Tensors and their Components
j
A14.6 Transformation Properties of the m under Rotations
A14.7 Isotropic Averages and Tensor Invariants
A14.7.2 Isotropic averages and rotational invariants G j in terms
j
of the m
of the
of the
A14.7.5 Isotropic averages and the rotational invariants G j ,
a, and
for the cartesian basis
A14.7.6 Isotropic averages and rotational invariants, G j , a,
and
for the spherical basis
References
450
451
454
457
460
461
462
466
470
471
471
472
472
473
475
477
477
477
479
481
481
482
486
486
486
487
488
490
490
494
495
Contents
xvii
A
A15 The Optical Activity Tensors, G, G , A and A
A15.1 Introduction
A15.2 Isotropic Averages of the Type haG 0 i
A15.3 Isotropic Averages of the Type haAi
A15.4 Other Tensor Invariants
Reference
497
497
497
499
500
501
A16 Maxwell Equations in Vacuum and in Media

A16.1 The Fundamental Equations
A16.2 Case I: the Maxwell Equations in Vacuum
A16.3 Case II: the Maxwell Equations in a Linear Medium
with D 0 and D 0
A16.4 Case III: the Maxwell Equations in a Non-Linear Medium
with D 0 and D 0
A16.5 Case IV: the Maxwell Equations in a Linear Medium
with 6D 0 and 6D 0
503
503
507
A17 Monochromatic Plane Harmonic Waves in Vacuum and in

a Non-absorbing Linear Medium
A17.1 General Wave Equation in Vacuum
A17.2 Monochromatic Plane Harmonic Electromagnetic Wave in
Vacuum
A17.2.1 Solution of the wave equation for the vector E
A17.2.2 Solution of the wave equation for the vector B
A17.2.3 Energy considerations for a plane harmonic
electromagnetic wave in vacuum
A17.3 The Exponential Representation of Harmonic Waves
A17.4 Monochromatic Plane Harmonic Wave in a Homogeneous,
Isotropic and Linear Medium
508
509
511
513
513
514
514
518
522
526
530
A18 The Transition Polarizability Tensor afi

A18.1 Introduction
A18.2 The Restrictions r 6D i, f
A18.3 The Relative Signs of ir
References
533
533
533
535
536
A19 ClebschGordan Coefficients and Wigner 3-j and 6-j Symbols

A19.1 Introduction
A19.2 ClebschGordan Coefficients
A19.3 Wigner 3-j Symbols
A19.4 Wigner 6-j Symbols
Reference
Reference Table A19.1
537
537
538
544
546
550
551
xviii
Contents
A20 Sources of Electromagnetic Radiation

A20.1 Introduction
A20.2 The Oscillating Electric Dipole as a Source
A20.3 The Oscillating Magnetic Dipole as a Source
A20.4 The Oscillating Electric Quadrupole as a Source
A20.5 Scattering from Chiral Molecules
555
555
555
561
562
564
A21 Polarization of Electromagnetic Radiation

A21.1 Introduction
A21.2 States of Polarization: Monochromatic Radiation
A21.2.1 Linear polarization
A21.2.2 Elliptical and circular polarization
A21.2.3 Stokes parameters
A21.2.4 Stokes parameters for scattered radiation
A21.3 States of Polarization: Quasi-Monochromatic Radiation
A21.4 Change of Polarization: Depolarization Ratios, Reversal
Coefficients and Degrees of Circularity
565
565
565
565
566
570
572
573
575
Further Reading
579
Index
585
xix
Preface
Many ingenious practizes in all trades, by a connexion
and transferring of the observations of one Arte, to the
use of another, when the experiences of severall misteries
shall fall under the consideration of one mans mind.
Francis Bacon
Raman spectroscopy is now finding wide-ranging application in pure and applied science
and the number of original papers devoted to this area of spectroscopy continues to grow.
This is largely the result of significant advances in the equipment available, particularly
laser excitation sources, spectrometers, detectors, signal processors and computers.
It seems timely, therefore, to provide an integrated treatment of the theory underlying
Raman spectroscopy. Of course there are already a number of edited books and reviews
dealing with various aspects of the subject, but this book is the result of the phenomenon of
Raman spectroscopy falling under the consideration of one mans mind as Francis Bacon
put it. My objective has been to present a unified theoretical treatment which is reasonably
complete and adequately rigorous but nonetheless readable. My hope is that this will
provide a sound basis for the effective use of more highly specialized review articles.
As to completeness, I have had to put some restrictions on the coverage, partly because
the subject is so vast and partly because of my own limitations. Therefore the treatments
developed here relate mainly to scattering by a system of freely orienting, non-interacting
molecules or by systems which approximate to this. As to rigour, I have endeavoured to
explain in words, as far as possible, the inwardness of the mathematics and physics which
are necessarily involved. I have particularly tried to avoid taking refuge behind that often
overworked phrase as is well known.
An effective theoretical treatment demands a variety of carefully honed mathematical
and physical tools. To keep the treatment in the main text uncluttered, these tools are
xx
Preface
developed in comprehensive Appendices to which cross-references are made in the main

text. These Appendices should also ensure that the main text is useful to readers with a
wide variety of scientific backgrounds and experience.
As far as possible the symbols used to represent physical quantities are based on the
IUPAC recommendations but to avoid excessive embroidery with subscripts and superscripts it has been necessary to introduce a few new symbols, all of which are clearly
defined. The SI system of units is used throughout, except in those few instances where
spectroscopists commonly adhere to historical units, as for example the unit cm1 for
wavenumber and related quantities.
In the main text I have limited citations of the literature. This has enabled me to use the
Harvard system and quote names of authors and the dates of publication directly in the
text. I find this preferable to the anonymity and lack of historical sequence which result
from the use merely of reference numbers in the text. More complete lists of publications
are provided in the section entitled Further Reading, located at the end of the book.
The writing of this book has been a somewhat lengthy process and I have had to learn a
great deal along the way despite more than 50 years of work in this field. Happily I have
been able to find the time and the energy required. The University of Bradford enabled me
to free myself of administrative responsibilities by allowing me to take early retirement
and then reappointed me in an honorary capacity and provided excellent facilities for
research and scholarship. I am very grateful for these arrangements.
I am also much indebted to a number of other universities and institutions which
invited me to spend short periods with them so that I could use their libraries and benefit
from discussions with colleagues. In France I would mention the National Laboratory for
Aerospace Research (ONERA), and the Universities of Bordeaux I, Lille, Paris VI and
Reims. In Italy I would mention the University of Bologna and the European Laboratory
for Nonlinear Spectroscopy (LENS), attached to the University of Florence. In this country
I have made frequent use of the excellent facilities of the Radcliffe Science Library,
Oxford University. My periods in Oxford were made all the more pleasurable by the
kindness of my old Oxford College, Jesus, in making me a supernumerary member of its
Senior Common Room for a period. Nearer home, the J. B. Priestley Library of Bradford
University has also been much consulted and I am particularly indebted to Mr John
Horton, deputy librarian, for his unstinting help.
The following friends in the community of Raman spectroscopists have kindly read
and commented fruitfully upon sections of this book: A. Albrecht, D. L. Andrews,
L. D. Barron, J. Bendtsen, H. Berger, A. D. Buckingham, R. J. H. Clark, T. J. Dines,
H. Hamaguchi, S. Hassing, L. Hecht, M. Hollas, W. J. Jones, W. Kiefer, I. M. Mills,
O. Mortensen, H. W. Schrotter, G. Turrell, A. Weber, R. Zare and L. D. Ziegler. I
acknowledge with gratitude their efforts which have eliminated many errors and
ambiguities. I am also grateful to Claude Coupry and Marie-Therèse Gousset for providing
Plate 5.1 and to H. G. M. Edwards for providing Plate 5.2.
I would also like to record my appreciation of the patience and continued support
of John Wiley and Sons. I would mention particularly Dr Helen McPherson, Chemistry
Publisher, who has been most considerate throughout; and Mr Martin Tribe who has been
an efficient and imperturbable Production Coordinator.
Preface
xxi
Very special thanks are due to my wife Moira. She has very ably undertaken most
of the word-processing of the text, helped considerably with the style and presentation,
and provided loving encouragement. Rupert, another member of the family, has also
earned honourable mention. His insistence upon regular walks has helped to offset the
sedentary effects of authorship and his relaxed presence in my study has been calming
and companionable.
It would be unrealistic to expect that this wide-ranging book will be entirely free of
errors. In the words of Evan Lloyd, an eighteenth century Welshman, also a graduate of
Jesus College, Oxford, I can only plead that
Earnest is each Research, and deep;
And where it is its Fate to err,
Honest its Error, and Sincere.
December 2001
Derek A. Long
Acknowledgements
Comme quelquun pourrait dire de moi
que jai seulement fait ici un amas de fleurs e trangères,
nayant fourni du mien que le filet a` les lier.
M. E. Montaigne
I am grateful for permission to reproduce the material listed below.

Figure 5.11, Plate 5.2 Edwards, H. G. M., Russell, N. C. and Wynn-Williams, D. D.
(1997) Fourier-Transform Raman Spectroscopic and Scanning Electron Microscopic Study
of Cryptoendolithic Lichens from Antarctica. J. Raman Spectrosc., 28, 685, John Wiley
& Sons, Ltd, Chichester. Reproduced with permission from John Wiley & Sons, Ltd,
Chichester.
Figure 6.13 from Bendtsen, J. (1974). J. Raman Spectrosc., 2, 133. Reproduced with
permission from John Wiley & Sons, Ltd, Chichester.
Figure 6.14 from Bendtsen, J. and Rasmussen, F. (2000). J. Raman Spectrosc., 31, 433.
Reproduced with permission from John Wiley & Sons, Ltd, Chichester.
Table 7.1 from Mortensen, O. S. and Hassing, S. (1980). Advances in Infra-red and
Raman Spectroscopy, volume 6, 1, eds. R. J. Clark and R. E. Hester, Wiley-Heyden,
London. Reproduced with permission from John Wiley & Sons, Ltd, Chichester.
Figures 7.67.12 inclusive from Clark, R. J. H. and Dines, T. J. (1986). Resonance
Raman Spectroscopy and its Application to Inorganic Chemistry. Angew. Chem. Int. Ed.
Engl., 25, 131. Reproduced with permission from Wiley-VCH Verlag GmbH, Weinheim.
xxiv
The Raman Effect
Figures 7.147.18 inclusive from Kiefer, W. (1995). Resonance Raman Spectroscopy

in Infrared and Raman Spectroscopy, ed. B. Schrader. VCH Verlag GmbH, Weinheim.
Reproduced with permission from Wiley-VCH Verlag GmbH, Weinheim.
Figures 8.2, 8.3, 8.5, 8.6 from Ziegler, L. D. (1986). J. Chem. Phys., 84, 6013. Reproduced with permission of the American Institute of Physics.
Table 8.2 from Ziegler, L. D. (1986). J. Chem. Phys., 84, 6013. Reproduced with permission of the American Institute of Physics.
Figure 10.1 from Nafie, L. A. and Che, D. (1994). Theory and Measurement of Raman
Optical Activity, in Modern Non-linear Optics, part 3, eds. M. Evans and S. Kielich.
Reproduced with permission from John Wiley & Sons, Inc., New York.
Figures A8.6, A8.7, A8.8 from Barron, L. D. (1983). Molecular Light Scattering and
Optical Activity. Cambridge University Press, Cambridge. Reproduced with permission
from Cambridge University Press.
Figure A8.10a, b, c from Atkins, P. W. (1983). Molecular Quantum Mechanics, Oxford
University Press, Oxford. Reproduced by permission of Oxford University Press.
Reference Table A19.1 from Zare, R. N. (1988). Angular Momentum, John Wiley &
Sons, Inc.: New York. Reproduced with permission from John Wiley & Sons, Inc., New
York.
I also acknowledge my indebtedness to the literature of Raman spectroscopy in general
and in particular to the following authors and their publications upon which I have
drawn for the topics stated. For scattering by chiral systems, reviews and books by
L. D. Barron, L. Hecht and L. A. Nafie; for rotational and vibration-rotation Raman scattering, reviews by S. Brodersen, by W. J. Jones and by A. Weber; for vibrational and
electronic resonance Raman scattering, reviews by R. J. H. Clark and T. J. Dines, by
H. Hamaguchi and by W. Kiefer; for rotational and vibration-rotation resonance Raman
scattering, reviews by L. Ziegler; and for irreducible transition polarizability tensors, a
review by O. S. Mortensen and S. Hassing. Full references will be found in the Section
entitled Further Reading.
The treatments of the classical theory of Rayleigh and Raman scattering, vibrational
Raman scattering and the properties of electromagnetic radiation and many of the Reference Tables are based upon my earlier book, the copyright of which has been assigned
to me by the original publishers.
D. A. Long.
The Raman Effect: A Unified Treatment of the Theory of Raman Scattering by Molecules. Derek A. Long
Part One
THEORY
A unified treatment of the theory of Raman scattering by molecules.
1
Survey of Light-scattering
Phenomena
La lumière est le personnage principal dans le tableau.
Claude Monet
1.1 INTRODUCTION
According to the traditional view of the creation, it was on the first day that light was
separated. Quite soon after, our forebears developed an appetite for knowledge which
their progeny have never lost. However, it is only relatively recently that they discovered
that, if light is reunited fleetingly with matter and then scattered again, it carries with it
detailed information albeit in code about the structure of matter. It is the purpose of
this book to provide the means for deciphering the information carried by the scattered
light.
A number of light-scattering phenomena which provide structural information are now
known and understood. The principal ones and those most used are Rayleigh scattering, Raman scattering, hyper-Rayleigh scattering, hyper-Raman scattering, coherent
anti-Stokes Raman scattering, coherent Stokes Raman scattering, and stimulated Raman
gain or loss spectroscopy. Others, which are less used, include second hyper-Rayleigh
scattering, second hyper-Raman scattering, coherent anti-Stokes hyper-Raman scattering,
coherent Stokes hyper-Raman scattering, and stimulated hyper-Raman spectroscopy. More
than 25 types of Raman spectroscopies are now known!
The Raman Effect
This book is concerned almost entirely with aspects of Rayleigh and Raman scattering.
To set the scene for the detailed developments in subsequent chapters we now give
some basic definitions and present a description of these two processes, together with
a simple model for their interpretation based on exchange of energy between molecules
and the incident radiation. We then describe briefly a few of the other light-scattering
phenomena listed above in order to show their generic relationship to Rayleigh and
Raman scattering.
It should be emphasized that the treatments used in this chapter, although correct as
far as they go, will not enable many of the distinctive properties of the various lightscattering phenomena to be deduced. This requires the more sophisticated methods which
are developed in later chapters.
1.2 SOME BASIC DEFINITIONS

At this point it is expedient to make some general definitions. The expression material
system will be used for the matter involved in the scattering act. This will always be
taken to be an assembly of freely rotating, non-interacting molecules.
The term radiation refers to electromagnetic radiation which will be treated in the main
as classical electromagnetic waves, characterized inter alia by their frequencies or related
quantities. For treatments of spectroscopic theory it is most convenient to use the angular
frequency (unit: rad s1 ), rather than the frequency (unit: Hz). Where we can rely
on the symbol or to indicate unambiguously whether we are using angular frequency
or frequency we employ the name frequency for both and in the text. However,
where spectra are being considered, it is standard practice to use the wavenumber Q ,
where D 2c0 Q and c0 is the speed of light in vacuum. We shall therefore use or Q
as the situation requires. The incident radiation will be taken to consist of one or more
monochromatic electromagnetic waves of frequencies 1 , 2 . . . with photon energies
h 1 , h 2 . If there is only one monochromatic component the interaction is sometimes
referred to as a one-colour process and if there are M distinct monochromatic components,
as an M-colour process. The frequency of the scattered radiation will normally be denoted
by s and the energy of the scattered photon is thus h s .
In the material system the energy of a molecule in its initial state i, before the interaction, is defined as Ei , and in its final state f, after the interaction, as Ef . The energy
of the ground electronic state of the molecule will be defined as Eg and the energy of
an excited electronic state by Ee , or Ee1 , Ee2 . . . if more than one electronic state needs
to be specified. In principle, the initial electronic state need not be the ground electronic state but this will be the case for almost all of the aspects of light scattering we
shall consider. For transitions between energy levels, as for example between the states
i and f, the transition energy is defined as Efi D Ef Ei , and the associated transition frequency as fi D f i so that Efi D h fi . As Ef can be greater or less than
Ei , fi can be positive or negative. It is convenient therefore to introduce M which
Chapter A17 discusses frequency and related quantities in detail.
Survey of Light-scattering Phenomena
is the magnitude of fi that is M D jfi j. The magnitude of a transition energy is

thus h M .
All light-scattering processes are characterized by the fact that, unlike direct absorption
processes, the energy of an incident photon is not required to be equal to the energy
corresponding to the difference between two discrete energy levels of the material system.
It is a matter of experimental observation however, that as the energy of an incident photon
gets closer to an electronic transition energy associated with a transition from the ground
electronic state to an excited electronic state of the material system, the intensity of the
scattering is enhanced. This enhancement increases rapidly as the energy of the incident
photon approaches an electronic transition energy. Such enhanced scattering is called
resonance scattering. The characteristic properties of resonance scattering differ in some
important respects from those of normal scattering.
1.3 RAYLEIGH AND RAMAN SCATTERING

1.3.1 Description
Throughout this Section and also in Sections 1.4, 1.5, 1.6 and 1.7, we use the angular
frequency as the treatment is largely theoretical.
When monochromatic radiation of frequency 1 is incident on systems such as dustfree, transparent gases and liquids, or optically perfect, transparent solids, most of it is
transmitted without change but, in addition, some scattering of the radiation occurs. If
the frequency content of the scattered radiation is analysed, there will be observed to
be present not only the frequency 1 associated with the incident radiation but also, in
general, pairs of new frequencies of the type 1 M (Fig. 1.1(a)). In molecular systems,
the frequencies M are found to lie principally in the ranges associated with transitions
between rotational, vibrational and electronic levels. The scattered radiation usually has
polarization characteristics different from those of the incident radiation. The scattering
occurs over all directions and both the intensity and polarization of the scattered radiation
depend on the direction of observation.
The scattering without change of frequency is called Rayleigh scattering, and that
with change of frequency is called Raman scattering after its discoverer C.V. Raman
(Plate 1.1). In the spectrum of the scattered radiation, the new frequencies are termed
Raman lines, or bands, and collectively are said to constitute a Raman spectrum. Raman
bands at frequencies less than the incident frequency (i.e. of the type 1 M ) are
referred to as Stokes bands, and those at frequencies greater than the incident frequency
(i.e. of the type 1 C M ) as anti-Stokes bands. The origin of this nomenclature is as
Traditionally, the frequency of the incident monochromatic radiation for Rayleigh and Raman scattering
has been denoted by 0 or . For consistency with the general classification of light-scattering phenomena,
we use 1 for all the processes for which M D 1. When M > 1 we use 1 and 2 for processes with M D 2,
and so on.
The history of the discovery of the Raman effect and its early development is described in Section 1.10.
The Raman Effect
Scattering at
w1 (Rayleigh)
and
w1 M (Raman)
(a)
w1 incident
Scattering at
2w1 (hyper-Rayleigh)
and
2w1 wM (hyper-Raman)
(b)
w1 incident
Scattering at
3w1 (second hyper-Rayleigh)
and
3w1 wM (second hyper-Raman)
(c)
w1 incident
Figure 1.1 Diagrammatic representations of: (a) Rayleigh and Raman scattering; (b) hyperRayleigh and hyper-Raman scattering; (c) second hyper-Rayleigh and second hyper-Raman
scattering.
follows. According to Stokes law the frequency of fluorescent light is always smaller
or at most equal to that of the exciting light. Stokes lines in fluorescence are thus those
that correspond to Stokes law, and anti-Stokes lines are those that contradict it. This
nomenclature has also been adopted for the Raman effect, in spite of its difference from
fluorescence.
1.3.2 Energy transfer model

This model is based on the photon description of electromagnetic radiation. We consider
that before the interaction of the radiation with the system there are n1 photons each
of energy h 1 , where 1 is the frequency of the incident monochromatic radiation, and
the molecule has energy Ei . The interaction of the radiation with a molecule leads to
annihilation of one photon of energy h 1 , creation of a new photon of energy h s , and
the transition of the molecule to a state with energy Ef . The radiation now consists of
n1 1 photons of energy h 1 and one photon of energy h s . In the overall process,
energy must be conserved and thus Efi D h 1 s or s D 1 fi . However, as
already emphasized, the energy h 1 does not correspond to any electronic transition energy
and the photon of energy h 1 is not absorbed in the strict spectroscopic sense, as there is
no conservation of energy in this stage of the process. The role of the incident radiation is
rather to perturb the molecule and open the possibility of spectroscopic transitions other
than direct absorption. If h 1 approaches an electronic transition energy, enhancement of
the scattered intensity is observed.
For classification purposes in this and subsequent applications of the energy transfer
model it is convenient to introduce N, defined as the total number of incident photons
involved in a single scattering act, that is the number of incident photons annihilated.
The overall number of photons involved in a single scattering act, that is the number of
photons annihilated and created, is thus N C 1.
The scattering process under consideration involves two photons overall, one incident and
one scattered, so that N C 1 D 2. It describes Rayleigh scattering when f D i and Raman
scattering when f 6D i. For Rayleigh scattering s D 1 . For Raman scattering if Ef > Ei ,
then s D 1 M and we have Stokes Raman scattering; if Ef < Ei , then s D 1 C M
and we have anti-Stokes Raman scattering. Rayleigh scattering is sometimes described as
elastic scattering and Raman scattering as inelastic scattering. Also, as Raman scattering
involves a net exchange of energy between the radiation and the molecule, it is often classified
as an active process.
The Rayleigh and Raman scattering processes are summarized in Table 1.1. This table
provides the following information for each scattering process: the values of N and M
in column (i); the initial and final energy states of the molecule in columns (iii) and (v),
respectively; the initial and final numbers of photons in columns (ii) and (iv) respectively;
the scattered frequency s determined by the requirement of overall energy conservation
in column (vi); an energy level diagram in column (vii); and the name of the process in
column (viii). In the energy level diagram an arrow pointing upwards indicates photon
annihilation and an arrow pointing downwards indicates photon creation. Unbroken horizontal lines denote discrete quantum states of the molecule and broken lines indicate
virtual states. In a real absorption process energy is conserved and the resulting state of
the system is a discrete state. Absorption without energy conservation is called virtual
absorption and the resulting state is described as a virtual state. At this early stage in
the development, the energy level diagrams are to be regarded simply as a diagrammatic
representation of photon annihilation and creation, the energies of the photons and the
energy levels of the molecule that are implicated in the scattering process. The diagrams
(ii)
n1h 1
n1h 1
n1h 1
ND1
MD1
ND1
MD1
ND2
MD1
Ei
Ei
Ei
(iii)
Molecule
Initial state
energy
Photons
(i)
N and M
n1 2h1 C h s
n1 1h1 C h s
n1 1h1 C h s
(iv)
Photons
Final state
energy
Ef D Ei
Ef D Ei C h fi
Ef D Ei
(v)
Molecule
s D 21
s D 1 fi
s D 1
Scattered
frequency s
determined by
overall energy
conservation
(vi)
w1
w1
w1
w1
ws = w1
i, f
ws = 2w1
wfi
i, f
ws = w1 wfi
(vii)
Energy level diagram
Hyper-Rayleigh
scattering
Raman
scattering
(Stokes)
Rayleigh
scattering
(viii)
Name
Table 1.1 Energy balance sheet for Rayleigh, Raman, hyper-Rayleigh, hyper-Raman, second hyper-Rayleigh and second hyper-Raman
processes
8
The Raman Effect
(ii)
n1h 1
n1h 1
n1h 1
ND2
MD1
ND3
MD1
ND3
MD1
Ei
Ei
Ei
(iii)
Molecule
Initial state
energy
Photons
(i)
N and M
Table 1.1 (Continued)
n1 3h1 C h s
n1 3h1 C h s
n1 2h1 C h s
(iv)
Photons
Final state
energy
Ef D Ei C h fi
Ef D Ei
Ef D Ei C h fi
(v)
Molecule
s D 31 fi
s D 31
s D 21 fi
Scattered
frequency s
determined by
overall energy
conservation
(vi)
w1
f
wfi
w1
i, f
ws = 3w1 wfi
ws = 3w1
wfi
ws = 2w1 wfi
w1
w1
w1
w1
w1
w1
(vii)
Energy level diagram
Second
hyper-Raman
scattering
(Stokes)
Second
hyper-Rayleigh
scattering
Hyper-Raman
scattering
(Stokes)
(viii)
Name

9
10
The Raman Effect
say nothing about the mechanisms of the photonmolecule interactions or the probability
of their occurrence.
Raman scattering is inherently incoherent and as a result the intensity of scattering
from a material system of N non-interacting molecules is simply N times that from one
molecule and is independent of the bulk structure of the material system.
In contrast, Rayleigh scattering from a material system depends on the nature and
structure of the system as well as the concentration of the scattering species. As a result
Rayleigh scattering can be coherent in some situations and incoherent in others. However
the intensity of Rayleigh scattering from an assembly of N freely rotating, non-interacting
molecules is N times that from a single molecule except when the scattering is observed
along exactly the same direction as that in which the incident radiation is propagated. If
we exclude this particular case, which is termed forward scattering, we can treat Rayleigh
scattering as incoherent. This will be our normal procedure henceforth.
The intensity of Rayleigh and Raman scattering is directly proportional to the irradiance
of the incident radiation and so such scattering can be described as a linear process.
1.4 HYPER-RAYLEIGH AND HYPER-RAMAN SCATTERING

1.4.1 Description
When a system is illuminated with monochromatic radiation of frequency which has
an adequately large irradiance the scattered radiation is now found to include additional
frequencies of the type 21 and 21 M . The scattering with frequency 21 is called
hyper-Rayleigh scattering and that with frequencies 21 M is called hyper-Raman scattering (Fig. 1.1(b)). This scattering also occurs over all directions and both the intensity
and polarization of the scattered radiation depend on the direction of observation.

Here again the model is based on the photon description of electromagnetic radiation.
We consider that before the interaction there are n1 photons each of energy h 1 and
the molecule has energy Ei . The interaction of the radiation with a molecule leads to
the successive annihilation of two photons each of energy h 1 , the creation of a new
photon of energy h s and the transition of the molecule to a state with energy Ef . The
radiation now consists of n1 2 photons each of energy h 1 and one photon of energy
h s . Overall energy conservation requires that Efi D h 21 s or s D 21 fi , but
neither h 1 nor 2h1 need correspond to an electronic transition energy of the molecule.
This process, which involves three photons overall N C 1 D 3 describes hyper-Rayleigh
scattering when f D i and hyper-Raman scattering when f 6D i. For hyper-Rayleigh scattering s D 21 . For hyper-Raman scattering, if Ef > Ei then s D 21 M and this
is termed Stokes hyper-Raman scattering; if Ef < Ei , then s D 21 C M and this is
Chapter A17 deals with irradiance.
11
termed anti-Stokes hyper-Raman scattering (see Table 1.1). Resonance enhancement can
arise if either h 1 or h 2 approaches an electronic transition energy of the molecule.
Hyper-Raman scattering is incoherent as is non-forward hyper-Rayleigh scattering from
a gas and the intensity of scattering is proportional to the number of scattering molecules.
Hyper-Raman scattering is also an inelastic or active process. However, it should be
noted that the intensity of hyper-Rayleigh and hyper-Raman scattering is proportional to
the square of the irradiance of the incident radiation of frequency 1 . In consequence,
such scattering is termed a non-linear process.
1.5 SECOND HYPER-RAYLEIGH AND SECOND HYPER-RAMAN

SCATTERING
1.5.1 Description
When a system is illuminated with monochromatic radiation of frequency 1 which has
adequately large irradiance, the scattered radiation also includes frequencies of the type
31 and 31 M corresponding to second hyper-Rayleigh and second hyper-Raman
scattering, respectively (Fig. 1.1(c)).

Here the interaction of the radiation with a molecule involves the successive annihilation of three photons, each of energy h 1 and should be readily understood from
Table 1.1. This process, which involves four photons overall N C 1 D 4, describes
second hyper-Rayleigh scattering when f D i and second hyper-Raman scattering when
f 6D i. For second hyper-Rayleigh scattering s D 31 . For second hyper-Raman scattering, if Ef > Ei then s D 31 M (Stokes) and, if Ef < Ei then s D 31 C M
(anti-Stokes). Resonance enhancement can occur if either h 1 , 2h1 or 3h1 approaches
an electronic transition energy of the molecule.
Second hyper-Raman scattering is incoherent as is non-forward second hyper-Rayleigh
scattering from a gas and the intensity of scattering is proportional to the number of
scattering molecules. Second hyper-Raman scattering is also an inelastic or active process.
It should be noted that second hyper-Rayleigh and second hyper-Raman scattering are
non-linear processes, the scattered intensity being proportional to the cube of the irradiance
of the incident radiation of frequency 1 .
1.6 COHERENT ANTI-STOKES AND COHERENT STOKES

RAMAN SCATTERING
In all the light-scattering processes treated so far, the incident radiation has consisted of
one monochromatic wave of frequency 1 . We now consider the experimental situation
illustrated in Fig. 1.2 where the incident radiation consists of two overlapping coherent
monochromatic beams of frequencies 1 and 2 , with 1 > 2 . As the overlapping
12
The Raman Effect

w2 wM (CSRS)
w1
w2 gain (SRGS)
w2 loss (SRLS)
w2
Lens
w1 > w2
Material system
w1 w2 = wM
w1+ wM (CARS)
Figure 1.2 Diagrammatic representation of CARS, CSRS, SRGS and SRLS.
beams of radiation propagate through the material system, new radiation is produced
with frequencies corresponding to various combinations of 1 and 2 . From amongst
the possible combinations we first consider the combination 21 2 . If we vary 2
while keeping 1 constant we find that the intensity of the scattering increases dramatically when 1 2 D M , where M is a molecular frequency that can be observed
in Raman scattering. When this frequency-matching condition is satisfied s D 1 C M ,
because s D 21 2 D 1 C 1 2 D 1 C M . The condition 1 2 D M can
be regarded as a Raman resonance. This is quite different from the electronic resonances
described earlier. The scattered frequency 1 C M has the form of an anti-Stokes Raman
frequency relative to 1 . As this scattered radiation is coherent, it is called Coherent antiStokes Raman Scattering, or CARS. By varying 2 over a range of values that covers
the desired values of M a CARS spectrum can be obtained. The CARS frequencies will
be superimposed on a background of weak non-resonant scattering given by 21 2 .
If alternatively we consider the scattered frequency combination 22 1 , then when
1 2 D M strong scattering now occurs at 2 C 2 1 D 2 M . This is Stokes
radiation relative to 2 and is called Coherent Stokes Raman Scattering or CSRS.
CARS and CSRS differ in many important respects from the Raman processes considered so far in Sections 1.31.5. Their coherent nature has already been emphasized. They
produce highly directional beams of scattered radiation with small divergences. The scattered intensity is proportional (a) to the square of the number of scattering molecules and
(b) to the square of the irradiance of the incident radiation at 1 and to the irradiance
of the incident radiation at 2 . CARS and CSRS are thus non-linear processes. Electronic resonance enhancement of the scattered intensity can also arise if either 1 or 2
approaches an electronic transition energy.
The theoretical treatment of CARS and CSRS involves concepts that lie outside the scope
of this book. Consideration of the interaction of the waves of frequencies 1 and 2 involves
the bulk or macroscopic properties of the material system which must then be related to the
individual or microscopic properties of the molecules. From such considerations emerge
the special properties of CARS and CSRS radiation which we have just outlined.
Here we just concern ourselves with energy considerations taking CARS as an example.
The process involved is the annihilation of one photon of energy h 1 , creation of a photon
of energy h 2 , annihilation of a second photon of energy h 1 and scattering of a second
photon of energy h s D h 1 C M . It is readily seen that the photon energies are selfbalancing as h 1 C 2 1 C s D 0 since s D 21 2 . Thus there is no net
w1
13
w2
w1
ws = w1 + wM
wM
(a) CARS
w2
w1
w2
ws = w2 wM
wM
(b) CSRS
Figure 1.3 Energy level diagrams for (a) CARS (b) CSRS. Both CARS and CSRS are
passive processes with no net energy change in the material system.
energy change in the material system and the process is said to be passive or parametric.
The material system acts as a facilitating agent as it were for the exchange of energy
between radiation of different frequencies and this is very effective when 1 2 D M .
CSRS is also of course a passive or parametric process. Energy level diagrams for CARS
and CSRS and given in Figs 1.3(a) and (b) respectively.
1.7 STIMULATED RAMAN GAIN AND LOSS SPECTROSCOPY

We now consider another interaction that can arise when there is present in the material
system radiation of frequencies 1 and 2 with 1 > 2 and 1 2 D M . Suppose
that before the interaction there are n1 photons each of energy h 1 and n2 photons
each of energy h 2 and the molecules have energy Ei . The interaction of the radiation with the material system can involve the annihilation of a photon of energy h 1
and creation of a photon of energy h 2 , so that the radiation now consists of n1 1
photons each of energy h 1 and n2 C 1 photons each of energy h 2 . The scattered radiation now has a frequency of s D 2 D 1 M and energy conservation requires that
14
The Raman Effect
Ef D Ei C h 1 2 D Ei C h M . Thus the overall process is a Stokes Raman process

relative to 1 because radiation at 1 M is produced but, unlike the Stokes Raman
process described in Section 1.4, this process is taking place in the presence of radiation of frequency 2 (that is 1 M ), the frequency of the Stokes Raman radiation
itself. This process is termed a stimulated one, the presence of radiation of frequency
2 D 1 M being said to stimulate the Stokes Raman process and produce a gain in
intensity of the radiation of frequency 2 . This gain can be exponential and lead to very
substantial transfer of energy from radiation of frequency 1 to that of frequency 2 and
consequently substantial population of the final state f.
It should be noted that as n2 tends to zero the stimulated Raman gain process tends
towards the normal Raman process. Consequently the theoretical treatment of the stimulated Raman effect can embrace the normal Raman process by treating the latter as a
limiting case.
The interaction of the radiation with the material system can also result in the creation
of a photon of energy h 1 and annihilation of a photon of energy h 2 , so that the radiation
now consists of n1 C 1 photons each of energy h 1 and n2 1 photons each of energy
h 2 . The scattered radiation now has a frequency of s D 1 and energy conservation
requires that Ef D Ei h 1 2 D Ei h M . The overall process represents a loss of
intensity at the Stokes Raman frequency 2 and is termed a stimulated Raman loss process.
In each of these processes, there is necessarily a compensating change in intensity at
the frequency 1 . Thus stimulated Raman gain at 2 involves a corresponding loss at
1 and stimulated Raman loss at 2 involves a corresponding gain at 1 . Stimulated
Raman scattering is also a coherent process. Its intensity is proportional to the number
of molecules involved and to the product of the irradiances of the two incident beams of
frequencies 1 and 2 .
The acronyms SRGS and SRLS (Fig. 1.2) are often used for stimulated Raman gain
spectroscopy and stimulated Raman loss spectroscopy, respectively. They are both nonlinear processes.
1.8 TYPICAL SPECTRA

The various light-scattering processes have so far been discussed in terms of the angular
frequency 1 . This is very convenient for theoretical treatments and will continue
to be used in subsequent chapters for such developments. However, when spectra are
being considered it is standard procedure in Raman spectroscopy (and elsewhere) to use
wavenumber Q rather than frequency and this practice will be followed throughout this
book. Furthermore, the results of the various theoretical treatments based on use of will
be re-expressed in terms of Q so that they relate directly to practical spectroscopic usage.
The spectra associated with a selection of light-scattering processes are displayed in
Fig. 1.4 using a wavenumber scale (unit: cm1 ). All the spectra illustrated relate to the
case of one participating molecular transition wavenumber Q M D M /2c0 D 1000 cm1 .
Chapter A17 discusses frequency and wavenumber in detail.
15
(2c0)1(w1 wM) (2c0)1w1 (2c0)1(w1+ wM)

Rayleigh
and Raman scattering
(Stokes and anti-Stokes)
19 000
(2c0)1w1
20 000
21 000
~
n/cm1
(2c0)1(2w1 wM) (2c0)12w1 (2c0)1(2w1 + wM)

Hyper-Rayleigh
and hyper-Raman scattering
(Stokes and anti-Stokes)
10000
19 000
20 000
21 000
~n/cm1
(2c0)1w2 (2c0)1w1 (2c0)1(w1+ wM)
19 000
20 000
21 000
~
n /cm1
Coherent anti-Stokes
Raman scattering (CARS)
ws = w1 + wM
when w1 w2 = wM
Figure 1.4 Diagrammatic spectra for selected light-scattering processes. O M D M /2c0 D

1000 cm1 throughout.
The various wavenumbers in each spectrum in Fig. 1.4 are denoted by vertical lines, the
position of each line being related to the value of Q /cm1 it represents. Each line is also
given an label to indicate its origin. In the treatment given in preceding sections all
the energy levels have been regarded as exactly defined so that a transition wavenumber
is also exactly defined and can be properly described as a line in a spectrum. In reality
this is certainly not the case. An excited state does not have an infinite lifetime so that its
energy is not exactly defined. There is consequently an uncertainty E in its energy so
that a transition is associated with a distribution of wavenumbers about a central value Q .
This so-called natural line broadening is but one of the many factors that can contribute to
intrinsic line shapes; and even under the most favourable conditions the instrumentation
used to obtain a spectrum will itself further modify the line shape.
The use of the word line in relation to a spectrum was a natural descriptive label
when spectra were observed visually in a spectroscope or recorded on a photographic
plate. It continues to be widely used in many forms of spectroscopy, and terms such as
line shape and line width are well established in the literature.
In this book we shall use line when the observed wavenumber arises from a single
transition, as for example a rotational or a rovibrational transition in a Raman spectrum or
an atomic transition associated with lasers or a mercury arc source. However, in the case of
Raman spectra of liquids where vibrational transitions are associated with unresolved rotational structure the use of band and related terms such as band width and band shape
is more appropriate. Similar considerations apply to Rayleigh scattering from liquids.
It might appear that the various light-scattering phenomena that we have considered
simply provide alternative methods of observing Q M . If this were the case only Rayleigh
Chapter 3, Section 3.9 discusses the presentation of experimentally observed Rayleigh and Raman spectra.
16
The Raman Effect
and Raman scattering would normally find practical application because the equipment
necessary for other types of light scattering is considerably more expensive. However,
knowledge of the theory underlying these phenomena will reveal that each process has its
own distinctive contribution to make. For example in some polyatomic molecules not all
their various transition frequencies are accessible using a single light-scattering process.
Also, each process can provide unique structural information relating to the scattering
molecule. In addition some processes can offer higher spectral resolution or experimental
convenience in particular situations, as for example in remote sensing. Taken together,
the various light-scattering processes constitute a formidable array of tools for use in a
great variety of scientific problems, both pure and applied.
1.9 BASES FOR THE THEORETICAL TREATMENT OF

RAYLEIGH AND RAMAN SCATTERING
The theoretical treatment of Rayleigh and Raman scattering may be approached in several
ways. One method is the purely classical one in which the radiation is treated as an
electromagnetic wave, and the material system as an assembly of independent classical
rotors and vibrators. This can explain satisfactorily some aspects of light scattering, both
linear and non-linear, particularly as regards the frequency dependence. But, of necessity,
this method is silent on how the molecular properties of the material system are involved.
However, if we adopt a quantum mechanical treatment of the material system but retain
a classical treatment of the radiation, all the properties of the light-scattering phenomena
we wish to consider can be treated quite satisfactorily. For brevity we shall refer to this
hybrid procedure as the quantum mechanical method. Yet another method is to quantize
both the radiation and the material system. This procedure, which is the basis of quantum
electrodynamics, we have used in a basic form in the classification of light-scattering
phenomena given earlier. However, its more detailed application need not concern us as
it yields the same results as the quantum mechanical theory for Rayleigh and Raman
scattering.
We shall use the quantum mechanical method for the unified detailed treatment of
Rayleigh and Raman scattering that constitutes most of this book (Chapters 410), but
we feel that there is some merit in including an account of the classical treatment of these
phenomena (Chapter 3). An introduction to both these theoretical treatments is given in
Chapter 2. It is possible to construct a completely unified treatment of all the various
Raman spectroscopies. However, the sophistication of such an approach is such that only
those already well versed in the subject are likely to find it useful. The words of the
French poet Valery come to mind: il y a deux dangers mortel pour lhomme, lordre et
le desordre. I hope the treatment in this book avoids these two dangers.
1.10 HISTORICAL PERSPECTIVE

Curiosity about the explanation of the blue colour of the sky led Lord Rayleigh to formulate a classical theory of light scattering without change of frequency (Rayleigh, 1871).
17
Fascination with the marvellous blue of the Mediterranean sea caused C. V. Raman to
investigate the scattering of light by liquids and so to discover experimentally the scattering of light with change of frequency (Raman and Krishnan, 1928). An independent
prediction of this phenomenon had been made a few years earlier (Smekal, 1923) using
classical quantum theory.
Shortly after Raman and Krishnans publication a report of light scattering with change
of frequency in quartz was reported by two Russian scientists, Landsberg and Mandelstam
(1928), and in France, Raman and Krishnans observations were soon confirmed by
Cabannes (1928) and Rocard (1928). The potential of the Raman effect in chemistry
and physics was realized very rapidly. By the end of 1928 some 70 papers on the Raman
effect had been published. One of these (Wood, 1928) was written in telegraphese and
sent to Nature where it was published as received. The early history of the Raman effect
has been described in some detail by Long (1988).
Light-scattering phenomena with N 2, were not discovered until laser sources became
available in the 1960s. Only with appropriate lasers is it possible to arrange for two or more
photons of the same or different energies to be incident essentially simultaneously on the
scattering system; and only with lasers is it possible to tune the frequency of the incident
radiation to achieve the conditions necessary for some of the higher-order processes such
as CARS. However some higher-order processes were predicted on theoretical grounds
as early as the 1930s by, for example, Goeppert-Mayer (1931) and Placzek (1934).
1.11 CAVEAT
Scattering of radiation without a change of frequency, that is elastic scattering, can also
arise from larger scattering centres such as dust particles. Such processes are called Mie
or Tyndall scattering, according to the size of the scattering particle and its refractive
index relative to the scattering medium.
There are also processes other than those considered so far that can produce light quanta
with energies different from that of the incident quantum. One example is Brillouin
scattering which arises from the interaction with acoustic waves in crystals. There are
other processes which occur with a time delay and are not scattering processes. Examples
of these are fluorescence and phosphorescence. Processes of the types just described fall
outside our considerations.
REFERENCES
Brillouin, L. (1922). Ann. Phys. (Paris) 88, 17.
Cabannes, J. (1928). Compt. Rend. 186, 1201.
Goeppert-Mayer, M. (1931). Ann. d. Phys. 9, 273.
Landsberg, G. and Mandelstam, L. (1928). Naturwiss. 16, 557, 772.
Long, D. A. (1988). Int. Rev. Phys. Chem. 7, 317.
Placzek, G. (1934). Rayleigh-Streuung und Raman-Effekt, in Handbuch der Radiologie, E. Marx
(ed.), 6, 205374. Academische Verlag: Leipzig.
18
Raman, C. V. and Krishnan, K. S. (1928). Nature 121, 501.
Rocard, Y. (1928). Compt. rend. 186, 1107.
Rayleigh, Lord (1871). Phil. Mag. 41, 274, 447.
Smekal, A. (1923). Naturwiss. 16, 873.
Wood, R. W. (1928). Nature 122, 349.
The Raman Effect
19
2
Introduction to Theoretical
Treatments of Incoherent Light
Scattering
Begin at the beginning, the King said, gravely,
and go on till you come to the end: then stop.
Lewis Carroll
2.1 GENERAL CONSIDERATIONS

A general incoherent light-scattering experiment is illustrated in Fig. 2.1. Monochromatic
radiation from a laser is incident on a material system. This radiation propagates as an
essentially non-divergent beam. We treat it as a plane monochromatic wave which is
completely characterized by its frequency 1 , its state of polarization pi , its propagation
vector n0i and its irradiance I . As a result of the interaction of the incident radiation
with the material system some scattered radiation is produced. This scattered radiation
consists of a number of different frequency components each of which, in general, will
be propagating in various directions in space. We shall also treat these components as
plane monochromatic waves. A component of the scattered radiation of frequency s is
similarly completely characterized by its state of polarization ps , its propagation vector
n0s and its intensity I.
Chapter A17 treats the properties of plane monochromatic waves, including propagation vector and intensity and irradiance of radiation. For the material system considered here (see page 4) we can treat the incident
and scattered radiation as propagating in a vacuum.
20
The Raman Effect

z
Scattered radiation, a
frequency component of
which is characterized by:
frequency ws (wavenumber ns)
polarization p s
wave vector n0s
intensity I
Material
system
O
y
x
Incident monochromatic
radiation, characterized by:
frequency w1(wavenumber n1)
polarization pi
wave vector n0i
irradiance
How do the properties

of the material system
and the illumination
observation geometry
determine the relation
between these two sets
of radiation parameters?
Figure 2.1 A general incoherent light-scattering experiment. The incident radiation propagates along some direction, say Oz. The scattered radiation propagates along some direction
Ov which has a general orientation relative to Oz.
This description of a general incoherent light scattering experiment has taken the incident radiation to originate from a laser. Of course incoherent linear light scattering experiments can be performed using other monochromatic light sources. Indeed, for decades
prior to the development of laser sources in the 1960s, Rayleigh scattering and Raman
light scattering were studied using traditional light sources, particularly high-power watercooled mercury arcs. However, lasers have unique properties that enable the potential of
Rayleigh and Raman scattering to be fully exploited and they are now invariably chosen
for incoherent light-scattering experiments. For non-linear light-scattering experiments,
either incoherent or coherent, lasers are of course essential.
The objectives of the theoretical treatments of incoherent light scattering which we are
going to present are to establish quantitatively the nature of the relationships between I
and I and between 1 and s and how they depend on the molecular properties of the material system. When such relationships are known, light-scattering experiments in which the
characteristics of the scattered radiation are measured as a function of the characteristics
of the scattered radiation and the illuminationobservation geometry can yield a great deal
of information about the properties of molecules which constitute the material system.
This chapter outlines the bases of the principal methods we shall use for treating
linear and non-linear incoherent light-scattering phenomena, namely classical theory and
21
quantum mechanical theory. The intention is to provide a guide to the more detailed
treatments in subsequent chapters.
These theoretical treatments of light scattering are all based on a microscopic approach
in which, initially, the scattering from an individual molecule is considered. For almost
all the cases we shall consider, the scattering is incoherent. The intensity of the scattering
from a material system of N non-interacting molecules is then simply N times that from
one molecule and is independent of the bulk structure of the material. In a few cases
light scattering can be coherent in some situations and incoherent in others. For coherent
scattering the intensity then depends on the nature and structure of the material system
as well as the molecular concentration.
2.2 INDUCED OSCILLATING ELECTRIC DIPOLES AS SOURCES

OF SCATTERED RADIATION
In both the classical and quantum mechanical treatments used here the origin of scattered
radiation is considered to be the oscillating electric and magnetic multipole moments
induced in a molecule by the electromagnetic fields of the incident light waves . Normally,
the most significant multipole source is the oscillating electric dipole. The oscillating
magnetic dipole and electric quadrupole are the next most important sources but their
contributions are, typically, several orders of magnitude smaller than that of the oscillating
electric dipole. Thus initially in this chapter we shall confine ourselves to the oscillating
induced electric dipole as the source of scattered radiation. Those aspects of incoherent
light scattering phenomena for which oscillating magnetic dipole and electric quadrupole
contributions must be taken into account are considered in Section 2.5.
The intensity I, or time-averaged power per unit solid angle, radiated by an oscillating
electric dipole induced in a molecule by the electric field of the incident radiation of
frequency 1 , along a direction making an angle with the axis of the dipole is given
by
2.2.1
I D k0 s4 p02 sin2
where
k0 D
1
322 0 c03
2.2.2
and p0 is the amplitude of the induced electric dipole with frequency s which is generally
but not necessarily different from 1 . Here, and in much of the subsequent theoretical
treatments it is convenient to use the frequency . However in some situations, particularly those where the emphasis is on the position of bands in a spectrum rather than their
intensities, we shall use the wavenumber Q in accord with normal practice. Therefore
Chapter A13 treats induced multipole moments.

Chapter A20 treats the oscillating electric dipole.
Chapter A17 discusses frequency and wavenumber in detail.
22
The Raman Effect
for future reference we give the following alternative forms of eqs (2.2.1) and (2.2.2),
namely
I D k0Q Q s4 p02 sin2
2.2.3
where
k0Q D
2 c0
20
2.2.4
These alternative forms are obtained from eqs. (2.2.1) and (2.2.2) by introducing the
relation
2.2.5
s D 2c0 Q s
The objectives of both the classical and quantum mechanical treatments of light scattering are to find how s (or Q s ) and p0 are determined by the properties of the scattering
molecule and the incident electromagnetic radiation of frequency 1 (or wavenumber Q 1 ).
In preparation for the detailed treatments in subsequent chapters we now consider in
outline how the classical and quantum mechanical treatments approach this problem.
2.3 BASIS OF THE CLASSICAL THEORY OF

LIGHT SCATTERING
We now consider the nature of the relationship between the induced electric dipole moment
of a molecule and the dynamic electric field which produces it. This dynamic electric
field is characterized by E, the time-dependent electric vector of the incident radiation
which, for the incoherent phenomena we are treating, is a monochromatic plane wave
of frequency 1 . The total time-dependent induced electric dipole moment vector of a
molecule may be written as the sum of a series of time-dependent induced electric dipole
moment vectors p 1 , p 2 , p 3 . . . as follows:
p D p 1 C p 2 C p 3 C . . .
2.3.1
As p 1 p 2 p 3 this is a rapidly converging series.

The relationships between p 1 , p 2 , p 3 and E have the following forms :
p 1 D a E
p
2
p 3 D
1
2b
1
6g
: EE
... EEE
2.3.2
2.3.3
2.3.4
Thus p 1 is linear in E and p 2 and p 3 are non-linear in E, with p 2 quadratic and p 3
cubic in E. In eqs (2.3.2) to (2.3.4) the relationships between p 1 , p 2 , p 3 and E involve
tensors ; a is the polarizability tensor, a second-rank tensor ; b is the hyperpolarizability
Chapter A13 treats induced electric dipole moments.

Chapter A10 treats the properties of tensors generally.
Chapter A14 treats the polarizability tensor specifically.
23
tensor, a third-rank tensor; and g is the second hyperpolarizability tensor, a fourth-rank

tensor. These tensors will be time-dependent if, for example, a molecular vibration is
involved.
It will prove useful to write eqs (2.3.2) to (2.3.4) in component form, as follows:
p1
D E
p2

p3

D
D
2.3.5
1
2
E E
1
6
E E
E
2.3.6
2.3.7
where , ,
, can be x, y or z and summation over repeated Greek subscripts is implied.
The time dependence of p 1 , p 2 or p 3 is determined by the time dependence of the
tensor involved and also that of the appropriate term in E, EE or EEE, respectively. In
general p 1 , p 2 or p 3 will have a number of frequency components.
For p 1 which is linear in E there can be three frequency components, one associated
with Rayleigh scattering with frequency 1 , and two associated with Raman scattering
at frequencies 1 M where M is a molecular frequency. As we have seen from
eq. (2.2.1) the scattered intensity at a particular frequency is determined by the amplitude
of the induced electric dipole at that frequency. Thus we may adapt eq. (2.3.2) and write
for Rayleigh and Raman scattering
p0 1 D aRay E0 1
Ram
p0 1 M D a
E0 1
2.3.8
2.3.9
where p0 1 and p0 1 M denote time-independent amplitudes associated with the

induced electric dipoles oscillating at the frequencies given in brackets; E0 1 denotes the
time-independent amplitude of the electric field of the incident radiation of frequency 1 ;
aRay is an equilibrium polarizability tensor and aRam is a polarizability tensor associated
with M . The component forms of eqs (2.3.8) and (2.3.9) are
p0 1 D Ray
E 0 1
2.3.10
p0 1 M D Ram
E 0 1
2.3.11
and
For p 2 , which is quadratic in E, its frequency components can include one associated
with hyper-Rayleigh scattering with frequency 21 and two associated with hyper-Raman
scattering with frequencies 21 M . Using time-independent amplitudes and adapting
eq. (2.3.3) we may write for hyper-Rayleigh and hyper-Raman scattering
p0 21 D 12 bHRay : E0 1 E0 1
2.3.12
p0 21 M D 12 bHRam : E0 1 E0 1
2.3.13
and
where bHRay is an equilibrium hyperpolarizability tensor and bHRam is a hyperpolarizability
tensor associated with M .
24
The Raman Effect
For p 3 which is cubic in E its frequency components can include one associated with
second hyper-Rayleigh scattering with frequency 31 ; and two associated with second
hyper-Raman scattering with frequencies 31 M . Using time-independent amplitudes
and adapting eq. (2.3.4) we may write for second hyper-Rayleigh and second hyperRaman scattering
p 3 D 1 gHHRay ... E E E
2.3.14
0
p0 31 M D
6
1 HHRam
6g
... E E E
0 1 0 1 0 1
2.3.15
where gHHRay is an equilibrium second hyperpolarizability tensor and gHHRam is a second

hyperpolarizability tensor associated with M .
The intensities of Rayleigh and Raman scattering from a single molecule can be
obtained by substituting into eq. (2.2.1) the squares of the appropriate components of
the classical electric dipole amplitudes given by eqs. (2.3.10) or (2.3.11). It follows that
the intensities of Rayleigh and Raman scattering are quadratic in components of E0 1
and so proportional to the irradiance I of the incident radiation. Similar considerations
involving eq. (2.2.1) with eqs. (2.3.12) or (2.3.13) show that the intensities of hyperRayleigh and hyper-Raman scattering is quartic in the electric field components and so
proportional to I 2 . Likewise, using eqs. (2.3.14) and (2.3.15), the intensity of second
hyper-Rayleigh and second-hyper Raman scattering can be shown to be proportional to
I 3 . In all cases geometrical factors are determined by the tensor properties of a, b or g
and the factor sin2 in eq. (2.2.1).
The procedure for a classical treatment of light scattering is now clear. Treating both
the radiation and the molecule classically, we have simply to calculate the frequency and
amplitude of the induced electric dipole of the appropriate order, and substitute these in
eq. (2.2.1) or eq. (2.2.3). For linear processes the frequency is obtained by combining
the frequency dependence of the polarizability tensor with the frequency dependence
of the incident electric field. For the non-linear processes hyper-Rayleigh and hyperRaman scattering, it is necessary to select the appropriate frequency component of the
quadratic incident electric field term and combine this with the frequency dependence of
the hyperpolarizability tensor and similarly for higher-order processes.
2.4 BASIS OF THE QUANTUM MECHANICAL TREATMENT OF

INCOHERENT LIGHT-SCATTERING PHENOMENA:
ELECTRIC DIPOLE CASE
In the quantum mechanical treatment of light-scattering phenomena, we treat the interacting molecule in the material system quantum mechanically but continue to treat the
electromagnetic radiation classically. We regard the electric and magnetic fields associated with the incident electromagnetic radiation as producing perturbations of the states
of the molecule, and use the methods of quantum mechanics to calculate the properties of
the perturbed system. We shall be concerned in particular with the permitted transitions
between states of the molecule when it is under the perturbing influence of the incident
25
radiation and the frequency-dependent multipole transition moments associated with such
transitions. The properties of the scattered radiation can then be determined by regarding
such frequency-dependent multipole transition moments as classical multipole sources of
electromagnetic radiation.
As in the case of the classical treatment we confine ourselves initially to the frequencydependent electric dipole transition moments, and defer consideration of frequencydependent magnetic dipole and electric quadrupole transition moments to Section 2.5.
In the quantum mechanical treatment the induced electric dipole of classical theory is
replaced by the transition electric dipole associated with a transition in the molecule from
an initial state, i, to a final state, f, which has been induced by the incident electric field
of frequency 1 .
Thus we write for the total induced transition electric dipole vector, the following
equation, which is the analogue of eq. (2.3.1) for the classical case:
pfi D p 1 fi C p 2 fi C p 3 fi C . . .
2.4.1
where p 1 fi is linear in E, p 2 fi , is quadratic in E, p 3 fi is cubic in E and so on.
The total transition electric dipole pfi is given by
pfi D h f0 jpj
O i0 i
2.4.2
where i0 and f0 are the time-dependent perturbed wave functions of the initial and final
states of the molecule, respectively, and pO is the electric dipole moment operator.
The perturbed time-dependent wave functions i0 and f0 may be represented formally
by the series expansions
and
i0 D i0 C i1 C i2 C C in
2.4.3
f0 D f0 C f1 C f2 C C fn
2.4.4
where i0 represents the unperturbed state, i1 is the first-order modification to i0
as a result of the perturbation, i2 the second-order modification to i0 and so on, and
similarly for the final state.
Time-dependent perturbation theory enables the perturbed wave functions to be expressed as linear combinations of the unperturbed wave functions. When the interaction
O involves only an electric dipole term, which we designate as H
O p , we
hamiltonian H
may write

ap1
r0
2.4.5
i1 D
ir
r
i2
ap2
r0
ir
2.4.6
apn
r0
ir
2.4.7
..
.
in D

r
26
The Raman Effect
and similarly for the final state, the general relation being

apn
0
fn D
fr r
2.4.8
In all cases the summations are over all states of the system. The apir and apfr are
coefficients, and the subscript p is introduced to indicate that they are generated by the
O p . The ap1 and ap1 are linear in E, the ap2 and
electric dipole interaction hamiltonian H
ir
fr
ir
2
apfr quadratic in E and so on. This is because the perturbation operator acts once to
give i1 or f1 , twice to give i2 or f2 and so on.
We now introduce the series expansions for i0 and f0 given by eqs. (2.4.3) and (2.4.4)
into eq. (2.4.2), collect the terms according to their dependence on the perturbing electric
field strength E and correlate these with the corresponding terms in eq. (2.4.1). For the
first three terms in eq. (2.4.1) we find that
O i1 i C h f1 jpj
O i0 i
p 1 fi D h f0 jpj
2.4.9
O i2 i C h f2 jpj

O i0 i C h f1 jpj
O i1 i
p 2 fi D h f0 jpj
2.4.10
p 3 fi D h f0 jpj

O i3 i C h f3 jpj
O f0 i C h f1 jpj
O i2 i C h f2 jpj
O i1 i
2.4.11
A further transition moment which we designate as p 0 fi also arises and is given by
p 0 fi D h f0 jpj
O i0 i
2.4.12
This transition moment is independent of E as it involves only unperturbed wave functions.

It relates to a direct transition between unperturbed initial and final states and not to light
scattering and so need not be considered further here.
The two terms in eq. (2.4.9) for p 1 fi , each involve only one first-order perturbed
wave function, either i1 or f1 . Thus each term is linearly dependent on coefficients
or ap1
and so p 1 fi is linear in E.
of the type ap1
ir
fr
The three terms in eq. (2.4.10) for p 2 fi involve either two first-order perturbed wave
a1 , or one second-order
functions, i1 and f1 , and hence coefficients of the type ap1
ir pfr
2
2
2
2
perturbed wave function i or f and hence coefficients of the type apir
or apfr
. Thus
p 2 fi is quadratic in E. It can be shown likewise that p 3 fi is cubic in E through
coefficients of the following types:
ap3
ir
or ap3
,
fr
ap2
a1
fr pir
or ap1
a2
fr pir
The transition moments given by eqs. (2.4.9) to (2.4.11) each contain more than one
frequency component. If we select from p 1 fi the components with the correct frequency
dependences for Rayleigh and Raman scattering, we find that the amplitudes of the transition electric dipole can be cast in the forms
1

p0 1 ii D aii E0 1
2.4.13

and
p 1 1 M

fi
D afi E0 1
27
2.4.14
where aii and afi are transition polarizability tensors associated with Rayleigh and
Raman scattering, respectively.
Proceeding in a similar way the amplitudes of the components of p 2 fi with the
correct frequency dependences for hyper-Rayleigh and hyper-Raman scattering can be
cast in the general form
2
p0 fi D 12 bfi : E0 E0
2.4.15
where bfi is a transition hyperpolarizability and f D i for hyper-Rayleigh scattering
and f 6D i for hyper-Raman scattering.
Likewise for second hyper-Rayleigh and second hyper-Raman scattering the appropriate
third-order induced transition moment amplitudes would have the general form
3
p0 fi D 16 gfi ... E0 E0 E0
2.4.16
where gfi is a transition second hyperpolarizability and f D i for second hyper-Rayleigh
scattering and f 6D i for second hyper-Raman scattering.
As for the classical theory, geometrical factors are determined by the tensor properties
of afi , bfi and gfi and by the factor sin2 in eq. (2.2.1).
2.5 EXTENSION OF QUANTUM MECHANICAL TREATMENT OF

INCOHERENT LIGHT SCATTERING TO INCLUDE
MAGNETIC DIPOLE AND ELECTRIC QUADRUPOLE CASES
In Section 2.4 we assumed that the interaction hamiltonian involves only an electric
O p . This, as already stated, is an approximation, albeit an excellent one.
dipole term H
More generally, the interaction hamiltonian is the sum of a number of terms of decreasing
importance. Restricting ourselves to the first three terms in this series we may write
O DH
OpCH
OmCH
Oq
H
2.5.1
O m is the magnetic dipole term and H

O q is the electric quadrupole term. All three
where H
terms are, in effect, linearly dependent on E, the perturbing electric field. The hamiltonian
O q are non-zero only if E is not constant over the molecule; that is, the
O m and H
operators H
electric field gradient is non-zero. In the visible range of the spectrum, the wavelength of
electromagnetic radiation is of the order of a thousand times the dimensions of a molecule,
O q is normally
O m and H
so that E is almost constant over a molecule and the neglect of H
justified. However, there are some light-scattering phenomena whose explanation requires
O m and H
O q . Raman scattering by chiral molecules is an example.
the inclusion of H
O q in the interaction
O m and H
We now consider the general implications of including H
hamiltonian. When the interaction hamiltonian defined in eq. (2.5.1) is used to calculate
28
The Raman Effect
the perturbed wave functions, then instead of eqs. (2.4.5) to (2.4.7) we have

1 0
1
ap1
C
a
C
a
r
i1 D
m
q
ir
ir
ir
r
i2 D
..
.
in D
2.5.2
0
2
ap2
C am
C aq2
r
ir
ir
ir
2.5.3
0
n
apn
C am
C aqn
r
ir
ir
ir
2.5.4

r
1 and a1 are linear in E,

, am
and similarly for f1 , f2 , . . . fn . The coefficients ap1
qir
ir
ir
2
2 and a
, am
the coefficients ap2
qfr are quadratic in E and so on.
ir
ir
It follows that the two terms in eq. (2.4.9) are now linearly dependent not only on
1 and a1 (or
(or ap1
) but also on coefficients of the type am
coefficients of the type ap1
qir
ir
fr
ir
1
1 , a ); and these give rise to additional contributions to the transition electric dipoles
am
qfr
fr
associated with Rayleigh and Raman scattering.
Furthermore, at this higher level of approximation it is also necessary to consider,
in addition to the transition electric dipole given by eq. (2.4.2), the transition magnetic
dipole mfi given by
O i0 i
2.5.5
mfi D h f0 jmj
where m
O is the magnetic moment operator, and the transition electric quadrupole moment
tensor fi , given by
O j i0 i
fi D h f0 j
2.5.6
O is the electric quadrupole operator. These additional transition moments can also
where
have frequency components corresponding to Rayleigh and Raman scattering. Additional
transition moments will also arise in a similar way for higher-order scattering phenomena
but these are not considered in this book.
2.6 COMPARISON OF THE CLASSICAL AND QUANTUM

MECHANICAL TREATMENTS OF LIGHT SCATTERING
The two treatments of light scattering are, formally, very similar. For example, in the
case of Rayleigh and Raman scattering, comparison of eq. (2.4.13) with eq. (2.3.8) and
eq. (2.4.14) with eq. (2.3.9) shows that aRay and aRam of the classical theory are simply
replaced by aii and afi in the quantum mechanical treatment. However, this is a
change of profound significance. The classical polarizability tensors can only be related
qualitatively to molecular properties. In contrast, the transition polarizability tensors can
be related quantitatively to fundamental molecular properties and provide a much deeper
insight into the factors that characterize scattered radiation. Similar considerations apply
to the classical and transition hyperpolarizability and second hyperpolarizability tensors.
Chapter A20 treats the oscillating magnetic dipole and electric quadrupole.
29
Classical theory also has other limitations which arise because a proper treatment of
molecular rotors and vibrators is only possible using quantum mechanics.
2.7 THE WAY AHEAD

The classical theory of the incoherent linear phenomena, Rayleigh and Raman scattering
is considered in Chapter 3.
The quantum mechanical theory of Rayleigh and Raman scattering, in the electric dipole
approximation is treated in detail in Chapter 4. The theoretical treatment in Chapter 4 is
based on perturbation theory and the formulae obtained describe steady state processes
and contain no explicit reference to time. Such formulae are then applied to specific types
of Raman scattering in Chapters 59. However, a time-dependent formulation is also
included in Chapter 7 and applied to some aspects of resonance scattering.
The quantum mechanical theory of Rayleigh and Raman scattering from chiral systems
is considered in Chapter 10.
The incoherent non-linear phenomena, hyper-Rayleigh and hyper-Raman scattering and
second hyper-Rayleigh and second hyper-Raman scattering are not considered further in
this book.
31
3
Classical Theory of Rayleigh and
Raman Scattering
Se non e` vero, e` molto ben trovato.
Giordano Bruno
3.1 INTRODUCTION
We now consider an approach to the theory of Rayleigh and Raman scattering in which
both the electromagnetic radiation and the material system are treated classically. Although
the classical theory cannot deal with all aspects of Rayleigh and Raman scattering it does
provide useful insights into some facets of such scattering, particularly the frequency
dependence, and also some aspects of the selection rules.
3.2 FIRST-ORDER INDUCED ELECTRIC DIPOLE

In this section we develop in more detail the procedures outlined in Chapter 2, Section 2.3,
confining ourselves to Rayleigh scattering and vibrational Raman scattering. Our objective
is to calculate for a molecule the frequency-dependent linear induced electric dipole
vectors p 1 , making use of the relationship
p 1 D a E
3.2.1
32
The Raman Effect
where E is the electric field vector of the incident, plane wave, monochromatic radiation
of frequency 1 , and a is the polarizability tensor of the molecule. The polarizability
tensor will, in general, be a function of the nuclear coordinates and hence of the molecular
vibrational frequencies. We may therefore obtain the frequency-dependent induced electric
dipole vectors p 1 , by introducing into eq. (3.2.1) the frequency dependence of E and a.
The amplitude of a linear induced electric dipole of a particular frequency enables the
scattered intensity to be calculated using eq. (2.2.1).
We shall consider the scattering system to be one molecule which is free to vibrate, but
does not rotate; that is, the molecule is space-fixed in its equilibrium configuration, but the
nuclei may vibrate about their equilibrium positions. The variation of the polarizability
with vibrations of the molecule can be expressed by expanding each component of
the polarizability tensor a in a Taylor series with respect to the normal coordinates of
vibration, as follows:

1 2
D 0 C
Qk C
Qk Ql . . .
3.2.2
Qk 0
2 k,l Qk Ql
k
0
where 0 is the value of at the equilibrium configuration, Qk , Ql . . . are normal

coordinates of vibration associated with the molecular vibrational frequencies k , l . . . ,
and the summations are over all normal coordinates. The subscript 0 on the derivatives
indicates that these are to be taken at the equilibrium configuration. We shall, for the
time being, neglect the terms which involve powers of Q higher than the first. This
approximation is often referred to as the electrical harmonic approximation. We shall
also fix our attention initially on one normal mode of vibration Qk . We may then write
eq. (3.2.2) in the special form
k D 0 C 0 k Qk
where
0 k

Qk
3.2.3
3.2.4
0
The 0 k are components of a new tensor a0k which we shall call a derived polarizability tensor, as all its components are polarizability derivatives with respect to the
normal coordinate Qk . The properties of the components of the derived polarizability
tensor differ in some respects from 0 , the components of the equilibrium polarizability tensor a0 , as we shall see subsequently. As eq. (3.2.3) is valid for all tensor
components we may write
ak D a0 C a0k Qk
3.2.5
where ak is a tensor with components k and Qk , a scalar quantity, multiplies all
components of a0k . Assuming simple harmonic motion, that is, mechanical harmonicity ,
Chapter A14 treats the polarizability tensor.

Mechanical harmonicity in a molecular vibration means that the restoring force is proportional to the first
power of the displacement Qk . Similarly, electrical harmonicity means that the variation of the polarizability
in a vibration is proportional to the first power of Qk .
33
the time dependence of Qk is given by

Qk D Qk0 cosk t C k
3.2.6
where Qk0 is the normal coordinate amplitude and k a phase factor. Combining eq. (3.2.6)
with eq. (3.2.5) we obtain the time dependence of the polarizability tensor resulting from
the kth molecular vibration:
ak D a0 C a0k Qk0 cosk t C k
3.2.7
We now introduce into eq. (3.2.1) the frequency dependence of E given by

E D E0 cos 1 t
3.2.8
and of ak given by eq. (3.2.7). We then obtain

p 1 D a0 E0 cos 1 t C a0k E0 Qk0 cosk t C cos 1 t
3.2.9
Using the trigonometric identity

cos A cos B D 12 fcosA C B C cosA Bg
3.2.10
the second term in eq. (3.2.9) may be reformulated and we may then write the expression
for p 1 in the form
p 1 D p 1 1 C p 1 1 k C p 1 1 C k
Here
Ray
p 1 1 D p0
with
Ray
p0
cos 1 t
D aRay E0
3.2.11
3.2.12
3.2.13
and
aRay D a0
3.2.14
p 1 1 k D pkRam
cos1 k k t
0
3.2.15
D aRam
E0
pkRam
k
0
3.2.16
aRam
D 12 a0k Qk
k
3.2.17
Also
with
and
The cosine functions in eqs. (3.2.12) and (3.2.15) define the frequencies of the induced
dipoles; and eqs. (3.2.14) and (3.2.17) define the classical Rayleigh and Raman scattering
tensors introduced in eqs. (2.3.8) and (2.3.9) in Chapter 2.
Chapter A17 treats the frequency dependence of E.
34
The Raman Effect
We have used the familiar trigonometric representation of sinusoidal phenomena in this

treatment of the classical theory. However, as the exponential representation is used in
all subsequent treatments, the appropriate equations are repeated below in the exponential
representation:
for eq. 3.2.6
3.2.18
Qk D 12 Qk0 fexp ik t C k C exp ik t C k g
for eq. 3.2.8
E D 12 E0 fexp i1 t C exp i1 tg
for eq. 3.2.12
3.2.19
Ray
3.2.20
p 1 1 k D 12 p0Ram fexp i[1 k t k ] C exp i[1 k t k ]g
3.2.21
p 1 1 D 12 p0 fexp i1 t C exp i1 tg

for eq. 3.2.15
3.3 FREQUENCY DEPENDENCE OF THE FIRST-ORDER

INDUCED ELECTRIC DIPOLE
We see from eq. (3.2.11) that the linear induced electric dipole has three distinct frequency
components: p 1 1 , which gives rise to radiation at 1 , and so accounts for Rayleigh
scattering; p 1 1 k , which gives rise to radiation at 1 k , and so accounts for
Stokes Raman scattering; and p 1 1 C k , which gives rise to radiation at 1 C k ,
and so accounts for anti-Stokes Raman scattering. It should be noted that, whereas the
induced dipole p 1 1 has the same phase as that of the incident field, the induced
dipoles p 1 1 k are shifted in phase relative to the incident field by k . The quantity
k defines the phase of the normal vibration Qk relative to the field and will be different
for different molecules.
The time dependence of the electric field of the electromagnetic radiation of frequency
1 and the linear induced electric dipoles p 1 which it can produce are compared in
Fig. 3.1. In Fig. 3.1(a) the scattering molecule is not vibrating k D 0 and the total
induced electric dipole then has only one frequency component p 1 1 In Fig. 3.1(b),
the scattering molecule is vibrating with frequency k and the total induced electric dipole
is resolvable into three frequency components:
p 1 1 , p 1 1 k
and p 1 1 C k
This relatively simple classical treatment provides us with a useful qualitative picture
of the mechanisms of Rayleigh and Raman scattering. Rayleigh scattering arises from the
electric dipole oscillating at 1 induced in the molecule by the electric field of the incident
radiation, which itself oscillates at 1 . Raman scattering arises from the electric dipoles
oscillating at 1 k , which are produced when the electric dipole oscillating at 1 is
Chapter A17 treats the trigonometric and exponential representations of sinusoidal phenomena such as
electromagnetic waves.
(a)
wk = 0
35
p(1)(w1)
w0
(b)
p (1)(w1)
p (1)(w1wk )
wk
p (1)
p (1)(w1+ wk )
Figure 3.1 Time dependence of the linear induced dipoles p 1 produced by electromagnetic radiation of frequency 1 : (a) scattering molecule not vibrating k D 0: p 1 D p 1 1 ;
(b) scattering molecule vibrating with frequency k : p 1 D p 1 1 C p 1 1 k C
p 1 1 C k .
modulated by the molecule oscillating at frequency k . The necessary coupling between

the nuclear motions and the electric field is provided by the electrons whose rearrangement
with nuclear motion impose a harmonic variation on the polarizability. Alternatively, we
may use a musical analogy and say that the frequencies observed in Raman scattering are
beat frequencies between the radiation frequency 1 and the molecular frequency k .
3.4 CLASSICAL SCATTERING TENSORS aRay AND aRam

k
It is evident that the necessary condition for Rayleigh scattering is that aRay be non-zero.
As all molecules are polarizable to a greater or lesser extent, the classical equilibrium
polarizability tensor a0 will always have some non-zero components and so aRay is always
non-zero. Thus all molecules exhibit Rayleigh scattering.
The corresponding necessary condition for Raman scattering associated with a molecular frequency k is that aRam
be non-zero. This requires that at least one of the compok
nents 0 k of the derived polarizability tensor a0k is non-zero. We recall from eq. (3.2.4)
that 0 k is the derivative of the component of the polarizability tensor with respect
to the normal coordinate of vibration Qk , taken at the equilibrium position. Thus, the
condition for Raman activity is that, for at least one component of the polarizability
tensor, a plot of that component against the normal coordinate must have a non-zero
gradient at the equilibrium position.
36
The Raman Effect
3.5 SELECTION RULES FOR FUNDAMENTAL VIBRATIONS

In principle we can now determine the Raman activity of vibrations in particular molecules.
However, the selection rules in their classical form, although appearing deceptively simple
in concept, become progressively more difficult to apply as the complexity of the molecule
increases. The ensuing discussion of a few specific cases will illustrate this.
In dealing with these specific cases it will be instructive to consider the infrared activity
as well as the Raman activity so that comparisons can be made. We note that according
to classical theory the condition for vibrational infrared activity is that at least one of the
dipole moment component derivatives with respect to the normal coordinate Qk , taken
at the equilibrium position, should be non-zero. This means that, for at least one of the
components of the dipole moment vector, a plot of that component against the normal
coordinate must have a non-zero gradient at the equilibrium position.
3.5.2 Diatomic molecules

We shall consider first a homonuclear diatomic molecule A2 which has just one mode of
vibration. Such a molecule has no permanent dipole moment in the equilibrium position
because of the symmetry of the electron distribution. This symmetry does not change
with small changes in the internuclear separation, and so the dipole remains zero during
a vibration and hence the derivative is zero. The vibration is therefore infrared inactive.
We turn now to the question of Raman activity. Clearly the molecule has a non-zero
polarizability, and we may represent this by a polarizability ellipsoid , so oriented at
equilibrium that it has one principal axis along the bond direction and its other two
principal axes at right angles to the bond direction. Such an ellipsoid is defined by a
maximum of three polarizability components. However, as we are considering a bond,
the polarizability will be the same in all directions at right angles to the bond, and the
polarizability tensor is then defined by just two components which we may designate
as jj , the polarizability along the bond, and ? , the polarizability at right angles to
the bond. For a given internuclear separation the mean polarizability a is then given
by 13 jj C 2? and the anisotropy by jj ? . We have now to ask how these
polarizability components change during the vibration of the A2 molecule. It is not as
easy as many texts would imply to deduce conclusively the properties of jj /Qk 0 and
? /Qk 0 from a priori considerations of how electron distributions are affected by a
change of internuclear distance.
Let us consider the specific case of the hydrogen molecule H2 . We can see that as the
internuclear separation tends to zero the polarizability tends towards that of a helium atom,
and as the internuclear separation tends to infinity the polarizability tends to that of two
hydrogen atoms. As we know that the polarizability of two hydrogen atoms is greater than
that of a helium atom, we can infer that, in general terms, the polarizability does change
with internuclear separation. However, these limiting cases do not define the form of the
Chapter A14 treats the polarizability ellipsoid, and also the mean polarizability and the anisotropy.
37
2 aH
?
a
Equilibrium
separation
aHe
Internuclear separation
Figure 3.2 Polarizability and internuclear distance in the H2 molecule with limiting values
for zero and infinite internuclear distances.
30
1041a /CV1 m2
a||
a
20
a
10
R eq
0
100
200
300
400
500
R /pm
Figure 3.3 Plots of , jj , and ? as functions of R, the internuclear distance, for the H2
molecule. The mean polarizability is shown only up to R 390 pm so that the very close
curves for jj and ? above this distance remain distinct.
variation of the polarizability with internuclear separation, and in principle the derivative
at the equilibrium internuclear separation could be positive, negative, or zero (Fig. 3.2).
Also, as atoms have isotropic polarizabilities with a 6D 0 and D 0, whereas near the
equilibrium position in H2 we have a 6D 0 and 6D 0, the forms of the variation of a and
with internuclear separation in H2 must be different. For the fullest information, we must
appeal to quantum mechanical calculations to ascertain the variation of the polarizability
components with internuclear separation in the neighbourhood of the equilibrium position.
The results of a typical calculation (Rychlewski, 1980) are presented in Fig. 3.3 and show
The Raman Effect
Dipole moment
38
Equilibrium
separation
Internuclear distance
Figure 3.4 Variation of dipole moment with internuclear distance in a diatomic molecule
AB (here the dipole moment derivative at the equilibrium separation is negative).
that both jj /Q0 and ? /Q0 are non-zero and positive for values of the internuclear
distance R up to about R D 175 pm. Further, the calculations show that jj /Q0 >
? /Q0 and so /Q0 is also non-zero and positive up to about R D 175 pm.
These results for H2 may be used to guide us towards some generalizations for other
homonuclear diatomic molecules. In such molecules, we may also reasonably expect
jj /Q0 and /Q0 to be non-zero and of different magnitudes and hence a/Q0
and /Q0 to be non-zero. Thus, the vibrations of A2 diatomic molecules will be
Raman active. However the signs and relative magnitudes of these derivatives need not
necessarily follow the pattern in H2 .
We next consider the case of a heteronuclear diatomic molecule AB which also has
just one mode of vibration. The arguments given above for polarizability changes in A2
molecules can be expected to apply to AB molecules, and thus the vibration will be Raman
active. The molecule AB will necessarily have a permanent dipole moment because there
will be an asymmetry in its electron distribution. As the dipole moment must be zero for
both infinitely large and zero internuclear separations, the normal form of the variation
of the dipole moment component along the bond direction with internuclear distance
will be as shown in Fig. 3.4. The components of the dipole at right angles to the bond
direction are, of course, always zero. For infrared activity, the maximum dipole moment
must occur at an internuclear distance different from the equilibrium distance, so that the
derivative at the equilibrium position is non-zero. This is the case for all heteronuclear
diatomic molecules and thus the vibration in AB molecules will be infrared active. The
form of the plot of dipole moment against internuclear distance will, however, vary from
one molecule to another, and thus the magnitude and sign of the derivative will also vary
considerably. The A2 and AB cases are compared in Fig. 3.5.
3.5.3 Polyatomic molecules

In polyatomic molecules, if we regard the total dipole moment as made up of contributions
from individual bond dipoles (at least to a first approximation), then we may regard each

Molecule
39
A
Mode of
vibration
Variation of
polarizability
with normal
coordinate
(schematic)
Polarizability
derivative at
equilibrium
position
Raman
activity
Yes
Yes
Dipole
moment
derivative
at equilibrium
position
=0
Infrared
activity
No
Yes
Variation of
dipole moment
with normal
coordinate
(schematic)
Figure 3.5 Comparison of polarizability and dipole moment variations in the neighbourhood of the equilibrium position and vibrational Raman and infrared activities for an A2 and
an AB molecule.
heteronuclear bond as having a non-zero bond dipole derivative at the equilibrium position
and combine such derivatives vectorially to determine infrared activities in particular
modes of vibration, taking into account the relative phases of the motions in each bond.
Similarly, we may regard the total molecular polarizability as made up of contributions
from individual bond polarizabilities (at least to a first approximation), and assume that
these bond polarizabilities show the same qualitative behaviour as in H2 (at least for
bonds). However, in polyatomic molecules the forms of the vibrations are relatively
complicated and involve, for example, the stretching and compression of more than one
bond. The Raman activity of such a vibration depends on the components of the overall
derived polarizability tensor which is formed by tensor addition of individual derived
40
The Raman Effect
bond polarizability tensors, taking into account the relative phases of the motions in each
bond. It is profitable to consider only very simple polyatomic molecules as examples.
We examine first a linear symmetric molecule ABA. Such a molecule has four modes of
vibration: a symmetric stretching mode Q1 , an antisymmetric stretching mode Q2 , and two
bending modes Q3a and Q3b which form a degenerate pair and have the same frequency
of vibration. These vibrations are illustrated in Fig. 3.6, and in the following discussion
we shall take the z axis as the bond axis. This ABA molecule has no permanent dipole
because of the symmetry of the electron distribution. For the symmetric stretching mode
Molecule
Mode of
vibration
Variation of
polarizability
with normal
coordinate
(schematic)
Polarizability
derivative at
equilibrium
position
=0
=0
Raman
activity
Yes
No
No
Dipole
moment
derivative at
equilibrium
position
=0
Infrared
activity
No
Yes
Yes
Variation of
dipole moment
with normal
coordinate
(schematic)
Figure 3.6 Polarizability and dipole moment variations in the neighbourhood of the equilibrium position and vibrational Raman and infrared activities for a linear ABA molecule.
41
Q1 , in which both AB bonds are stretched simultaneously in one phase and compressed
simultaneously in the other phase, the non-zero bond dipole derivatives in the two AB
bonds always act in opposition and cancel each other exactly; the vibration is therefore
infrared inactive. However, for this vibration the non-zero bond polarizability derivatives
in the two AB bonds are additive and the vibration is Raman active. Specifically, as the
axes of the molecular polarizability ellipsoid do not change in this vibration, the molecular
polarizability derivatives xx /Q1 0 D yy /Q1 0 and zz /Q1 0 are non-zero; also,
xy /Q1 0 D yz /Q1 0 D zx /Q1 0 D 0. Thus, a/Q1 0 and /Q1 0 are both
non-zero.
The situation is quite different for the antisymmetric stretching mode Q2 , in which
one AB bond is stretched (or compressed) as the other AB bond is compressed (or
stretched). The non-zero bond dipole derivatives are additive and thus this vibration is
infrared active. However, the bond polarizability derivatives cancel each other and we have
for the molecular polarizability derivatives xx /Q2 0 D yy /Q2 0 D zz /Q2 0 D
xy /Q2 D yz /Q2 0 D zx /Q2 0 D 0 and so a/Q2 0 and /Q2 0 are both
zero. Thus, the Q2 mode is Raman inactive.
For the degenerate bending modes Q3a and Q3b , as far as infrared activity is concerned,
it is readily seen that the dipole moment derivatives at right angles to the molecular
axis are non-zero and thus these vibrations are infrared active. However, all six molecular polarizability derivatives are zero for both modes. Thus, a/Q3a 0 D a/Q3b 0 D
/Q3a 0 D /Q3b 0 D 0 and both modes are Raman inactive. The general forms of
the dipole moment and polarizability changes with the various normal coordinates are
included in Fig. 3.6.
Finally, we consider the case of a non-linear ABA molecule. Such a molecule has three
modes of vibration: Q1 , a symmetric stretching mode; Q2 , a symmetric bending mode; and
Q3 , an antisymmetric stretching mode. The forms of these modes and the axis system are
given in Fig. 3.7. We may perhaps leave it to the reader to endeavour to convince himself
that: (a) Q1 is infrared active and also Raman active with xx /Q1 0 , yy /Q1 0 and
zz /Q1 0 non-zero, and xy /Q1 0 D yz /Q1 0 D zx /Q1 0 D 0 so that both
a/Q1 0 and /Q1 0 are non-zero; (b) Q2 is also both infrared active and Raman
active with the same dipole and polarizability components non-zero. The case of Q3 calls
for a little more explanation. It is quite easy to see that Q3 is infrared active; Q3 is also
Raman active. This activity arises because the space-fixed z and y axes no longer remain
axes of the polarizability ellipsoid during the whole of the vibration. Thus, although yz
is zero, in the equilibrium configuration yz /Q3 0 is non-zero. This is the only nonzero component of the derived polarizability for this mode, and hence a/Q3 0 D 0,
although /Q3 0 is non-zero. Figure 3.7 shows the forms of the dipole moment and
polarizability changes with the various normal coordinates.
Comparison of the vibrational activities in the molecules considered above shows that,
for those molecules with a centre of symmetry, those vibrations which are Raman active
are infrared inactive, and vice versa. This can be shown to be a general rule which is often
termed the rule of mutual exclusion. It can form the basis for distinguishing between two
alternative configurations of a molecule, for instance between a linear and a non-linear
configuration for an ABA molecule. In many other cases, it is also possible to distinguish
42
The Raman Effect

z
Molecule
x out of
plane
Mode of
vibration
Variation of
polarizability
with normal
coordinate
(schematic)
Polarizability
derivative at
equilibrium
position
Raman
activity
Yes
Yes
Yes
Dipole
moment
derivative at
equilibrium
position
Infrared
activity
Yes
Yes
Yes
Variation of
dipole moment
with normal
coordinate
(schematic)
Figure 3.7 Polarizability and dipole moment variations in the neighbourhood of the equilibrium position and vibrational Raman and infrared activities for a non-linear A B A
molecule.
between alternative configurations of a molecule by comparing the predicted numbers

of vibrational modes which are Raman and/or infrared active. This has proved to be a
valuable method of structural elucidation. Its main limitation is that, whereas observation
of a band is proof of its activity, the converse is not necessarily true. Some bands, although
permitted in principle, may fail to be observed, either because they are inherently weak or
because of limitations in the experimental technique. A careful study of the literature will
teach the spectroscopist the importance of tempered judgement and guarded optimism!
43
Broadly speaking, with molecules of relatively low symmetry all, or nearly all, vibrations are both infrared and Raman active. In molecules containing many atoms this might
be thought to lead to vibrational spectra so rich in bands as to be unintelligible. In fact, in
such molecules many of the vibrational frequencies are very close in value so that bands
overlie each other and a relatively simple spectrum results. Thus even molecules such as
DNA yield vibrational spectra with identifiable and characteristic features.
In some molecules, especially those of high symmetry, some vibrations may be both
infrared and Raman inactive; however, such modes may be active in light-scattering
spectra of non-linear origin, as for example in hyper-Raman spectra.
These qualitative considerations serve to indicate the importance of the symmetry of a
molecule and its vibrational modes in determining infrared and Raman activity. It is also
evident that the qualitative arguments used here would be difficult to apply to more complicated molecules. It would be difficult to infer with certainty the symmetry properties of the
modes of vibration, let alone the behaviour of the dipole moment and the polarizability.
Fortunately, there exist mathematical procedures which enable the symmetry properties of vibrational modes, dipole moment, and polarizability derivatives for molecules
to be predicted for an assumed equilibrium configuration of the nuclei. We shall consider
how such procedures operate after we have discussed quantum mechanical approaches
to light-scattering phenomena. It should be noted in advance that these new procedures
are completely independent of special assumptions regarding mechanical and electrical
anharmonicity. They are derived solely on the basis of the symmetry of the vibrations and
the electrical properties of the molecules and, consequently, have wide validity. However,
these symmetry arguments tell us nothing about the intensity with which a vibrational
band will appear in infrared or Raman spectra. Once again we emphasize that a vibration,
while formally active, may be so weak as to be undetectable in practice.
3.6 SELECTION RULES FOR OVERTONES AND

COMBINATIONS
In the foregoing we have been concerned entirely with fundamental vibrations; indeed,
in the approximation of mechanical and electrical harmonicity, it would appear that only
fundamental vibrations can occur. When anharmonicities are taken into account, overtone
and combination bands are also permitted. Fortunately, in Raman spectra, unlike infrared
spectra, such bands are almost invariably much weaker than the fundamentals. Vibrational Raman spectra are therefore usually much simpler than infrared spectra. However,
it is worthwhile examining briefly the factors controlling the activity of overtone and
combination bands in the Raman effect.
If mechanical anharmonicity is taken into account, the time dependence of the normal
coordinate Qk will not depend solely on cosk t C k as in eq. (3.2.6), but will include
terms involving cos2k t C 2k , cos3k t C 3k etc., which relate to overtones, and
also terms involving cosk t C kl cosl t C 0kl which relate to combination tones. As
a consequence of the presence of these terms, there will be additional induced electric dipoles with frequencies 1 2k etc., and 1 k l etc. Thus, mechanical
anharmonicity can lead to the observation of overtones and combinations in the Raman
44
The Raman Effect
effect. As a0k is unaffected by the introduction of mechanical anharmonicity, the same

selection rules apply to the overtones and combination tones as to the fundamentals. For
example, in a linear symmetric molecule ABA overtones of 1 (i.e 21 , 31 etc.) would
be Raman active, but not overtones of 2 or 3 . Similarly, some combinations of 1 with
2 or 3 would be Raman active, but no combinations of 2 with 3 .
The situation is different if electrical anharmonicity is taken into account. This means
that the third (and possibly higher) terms in the Taylor series expansion of the polarizability
in terms of the normal coordinates eq. (3.2.2) must be considered. The consequence of this
is that there will again be additional induced electric dipoles with frequencies 1 2k
etc. and 1 k l etc., but in this case these dipoles involve a new derived tensor
with components of the type 2 /Qk2 0 , 2 /Qk Ql 0 and so on.
Thus overtones and combinations arising from electrical anharmonicity can be Raman
active even if the fundamental vibration is not. For example, in a linear symmetric
molecule ABA, although 2 and 3 are not Raman active, 22 and 23 are Raman active
because, as can be seen from Fig. 3.6, although the first derivative of the polarizability is
zero, the second derivative is not for 2 or 3 .
3.7 COHERENCE PROPERTIES OF RAYLEIGH AND RAMAN

SCATTERING
The classical treatment given earlier in Section 3.2 is based on a single scattering molecule
which, although free to vibrate, is space-fixed in its equilibrium configuration. However
in any experimental study scattering will be observed from a relatively large number of
molecules and their orientations will not necessarily be fixed. For such an assembly of
molecules there is an important difference between Rayleigh and Raman scattering. This
difference arises because, as emphasized in Section 3.3, Rayleigh scattering is in phase
with the incident radiation, whereas Raman scattering bears an arbitrary phase relation
to the incident radiation. This is because the phase of Raman scattering depends on the
phase of the molecular vibration (see eqs. 3.2.15 and 3.2.21) and this to a very good
approximation varies arbitrarily from molecule to molecule. Thus the molecules act as
independent sources of radiation irrespective of the degree of correlation between their
positions. The situation is different for Rayleigh scattering because interference between
the scattering from different molecules is possible and so the structural arrangement of
the scattering molecules will play a role. For example, in the idealized case of a perfect
crystal at absolute zero, this interference could result in the Rayleigh scattering averaging
to zero. The strong dependence of Rayleigh scattering on the state of matter is illustrated
by the fact that Rayleigh scattering from water is only 200 times stronger than from air,
even though water contains 1200 times more molecules in the same volume. However,
for scattering by an ideal gas, Rayleigh scattering is incoherent in non-forward directions
and the intensity scattered by N molecules is simply N times that scattered by a single
This statement specifically excludes higher-order coherent Raman-type processes such as CARS and
CSRS, (see Chapter 1, Section 1.6) and SRGS and SRLS (see Chapter 1, Section 1.7).
45
molecule. In the forward direction the situation is less straightforward and depends on
the relative contributions of the isotropic and anisotropic parts of the Rayleigh scattering.
With the coherence properties of Rayleigh and Raman scattering established we could
proceed from the simple case of one space-fixed molecule to an assembly of freely rotating
molecules and calculate the intensity and polarization properties of the scattered radiation
as a function of the illuminationobservation geometry.
However, these developments are not pursued here for the classical theory because of
its limited validity, as explained in the next section.
3.8 LIMITATIONS OF THE CLASSICAL THEORY

The classical theory gives the correct frequency dependence for Rayleigh scattering and
vibrational Raman scattering. It also shows, correctly, the dependence of the Rayleigh scattering tensor on the equilibrium polarizability tensor a0 , and its prediction of the dependence of the vibrational Raman scattering tensor on the derived polarizability tensor a0k
turns out to be correct in certain situations. For those who want to use Raman spectroscopy
merely to obtain characteristic molecular vibrational frequencies and use them as molecular signatures for qualitative analysis, the classical theory might be deemed adequate.
However, as might be expected, the classical theory has many limitations. It cannot
be applied to molecular rotations as classical theory does not ascribe specific discrete
rotational frequencies to molecules. The result for the vibrational Raman scattering tensor
given by eq. (3.2.17) is only partly correct. When quantum mechanics is used for the
treatment of molecular vibrations, the classical amplitude Qk is replaced by a quantum
mechanical amplitude. Also, the classical theory cannot provide information as to how a0k
is related to the properties of the scattering molecule, in particular its characteristic transition frequencies, and to the frequency of the incident radiation. The quantum mechanical
theory which will be developed in subsequent chapters provides this information and
forms the basis for a complete treatment of all aspects of Raman scattering. This treatment will reveal Raman scattering as a powerful and versatile tool which can be used to
determine molecular parameters and to explore in some detail the spectroscopic properties
not only of the ground electronic state but also of upper electronic states of molecules.
Like our own signatures, Raman spectra, when properly analysed, can reveal much about
the character as well as the identity of the molecular signatory.
3.9 EXAMPLE OF RAYLEIGH AND RAMAN SCATTERING

In this section and the following one, both of which deal with observed spectra, we
use wavenumber Q instead of the angular frequency which was appropriate for the
theoretical treatment earlier in this chapter.
The phenomena of Rayleigh and Raman scattering are illustrated in Fig. 3.8,
which includes facsimiles of two of the first spectra published by Raman and
Krishnan. Figure 3.8(a) shows the photographically recorded spectrum of the essentially
monochromatic radiation from a mercury arc used to produce the scattering; the
46
The Raman Effect

Mercury arc line
at 4358.3 (435.83 nm, 22 938 cm1)
(a)
n~1 762 cm1

n~1 790 cm1
~n + 314 cm1
1
~n + 218 cm1
1
~n + 459 cm1
1
n~1
~n + 790 cm1
1
~n + 762 cm1
~n 459 cm1
1
~n 218 cm1
1
~n 314 cm1
(b)
Stokes Raman bands

(type n~1 n~M)
anti-Stokes Raman bands

(type n~1 + n~M)
(c)
Rayleigh
band, n~
1
(435.83 nm,
Figure 3.8 (a) Spectrum of a mercury arc in the region of 4358.3 A
Q 1 D 22 938 cm1 . (b) Rayleigh and Raman spectra of carbon tetrachloride (liquid) excited
by mercury arc radiation, Q 1 D 22 938 cm1 . (c) Rayleigh and Raman spectra of carbon
487.99 nm)
tetrachloride (liquid) excited by an argon ion laser, Q 1 D 20 487 cm1 (4879.9 A,
and recorded directly. The spectra in (a) and (b) are facsimiles of spectra reported by Raman
and Krishnan (1929) and were photographically recorded.
47
spectrum has one intense line at wavenumber Q 1 D 22 938 cm1 (or wavelength 4358.3 A,
435.83 nm). The mercury spectrum contains other lines in this region, but they are
much weaker and need not be considered here. Figure 3.8(b) shows the photographically
recorded spectrum of the radiation after scattering by carbon tetrachloride (liquid). This
spectrum contains a strong band at Q 1 D 22938 cm1 due to Rayleigh scattering of the
incident radiation, and a number of weaker bands whose absolute wavenumbers conform
to the pattern Q 1 Q M and are given by Q 1 218, Q 1 314, Q 1 459, Q 1 762 and
Q 1 790 cm1 . The first three pairs of bands arise from Stokes and anti-Stokes Raman
scattering associated with Q M values of 218, 314 and 459 cm1 . The remaining two bands
arise from Stokes Raman scattering associated with Q M values of 762 and 790 cm1 ; the
corresponding anti-Stokes bands are not observed. The Q M values relate to fundamental
vibrations of the carbon tetrachloride molecule.
Rayleigh and Raman scattering are now invariably excited by monochromatic radiation
from a suitable laser and recorded directly, for example, photoelectrically. For comparison
with Fig. 3.8(b), the Rayleigh and Raman spectra of carbon tetrachloride (liquid) excited
487.99 nm) from an argon ion laser and directly
with 20 487 cm1 radiation (4879.9 A,
recorded are shown in Fig. 3.8(c). In addition to all the bands observed previously, this
spectrum contains the anti-Stokes bands at Q 1 C 762 and Q 1 C 790 cm1 , which were
missing from the spectrum in Fig. 3.8(b). Their very low intensity explains why they
were not observed in the early work.
The patterns of bands observed in the spectra in Figs. 3.8(b) and 3.8(c) are entirely
consistent with the theory developed in Sections 3.2 and 3.3. These spectra also show that
Rayleigh scattering is much more intense than Raman scattering, and that Stokes Raman
scattering is more intense than anti-Stokes Raman scattering with the ratio of the intensity
of anti-Stokes to Stokes Raman scattering decreasing rapidly as Q M increases. However,
in order to interpret and understand these observations and other characteristic properties
of the scattered radiation which can be measured by appropriately designed experiments,
a knowledge of the theoretical developments treated in subsequent chapters is required.
3.10 PRESENTATION OF RAMAN SPECTRA

For a given Raman band let us denote its absolute peak wavenumber by Q 0 and define its
wavenumber shift from Q 1 , the wavenumber of the exciting radiation, as Q D Q 1 Q 0 . With
this definition Q is positive for Stokes Raman scattering Q 0 D Q 1 Q M and negative for
anti-Stokes Raman scattering Q 0 D Q 1 C Q M . The characteristic molecular property, the
wavenumber Q M is given by jQ j; for Stokes Raman scattering Q D Q M .
Raman spectra are normally presented in terms of Q and not Q 0 unless some dependence on Q 0 needs to be emphasized. For Stokes Raman scattering Q M can be used as an
equivalent alternative to Q . It has been recommended by IUPAC that Raman spectra
should be plotted with the abscissa linear in the wavenumber shift Q (unit: cm1 )
increasing to the left, and the ordinate linear and proportional to the intensity. As an
See Chapter 1, Section 1.8.

See also Chapter 1, pages 4 and 5.
48
The Raman Effect

Stokes Raman
bands (n~ positive)
anti-Stokes Raman
bands (n~ negative)
800
600
400
200
0
200
~ 1
Wavenumber shift n/cm
400
762
790
459
790
762
314
Intensity
218
218
314
459
Rayleigh
band
n~1
600
800
Figure 3.9 Rayleigh and Raman spectra of carbon tetrachloride (liquid) presented according
to IUPAC recommendations. Note that normally only Stokes Raman bands are studied. It is
then acceptable to label the abscissa as wavenumber/cm1 .
example the directly recorded Raman spectrum of Fig. 3.8(c), which has the wavenumber
shift increasing to the right in order to match the photographically recorded Raman spectrum of Fig. 3.8(b), has been redrawn in Fig. 3.9 to conform with the recommended
conventions. The use of a quotient of a physical quantity and a unit, that is quantity
calculus, in the labelling of the wavenumber axis should be noted. This practice means
that the values of the quantities are presented as pure numbers. Quantity calculus should
also be used for the intensity axis when the intensity has a specific unit; otherwise, the
ordinate axis should be labelled intensity with an arrow indicating that the intensity
increases upwards as in Fig. 3.9.
For illustrative Raman spectra specially prepared for this book we shall follow the
IUPAC conventions but Raman spectra taken from the literature will not be redrawn for
conformity.
REFERENCES
Raman, C. V. and Krishnan, K. S. (1929). Proc. Roy. Soc. Lond. 122, 23.
Rychlewski, J. (1980). Molec. Phys. 41, 833.
49
4
Quantum Mechanical
Theory of Rayleigh and
Raman Scattering
Oh polished perturbation! golden care!
That keepst the ports of slumber open wide
To many a watchful night!
William Shakespeare
4.1 INTRODUCTION
In this chapter we show how quantum mechanics, on the basis of certain assumptions,
enables us to obtain expressions for the first-order induced transition electric dipole
moment p 1 fi and hence the transition polarizability afi for Rayleigh and Raman scattering. The induced transition magnetic dipole and quadrupole moments are considered
in Chapter 10.
Our treatment is based on time-dependent perturbation theory as outlined in Chapter 2.
The treatment follows closely the now classic treatment of Placzek (1934). The general
formula for afi that results is not particularly tractable and it is necessary to introduce a
number of simplifications to obtain more useful expressions for afi . A substantial part
of this chapter is therefore concerned with consideration of such approximations and the
conditions under which they are valid. The formulae obtained in this way are then used in
later chapters which are devoted to particular types of Raman scattering, namely rotational,
vibrational and electronic Raman scattering, under normal and resonance conditions.
50
The Raman Effect
To cater for the reader who is prepared to take the results of this chapter largely on
trust, each of the subsequent chapters begins with a summary of the relevant formulae
and the approximations involved.
4.2 TIME-DEPENDENT PERTURBATION THEORY AND af i

We showed in Chapter 2 that the first-order induced transition electric dipole moment is
given by

1

p fi D f1 pO i0 C f0 pO i1
4.2.1
where i0 and f0 are the unperturbed time-dependent wave functions of the initial state
i and final state f, respectively, i1 and f1 are the corresponding first-order perturbed
time-dependent wave functions, and pO is the electric dipole moment operator.
In outline, the procedure for evaluating eq. (4.2.1) is as follows. The first step involves
obtaining the relationships of the perturbed time-dependent wave functions i1 and f1
to the unperturbed time-dependent wave functions i0 and f0 of the system. These
relationships are derived on the following assumptions: the perturbation is first order,
the interaction hamiltonian for the perturbation is entirely electric dipole in nature , and
the perturbation is produced by the time-dependent electric field associated with a plane
monochromatic electromagnetic wave of frequency 1 . These relationships are then used
to substitute for i1 and f1 in eq. (4.2.1). Subsequently, the resulting terms are collected
according to their frequency dependence and those terms which correspond to Rayleigh
and Raman scattering are identified.
Before we set out the results of this procedure we draw attention to the following points.
To maintain generality in the treatment of the incident electromagnetic radiation we shall
consider the electric field amplitudes to be complex. However, for much of the ensuing
development we take the time-independent wave functions to be real. Complex wave functions only arise in a limited number of special cases, namely when magnetic perturbations
are present or the material system has degenerate states. The results of the expansion of
the treatment to include the case of complex wave functions are included in Section 4.4.
It must be emphasized that the terms which result from the substitutions for i1 and
1
f in eq. (4.2.1) are all inherently complex in nature by virtue of the forms of their
exponential time-dependent parts, even if all other parts are real. Therefore, in anticipation
of the complex nature of eq. (4.2.1) we rewrite it as

1

p
D 1 p 0 C 0 p 1
4.2.2
fi
and
define
the real induced transition electric dipole moment p 1 fi as

1

p fi D p 1 fi C p 1 fi
4.2.3
where p 1 fi is the conjugate complex of p 1 fi .
The case where the interaction hamiltonian involves electric dipole, magnetic dipole and electric
quadrupole terms is considered in Chapter 10.
This definition of the real induced transition electric dipole moment follows that generally used in the
literature and does not involve the usual factor of 1/2.
51
The procedure we have just outlined leads to the following expression for the component of p 1 fi for real wave functions:
1
1
p fi D
2h r6Di
O j r ih r jpO j i i
f jp
ri 1 ir
O j r ih r jpO j i i
f jp
EQ 0 exp i1 fi t

EQ 0 exp i1 C fi t
ri C 1 C ir

1 h f jpO j r ih r jpO j i i Q
E0 exp i1 C fi t
C
2h r6Df
rf 1 ir

h f jpO j r ih r jpO j i i Q
C
E0 exp i1 fi t
rf C 1 C ir
C complex conjugate
4.2.4
The restrictions on the states involved in the summation here and in subsequent
formulae are discussed later, following eq. (4.2.11).
For convenience we explain below the symbols used in eq. (4.2.4). The wave functions
i , r and f which occur in eq. (4.2.4) are time-independent unperturbed wave functions
of the states i, r and f, respectively. The relation between r and the corresponding
time-dependent wave function r is given by
r D
exp ir ir t
where
r D
Er
h
4.2.5
4.2.6
Here, Er is the energy (unit: J) of the state r, and 2r (unit: rad s1 ) relates to the full
width of the level r. If this width is expressed as an energy (unit: J) it is given by 2hr .
However, it is common practice to express the width in wavenumbers (unit: cm1 ). If we
use self-evident labels to distinguish r in different units we have 2r E D 2hr D
4c0h r Q . The lifetime
r of the state r is related to 2r through the uncertainty
principle. Thus
r D h /2r E or, alternatively,
r D 1/2r D 1/4c0 r Q . In
practice the units applicable to 2r will be clear from the context and so we shall avoid
labelling 2r unless clarity requires this.
For the initial and final states i and f we assume that their lifetimes are infinite so that
i D r D 0. As a result
4.2.7
i D i exp ii t
The use of 2r for the full width is convenient because many of the subsequent formulae involve the
half-width which in this convention is r . However, some authors use r for the full width and hence r /2
for the half-width.
52
The Raman Effect
and similarly for f and f . Also, the denominators in eq. (4.2.4) contain only r and
not ri or rf . A double subscript on indicates a frequency difference; for example,
ri D r i
4.2.8
pO and pO are the and components of the electric dipole moment operator . EQ 0 is
the component of the complex amplitude of the plane harmonic electromagnetic wave
of frequency 1 associated with the incident radiation.
The terms in eq. (4.2.4) are of two types as regards their frequency dependence, namely
those with the frequency 1 fi in the exponential term and those with the frequency
1 C fi in the exponential term.
According to Placzek (1934) who invoked a theorem due to Klein (1927), the terms
in eq. (4.2.4) involving 1 fi describe the generation of Rayleigh and Raman scattering provided that 1 fi > 0. If fi is negative, that is the final state is lower in
energy than the initial state, as in anti-Stokes Raman scattering, this condition is always
satisfied. Similarly, if fi is zero, that is the initial and final states have the same energy,
as in Rayleigh scattering, this condition is always satisfied. If fi is positive, that is the
final state is higher in energy than the initial state, as in Stokes Raman scattering, then
the condition h 1 fi > 0, or equivalently h 1 > h fi means that the energy of the
incident quantum must be more than sufficient to reach the final state f from the initial
state i. For the rotational and vibrational transitions which involve no change of electronic state this condition is always satisfied for excitation frequencies in the visible and
ultraviolet regions of the spectrum.
The terms in eq. (4.2.4) involving 1 C fi describe induced emission of two quanta,
1 C fi and 1 , from an initial state, which is an excited level i , to a lower level f
if 1 C fi > 0. These will not be considered here.
With these considerations in mind, it follows that the Stokes and anti-Stokes Raman
part of the component of the real induced transition electric dipole moment is

1
1 hfjpO jrihrjpO jii hfjpO jrihrjpO jii Q
p fi D
C
E0 exp is t
2h r6Di,f ri 1 ir
rf C 1 C ir
C complex conjugate
4.2.9
In eq. (4.2.9) we have simplified the notation somewhat. We have written jii for j i i,
hfj for h f j and so on, and introduced the absolute frequency of the scattered radiation
s , defined as
4.2.10
s D 1 fi
We now introduce a general transition polarizability afi with components fi
defined as

1 hfjpO jrihrjpO jii hfjpO jrihrjpO jii
fi D
C
4.2.11
h r6Di,f ri 1 ir
rf C 1 C ir
Chapter A13, Section A13.4 treats the electric dipole moment operator.
Chapter A17 treats plane harmonic electromagnetic waves.
53
Although for the time being we are taking the wave functions to be real, this general
transition polarizability is in fact complex because of the presence of i in the denominators. However, we shall avoid calling it a complex transition polarizability and reserve
that usage for the case where the wave functions are complex.
Two matters relating to eq. (4.2.11) and some earlier equations involved in its derivation
call for comment.
We consider first the summations over the states r which carry the restriction r 6D i, f.
In principle all states including the initial state i and final state f should be included.
However, when the contributions from the states i and f are analysed they are found
to be zero in most situations. In the few rather special cases where they are non-zero
they can be shown to be very small relative to the terms with r 6D i, f. The arguments
underlying these statements have been detailed by Svendsen and Mortensen (1981) and
are summarized in Chapter A18. Thus for simplicity we shall always make the good
approximation that the initial and final states may be excluded from the summation over
the states r and indicate such exclusions with the summation sign as in eq. (4.2.11).
We now consider the relative signs of ir in the denominators. In many accounts
of the theory of Raman scattering the i terms are given the same sign. However, in
the classic work of Placzek (1934) the i terms have opposite signs and recent papers
by Buckingham and Fischer (2000) and Hassing, Mortensen and Svendsen (2000) have
shown that fundamental physical principles do require the i terms to have opposite
signs. As we shall see, the second term in eq. (4.2.11) is normally unimportant so that an
incorrect relative sign of i is often (but not invariably) of no consequence. However, it is
both curious and unsatisfactory that the use of incorrect relative signs of i has become
established in the literature. The relative signs of i are considered in more detail in
Chapter A18.
We shall see later in this chapter that if certain frequency conditions can be satisfied ir
can be ignored and then, of course, the transition polarizability becomes unambiguously
real for real wave functions and is given by

fi D
C
4.2.12
h r
ri 1
rf C 1
r6Di,f
Using this real transition polarizability we can express eq. (4.2.9) in the form
1

p fi D 12 fi EQ 0 1 exp is t C EQ 0 1 exp is t

4.2.13
It will prove convenient to introduce the component of the time-independent
complex transition moment amplitude pQ 1
and its conjugate complex pQ 1
and
0 fi
0 fi
write eq. (4.2.13) in the form
1

p fi D 12 pQ 1
exp is t C pQ 1
exp is t
0 fi
0 fi
4.2.14
An expanded nomenclature for the amplitude EQ 0 1 is introduced here, instead of EQ 0 used earlier, to
avoid ambiguity.
54
The Raman Effect
where
and
pQ 1
0

fi
pQ0 1

fi
D fi EQ 0 1
4.2.15
D fi EQ 0 1
4.2.16
When the electric field amplitude is real so that EQ 0 1 D EQ 0 1 D E0 1 the timeindependent transition moment amplitude is also real. Then pQ 1
D pQ 1
and may
0 fi
0 fi
1
be replaced by p0 fi . Thus eq. (4.2.14) becomes
p1

with

fi
D
1
2
p1
0
p1
0

fi

fi
fexp is t C exp is tg
D fi E0 1
4.2.17
4.2.18
If we transform eq. (4.2.17) to the trigonometric representation we obtain

p1

fi

D p1
cos s t
0 fi
4.2.19
or, introducing eq. (4.2.18),

1
p fi D fi E0 1 cos s t
4.2.20
The results of the quantum mechanical treatment are generally similar in form to those
obtained from the classical treatment described in Chapter 3, but with a transition electric
dipole and polarizability replacing an oscillating electric dipole and polarizability. The
electric field has the same classical form in both treatments.
In contrast to the classical polarizability, the transition polarizability is defined in terms
of the wave functions and energy levels of the system and thus it becomes possible,
at least in principle, to establish how the characteristics of the scattered radiation are
determined by the properties of the scattering molecules.
4.3 QUALITATIVE DISCUSSION OF ars f i

According to the definition of the general transition polarizability given by eq. (4.2.11)
a knowledge of the wave functions, energies and lifetimes of all the states of a system
would be needed to determine fi and at first sight this would appear to present a
formidable, if not insurmountable, barrier to further progress.
However, a closer examination will reveal that the problem becomes quite tractable
if certain simplifying assumptions are introduced. Happily, these assumptions can be
justified for realistic experimental conditions.
In this section we make a qualitative survey of the problem before proceeding to a
stage by stage quantitative analysis in subsequent sections.
55
4.3.1 Frequency denominators

We consider first the frequency denominators ri 1 ir which appear in the first
term in the definition of fi given by eq. (4.2.11). In obtaining this equation there
was no requirement that h 1 , the energy of a photon of the incident exciting radiation,
should bear any specific relation to any absorption energy h ri of the scattering molecule.
Also, no restriction was made about the energies of the states jri. Indeed, in principle the
state jri can lie above the final state jfi, below the initial state jii or between jii and
jfi as Fig. 4.1 illustrates. We shall however conduct the ensuing discussions assuming
that jri lies above jii and jfi which is normally the case. Extension to other cases is
straightforward. It is evident that the relative magnitudes of 1 and r will play an
important part in this denominator because their difference is involved. To explore this
role qualitatively we now consider two limiting cases. The first case is when the frequency
of the exciting radiation 1 is very much smaller than any absorption frequency ri of
the molecule, that is 1 ri for all r. Then ri 1 ri for all states jri and the
r can be neglected because they are small relative to the ri . The second case is when
1 is close to one or more particular absorption frequencies of the molecule ri . Then,
1 ri and ri 1 ir tends to ir for a particular state jri. Terms with such
denominators will dominate in the sum over r.
When 1 ri the foregoing perturbation treatment of Raman scattering is illustrated
by the energy level diagram in Fig. 4.2(a). Here the molecule is represented as interacting
with incident radiation of frequency 1 and making a transition from an initial stationary
state jii to a so-called virtual state from which it subsequently makes a transition to a final
stationary state jfi. This virtual state which is indicated by a broken line in Fig. 4.2(a) is
not a stationary state of the system. It is not a solution of a time-independent Schrodinger
equation and so does not correspond to a well-defined value of the energy. The process of
absorption without energy conservation which gives rise to such a state is called virtual
absorption.
r3
f
r2
hw1
h(w1wfi)
hwfi
r1
Figure 4.1 The states jri for a Stokes transition jfi

required by the Klein condition.
jii. Note that h 1 > h fi , as
56
The Raman Effect
w1
w1
w1
w1
(a) Normal
Raman
scattering
w1 << wri
(b) Pre-resonance
Raman
scattering
w1 wri
(c) Discrete
resonance
Raman
scattering
w1 wri
(d) Continuum
resonance
Raman
scattering
w1 in range of
dissociative continuum levels.
Figure 4.2 Four types of Raman scattering processes.
As 1 approaches a molecular transition frequency ri the Raman scattering process

is illustrated by the energy diagram of Fig. 4.2(b) and is termed pre-resonance Raman
scattering. When 1 ri the Raman process is illustrated in Fig. 4.2(c) and is termed
discrete resonance Raman scattering. If h 1 is large enough to reach the dissociative
continuum energy levels of the system the Raman process is illustrated in Fig. 4.2(d) and
is termed continuum resonance Raman scattering.
Turning now to the frequency denominator rf C 1 C ir in the second term in the
definition of fi given by eq. (4.2.11) we see that, because the sum of rf and 1
is involved, this denominator cannot become small and cannot lead to dominant terms in
the sum over r. Thus when the first term in fi has one or more dominant terms the
second term will be relatively unimportant.
The intensity of resonance Raman scattering can be expected to be orders of magnitude
greater than normal Raman scattering because, as already explained, the denominator of
the first term in the expression becomes very small, when 1 ri , but remains large if
1 ri . The second term in the expression for fi is unimportant relative to the
first term in the resonance case except for stimulated emission which does not concern
us here.
4.3.2 Transition electric dipole numerators

By way of illustration we consider a numerator in the first term in eq. (4.2.11), namely
hfjpO jrihrjpO jii. This is the product of two transition electric dipole terms, one for a
transition from a state jii to a state jri and one for a transition from from a state jri to a
57
|r
rprf
|f
rp^ si
|i
Figure 4.3 Linked electric dipole transitions. The product of their transition moments
constitutes the numerator of the first term in eq. (4.2.11). Here, r > f and i .
state jfi. Provided r is greater than i and f the first transition is an absorption and
the second an emission. Figure 4.3 illustrates such a pair of linked transitions.
For normal Raman scattering where 1 ri we can say in the language of timedependent perturbation theory that fi is determined by a weighted sum over the
states jri of the products hfjpO jrihrjpO jii, the weighting for each state jri being inversely
proportional to ri 1 ir . Thus normal Raman scattering can be regarded as
involving all possible pathways through states jri which connect the initial state jii and the
state jfi, a possible pathway being defined as one for which the electric dipole transition
moments between the state jri and both the state jii and the state jfi are non-zero.
In contrast, for resonance Raman scattering, the states jri for which 1 ri will
predominate in the sum over the states jri and fi will be determined by the properties
of a limited number of states jri, indeed, perhaps by just one such state.
We see that for normal Raman scattering, although the magnitude of fi is determined by the states jri, direct information regarding these states cannot be obtained.
Normal Raman scattering is essentially a ground electronic state property and it is the
states jii and jfi that play a determining role. However, in resonance Raman scattering
detailed information about the states jri and their lifetimes can be obtained and we shall
find that such scattering offers a valuable method of investigating excited vibronic and
rovibronic states.
Our discussion so far has centred largely on the first, potentially resonant, term in
the expression for fi . The second, non-resonant, term can be interpreted as relating
to a virtual process in which emission of the scattered photon precedes the absorption
of the incident photon. The two different time-orderings of the virtual absorption and
emission processes can be represented as in Fig. 4.4 using diagrams of the type first
introduced by Feynman (1962). In this volume we use Feynman diagrams just to illustrate
the time-ordered processes involved in Raman scattering. However, such diagrams, used
in association with a set of appropriate rules, can also provide a powerful method for
generating formulae for transition moments in multiphoton processes, but this aspect is
58
The Raman Effect

s
w1
i
f
s = w1 wfi
w1
Figure 4.4 Feynman diagrams for Raman scattering.
not considered here. Objections can be raised against applying time-ordering concepts
to the perturbation treatment of non-resonance Raman scattering because of the nature
of the virtual state. The purist might prefer the somewhat theological statement that the
several processes occur indivisibly in time. Such problems of semantics disappear when
resonance Raman scattering is being described.
4.3.3 Selection rules

It is evident from the definition of a component of a general transition polarizability tensor
given by eq. (4.2.11) that the xy component (xy fi will have the same transformation
properties under a molecular symmetry operation as
hfjxyjii
4.3.1
Thus the general condition for xy fi to be non-zero is that the product f xy i belongs
to a representation which contains the totally symmetric species. This symmetry condition
is discussed in detail for vibrational transitions in Chapter 5, Section 5.8.
4.4 TENSORIAL NATURE OF THE TRANSITION

POLARIZABILITY AND ITS SYMMETRY
Before proceeding to the various stages in the simplification of the general transition
polarizability we consider its tensorial nature and symmetry. The quantity fi defined
in eq. (4.2.11) is the component of the general transition polarizability tensor which
is a second-rank tensor. We shall refer to this tensor as a Rayleigh scattering tensor when
f is degenerate with i and fi D 0, and as a Raman scattering tensor when f 6D i and
fi 6D 0.
For the Raman scattering tensor we now introduce the symbol afi and denote the
corresponding tensor components by fi . For the Rayleigh scattering tensor we first
introduce a general symbol aii0 . When i0 is degenerate with i but i0 6D i the Rayleigh
scattering tensor is associated with transitions between degenerate levels of the scattering
59
system. A classic example of this is provided by atomic sodium vapour as the ground
state of sodium has a two-fold degeneracy. When i0 is degenerate with i and i0 D i we
shall use aii as the symbol for the Rayleigh scattering tensor. In this book we shall be
concerned almost entirely with aii which represents the most commonly occurring case.
We now consider the symmetry of the tensors aii , aii0 and afi with respect to
interchange of the subscripts and of their components. For real time-independent
wave functions the hermitian property of the transition electric dipole operator means
that
p ir D p ri
4.4.1
Using this property with eq. (4.2.11) it is straightforward to show that for the Rayleigh
scattering tensor aii
ii D ii
4.4.2
Thus the Rayleigh scattering tensor aii is symmetric. However, when i0 is degenerate
with i but i0 6D i, the Rayleigh scattering tensor aii0 can have an antisymmetric part.
For the Raman scattering tensor, however,
fi 6D fi
4.4.3
so that the Raman scattering tensor is not in general symmetric. It will transpire, however,
that the Raman scattering tensor can become symmetric provided certain additional
conditions are satisfied. These conditions are often met, or at least closely approximated to.
For some subsequent developments it will prove convenient to make a formal separation
of the scattering tensors into symmetric and antisymmetric parts. For example, for the
case of fi defined in eq. (4.2.11), using the algebraic identities
aX C bY D 12 a C bX C Y C 12 a bX Y
4.4.4
rf ri D if
4.4.5
and
we can rearrange the right-hand side of eq. (4.2.11) to obtain
fi D sfi C afi
4.4.6
where the superscripts s and a denote the symmetric and antisymmetric parts and
1
ri C rf
sfi D
2h r6Di,f ri 1 ir rf C 1 C ir
fhfjpO jrihrjpO jii C hfjpO jrihrjpO jiig
1
21 C if C 2ir
afi D
2h r6Di,f ri 1 ir rf C 1 C ir
fhfjpO jrihrjpO jii hfjpO jrihrjpO jiig
4.4.7
4.4.8
60
The Raman Effect
It is straightforward to show as a result of the hermitian property of the transition

electric dipole defined in eq. (4.4.1), that when f 6D i corresponding to Raman scattering sfi D sfi and afi D afi in accord with the labels they have been
assigned. However, for Rayleigh scattering with f D i, sii D sii but a D 0.
For Rayleigh scattering with f D i0 where i0 is degenerate with i but i0 6D i, there can be
both symmetric and antisymmetric parts of the Rayleigh scattering tensor.
For completeness we now extend these considerations to the case where the wavefunctions are complex. We then have complex transition tensor components which we denote
by Q fi . We define Q fi in terms of its real and imaginary parts, as follows:
Q fi D fi i0 fi
4.4.9
Both the real and imaginary parts of O fi can be separated similarly into symmetric
and antisymmetric parts. Thus in place of eq. (4.4.6) we now write
Q fi D Q sfi C Q afi D sfi C afi i0 sfi i0 afi
4.4.10
To simplify matters we restrict the formulae that follow to the case of normal Raman
scattering and set r D 0. Then we have
1
ri C rf
sfi D
2h r6Di,f ri 1 rf C 1
RehfjpO jrihrjpO jii C hfjpO jrihrjpO jii
1
21 C if
afi D
4.4.11
RehfjpO jrihrjpO jii hfjpO jrihrjpO jii

1
ri C rf
0 sfi D
4.4.12
ImhfjpO jrihrjpO jii C hfjpO jrihrjpO jii

1
21 C if
0 afi D
4.4.13
ImhfjpO jrihrjpO jii hfjpO jrihrjpO jii
4.4.14
Thus when f 6D i, corresponding to Raman scattering, both the real and imaginary
parts of the complex transition tensor can have symmetric and antisymmetric parts. For
Rayleigh scattering with f D i the situation is different. Using in place of eq. (4.4.1) the
general hermitian property
4.4.15
pO ir D pO ri
it follows that
RehijpO jrihrjpO jii D RehijpO jrihrjp jii
4.4.16
ImhijpO jrihrjpO jii D ImhijpO jrihrjpO jii
4.4.17
but

Using these relationships, aii D 0 and 0 sii D 0 but
2
ri
RehijpO jrihrjpO jii D sii
sii D
h r6Di ri2 12
2
1
0 aii D
ImhijpO jrihrjpO jii D 0 aii
2
h r6Di ri 12
61
4.4.18
4.4.19
Thus for Rayleigh scattering with f D i, the real part of the scattering tensor is pure
symmetric and the imaginary part pure antisymmetric; but for Rayleigh scattering with
f D i0 i0 6D i, the scattering tensor has the same symmetry properties as the Raman
scattering tensor.
4.5 BORNOPPENHEIMER APPROXIMATION AND THE

TRANSITION POLARIZABILITY TENSOR
We indicated in the introduction to this chapter that it is necessary to introduce a number
of simplifications to make the general formula for fi more tractable. In this section
and the subsequent ones we consider these simplifications in some detail. Throughout the
rest of this chapter we take the time-independent wave functions to be real.
The first and fundamental step in unravelling the general expression for fi given
by eq. (4.2.11) is to introduce the adiabatic approximation (Born and Oppenheimer, 1927)
which enables the electronic and nuclear motions to be separated.
We first expand the notation for a general jth electronicnuclear state so that its
electronic, vibrational and rotational parts are represented by their respective quantum
numbers ej , vj and Ri (or sets thereof) and write

4.5.1
jji D ej vj Rj
We also expand the notation for the energy of the state in a corresponding way, putting
j D ej vj Rj
The BornOppenheimer, or adiabatic, approximation enables us to set

jji D ej vj Rj
4.5.2
4.5.3
for the state jji and, for its energy,

ej vj Rj D ej C vj C Rj
4.5.4
The general state function is now a product of the separate electronic, vibrational and
rotational parts; and the energy is the sum of separate electronic, vibrational and rotational
j j j
parts. Note however that e1
j vj Rj refers to the lifetime of the state je v R i.
Here we express energy in units of rad s1 D E/h.

Less-relaxed approximations are needed in some situations. For example, the stricter approximation
jji D jej ijvi Ri i will encompass vibration rotation interaction.
62
The Raman Effect
We now apply the BornOppenheimer approximation to eq. (4.2.11) by introducing

eqs. (4.5.2) to (4.5.4). We express the result as

4.5.5
ef vf Rf :ei vi Ri D Rf vf ef O er , vr , Rr ei vi Ri
where
1
O e , v , R D
h
r
pO jer ijvr ijRr ihRr jhvr jher jpO

er ei C vr vi C Rr Ri 1 ier vr Rr
er vr Rr 6D ei vi Ri ,
ef vf Rf
pO jer ijvr ijRr ihRr jhvr jher jpO
er ef C vr vf C Rr Rf C 1 C ier vr Rr
4.5.6
Here O er , vr , Rr is the component of a polarizability tensor operator whose definition involves the states jer i, jvr i and jRr i as given in eq. (4.5.6); and ef vf Rf :ei vi Ri
is the component of the corresponding polarizability tensor associated with the transition from ei vi Ri to ef vf Rf . We could simplify eq. (4.5.6) by effecting closure over the
complete set of rotational states associated with each electronicvibrational level if the
denominators did not involve Rr Ri , Rr Rf and er vr Rr which are sensitive to the rotational
quantum numbers and so give rise to different weightings for each term in the sums

r
r
Rr jR ihR j.
Some reasonable assumptions enable us to eliminate this difficulty. Let us first assume
that the initial electronic state is a non-degenerate ground electronic state and put ei D eg .
We shall also assume that the er vr Rr may be neglected in the arguments that follow, but
we shall justify this later.
We now consider the denominator of the first term in eq. (4.5.6). If the electronic state
jer i is not the ground electronic state, that is er 6D eg , then, as the magnitudes of the
h Rr Ri are small relative to the vibronic energies, they may be neglected provided
er eg C vr vi 1 0
4.5.7
This inequality is satisfied if 1 is much smaller than any vibronic absorption frequency
er eg C vr vi (or electronic absorption frequency er eg if vr vi D 0). If the electronic state
er is the ground electronic state, that is er D eg , and hence er eg D 0, then the Rr Ri may
be neglected provided
4.5.8
1 vr vi C Rr Ri
These conditions on the frequency denominator in the first term in eq. (4.5.6) do not
involve the final state and so are independent of whether ef D eg or not.
We turn now to the frequency denominator of the second term in eq. (4.5.6). The Rr Rf
can be neglected if
er ef C vr vf C 1 0
4.5.9
This condition will always be met provided that 1 is sufficiently well removed from the
microwave region. This requirement is a necessary one if er ef D 0, as for example when
63
both the final and intermediate electronic states are the ground states. The requirement
that 1 is not small can be relaxed in some situations, for example if er ef is sufficiently
large.
For most situations met in practice, we may summarize the above conditions on the
frequencies as follows: 1 must be much smaller than any electronic absorption frequency
and much greater than any vibrational or rotational frequency of the material system.
Finally, we consider the er vr Rr terms in the denominators. Any dependence of these
terms on the rotational quantum numbers will be small and we can safely replace er vr Rr
by er vr . Furthermore, the magnitudes of the er vr themselves are only of the order of
tens of wavenumbers. Thus their inclusion will not affect the frequency conditions on 1
that have been imposed for the neglect of the Rr Ri and Rr Rf .
If the initial state of the system is not the ground electronic state similar arguments can
be developed for neglecting the dependence of the denominators on the rotational quantum
numbers. Thus if greater generality were needed we could revert to using ei instead of
eg . However it is realistic to continue to set ei D eg in subsequent developments.
Assuming the above frequency conditions are satisfied, we can now apply the closure
theorem to the numerators of the expression for O er , vr , Rr given by eq. (4.5.6) and
(with ei D eg ) obtain

p
1
pO jer ijvr ihvr jher jpO
O jer ijvr ihvr jher jpO
r r
O e , v D
C
h er vr 6D eg vi , er eg C vr vi 1 ier vr
er ef C vr vf C 1 C ier vr
ef vf
4.5.10
Thus (with ei D eg ) eq. (4.5.5) becomes

ef vf Rf :eg vi Ri D Rf vf ef O er , vr eg vi Ri
4.5.11
The next step is to transform the polarizability tensor operator O er , vr defined in
relation to space-fixed axes , . . . to O 0 0 er , vr defined in relation to molecule-fixed
axes 0 , 0 . . . , by using direction cosines l0 , l 0 . . . . After applying this transformation
to eq. (4.5.11) and making some rearrangements, we have

4.5.12
ef vf Rf :eg vi Ri D Rf l0 l 0 Ri vf ef O 0 0 er , vr eg vi
where O 0 0 er , vr is obtained from eq. (4.5.10) by replacing and by 0 and 0 , respectively. We see that the right-hand side of eq. (4.5.12) involves the product of two quite
distinct terms. This is because only the direction cosine operators can effect rotational
transitions and hvf j, hvi j, hef j, heg j and O 0 0 er , vr depend only on internal molecular
coordinates. The characteristic properties of normal rotational and vibrationrotation
Raman scattering can be obtained from a detailed analysis of eq. (4.5.12) and this is
dealt with in Chapter 6. In the case of resonance rotation and vibrationrotation Raman
scattering which is considered in Chapter 8, closure over the states Rr cannot be made.
It is necessary to revert to eqs. (4.5.5) and (4.5.6) and then er vr Rr will play a role.
Chapter A10, Section A10.3 deals with such tensor transformations.
64
The Raman Effect
Most Raman scattering is observed under experimental conditions in which the rotational structure is not resolved. Then the scattering may be described as associated with
transitions of the type eg vf
eg vi and is termed pure vibrational Raman scattering. In
treating such scattering, isotropic averages of intensity expressions are ultimately formed.
Such isotropic averaging is a purely classical process in which all molecular orientations
are assumed to be equally possible. This procedure can be shown to give results which are
identical with those obtained by using eq. (4.5.12) and summing the intensity expressions
over the complete set of initial and final rotational states.
We shall assume henceforth that for vibrational transitions with no resolved rotational
structure, and ei D eg , we may use

ef vf :eg vi D vf ef O er1 vr eg vi
4.5.13
instead of eq. (4.5.12) with the understanding that, in due course, isotropic averaging will
be performed if necessary. It is useful to introduce eq. (4.5.10) into eq. (4.5.13) and this
gives the following result:

1
hvf jhef jpO jer ijvr ihvr jher jpO jeg ijvi i
ef vf :eg vi D
h er vr 6D eg vi ,
er eg C vr vi 1 ier vr
ef vf
hvf jhef jpO jer ijvr ihvr jher jpO jeg ijvi i
C
4.5.14
This definition of the component of the transition polarizability will be the starting
point for all the subsequent simplifications considered in the sections that follow.
4.6 SIMPLIFICATION OF ars e f vf :e g vi : GENERAL

CONSIDERATIONS
We now consider how ef vf :eg vi defined in eq. (4.5.14) can be simplified. We shall
consider two different approaches. In the first, we introduce a number of quite radical
approximations, more or less simultaneously, and obtain relatively simple results. These
are necessarily of restricted validity. This procedure, which is considered in Section 4.7,
is essentially that used by Placzek (1934) and leads to results that are applicable to Raman
scattering for which ef D ei D eg , er eg 1 vr vi , and the ground electronic state is
non-degenerate.
The alternative approach, which is considered in Section 4.8, involves introducing the
approximations in stages. This yields a variety of special forms of ef vf :eg vi , which are
applicable to Raman scattering under the conditions consistent with the approximations
made in each case. This procedure is particularly important for the treatment of resonance
Raman scattering and electronic Raman scattering. If we pursue this approach to the limit
we arrive at the formulae given by the first procedure and also obtain additional insights
into the nature of the approximations used in the Placzek approach.
65
4.7 SIMPLIFICATION BY RADICAL APPROXIMATION: THE

PLACZEK TRANSITION POLARIZABILITY
We first confine ourselves to those cases for which ef D ei D eg ; that is, the initial and
final vibrational states with vibrational quantum numbers vi and vf , respectively, both
belong to the ground electronic state which we take to be non-degenerate. For these
conditions eq. (4.5.14) becomes

1
hvf jheg jpO jer ijvr ihvr jher jpO jeg ijvi i
eg vf :eg vi D
h er vr 6D eg vi ,
eg vf

C
er eg C vr vf C 1 C ier vr
4.7.1
We now split the above summation into two parts, one where er D eg and so er eg D 0,
and the other where er 6D eg , and write eq. (4.7.1) as
r
er 6Deg

eg vf :eg vi D eeg vDe
f :eg vi C eg vf :eg vi
It follows that
er Deg
eg vf :eg vi
1
D
h r i f
4.7.2
hvf jheg jpO jeg ijvr ihvr jheg jpO jeg ijvi i
vr vi 1
v 6Dv ,v

hvf jheg jpO jeg ijvr ihvr jheg jpO jeg ijvi i
C
vr vf C 1
4.7.3
where the vr vi and vr vf relate to vibrational levels in the ground electronic state, and
we have set er vr D eg vr D 0.
When er 6D eg , that is when the states jer i are excited electronic states, it is a good
approximation to neglect the vr vi and vr vf relative to the er gg provided that
er eg 1 vr vi or vr vf
4.7.4
This follows from consideration of eqs. (4.5.7) and (4.5.9). Under these conditions we
can also ignore the er vr because they are small compared to the er eg . Then we have

1
er 6Deg
eg vf :eg vi D
h er 6D eg ,
er eg 1
vr 6D vi vf
C
er eg C 1
4.7.5
66
The Raman Effect
In both eq. (4.7.3) and eq. (4.7.5) the electric dipole moment operator is a function of
both the set of electronic coordinates , and the set of nuclear coordinates Q, that is
pO D pO , Q
4.7.6
and the matrix elements are formed with complete adiabatic wave functions of the form
given in eq. (4.5.3). The set of nuclear coordinates Q is that particular set of internal
nuclear coordinates known as normal coordinates. These coordinates involve displacements from the equilibrium nuclear configuration which we denote by Q0 . Particular
members of the set of normal coordinates will be denoted by Qk , Ql . . . and are associated
with the kth, lth . . . vibrational frequencies, respectively.
To make further simplifications we introduce the adiabatic permanent electric dipole
moment and the adiabatic dynamic polarizability of the system in the lowest electronic
level, the nuclei being held fixed in a configuration Q so that only the electrons are free
to move. Both these adiabatic quantities are clearly functions of the nuclear coordinates
Q and will be denoted by pO Q and O Q, respectively. Introducing explicitly the
coordinate dependence of the wave functions, the adiabatic permanent electric dipole is
given by
pO Q D h 0 , Qjp,
O Qj 0 , Qi
4.7.7
and the adiabatic dynamic polarizability by

1 h 0 , QjpO , Qj
O Q D
h er 6Deg
C
O , Qj 0 , Qi
er , Qih er , Qjp
er eg 1
O , Qj er , Qih er , QjpO , Qj 0 , Qi

0 , Qjp
er eg C 1
4.7.8
Here j 0 , Qi corresponds to jeg i, j er , Qi to jer i and so on. It should be noted that
in both eq. (4.7.7) and eq. (4.7.8) the integration is over the electronic coordinates only.
Using eq. (4.7.7) we can write eq. (4.7.3) as follows:

f jp
r ihvr jp
ii
f jp
r ihvr jp
ii

r
g
h
v
O
Qj
v
O
Qj
v
v
O
Qj
v
O
Qj
v
1
h

eeg vDe
C
f :eg vi D
h r i f
vr vi 1
vr vf C 1
v 6Dv ,v
4.7.9
This is usually called the ionic part of the vibrational transition polarizability. It
describes Raman scattering which involves only virtual excited vibrational states, with
the molecule remaining in the ground electronic state throughout. Provided 1 is very
much rgreater
than molecular vibrational frequencies, this term can be ignored relative to
e 6Deg
( eg vf :eg vi .
Returning now to eq. (4.7.5) we can simplify this further by invoking the closure
theorem with respect to the complete set of vibrational states in each electronic state. We

then obtain
67

hvf jheg jpO jer iher jpO jeg ijvi i hvf jheg jpO jer iher jpO jeg ijvi i
C
er eg 1
er eg C 1
4.7.10
We now introduce the adiabatic dynamic polarizability defined in eq. (4.7.8), and
eq. (4.7.10) becomes

f
O Qvi
D
4.7.11
v
electronic
g
f
g
i
e v :e v
er 6Deg
eg vf :eg vi
1
D
h er 6Deg
The label er 6D eg has now been replaced by the label electronic because eq. (4.7.11)
may be described as defining the electronic part of the vibrational transition polarizability.
For normal conditions with 1 vr vi it is the only significant part of the vibrational
transition polarizability and so the distinguishing label, electronic, can be dropped. Also,
the transition-defining label can be conveniently shortened and written as vf vi . Thus

4.7.12
vf vi D vf O Qvi
it being understood that ef D ei D eg .
We see from eq. (4.7.12) that the components of the Placzek pure vibrational transition
polarizability are matrix elements of the adiabatic polarizability, defined by eq. (4.7.8),
which is a function of the nuclear coordinates Q only. This property will be used subsequently to establish the relationships between vf vi and the equilibrium dynamic
polarizability and its derivatives with respect to normal coordinates.
The Placzek pure vibrational transition polarizability given by eq. (4.7.12) in which the
wave functions are taken to be real has a valuable property not possessed by the general
transition polarizability given by eq. (4.5.5): it is symmetric with respect to interchange of
the indices and , and so has no antisymmetric part. Because of the frequency conditions
defined in eq. (4.7.4) the Placzek transition polarizability applies only to Raman scattering
by molecules which do not absorb in the visible region and where the Raman excitation
frequency is in the visible region. Such Raman scattering which is termed normal pure
vibrational Raman scattering, is treated in detail in Chapter 5.
The analogue of eq. (4.5.11) using the Placzek pure vibrational transition polarizability
defined in eq. (4.7.12) is given by

4.7.13
vf Rf :vi Ri D Rf O vf vi Ri
where the transitions are confined to the ground electronic state. Equation (4.7.13) can
also form the basis for a treatment of normal rotational Raman scattering. Following the
transformation procedure used to obtain eq. (4.5.12) we obtain

4.7.14
vf Rf :vi Ri D Rf 0 0 Ri 0 0 vf vi
where 0 0 vf vi is given by equation (4.7.12) with the molecule-fixed axes 0 , 0 replacing
the space-fixed axes . Equation (4.7.14) is the analogue of eq. (4.5.12).
Detailed analysis of eq. (4.7.14) provides the characteristic properties of normal rotational and vibrationrotation Raman transitions on the basis of the Placzek pure vibrational
transition polarizability.
68
The Raman Effect
For vibrational transitions with no resolved rotational structure, and ei D eg we drop

the rotational states, and write the vibrational Raman transition tensors using space-fixed
axes. Thus instead of eq. (4.7.14) we use

vf vi D vf O vi
4.7.15
and take isotropic averages to obtain the intensities. The reasons for this are the same as
those developed in Section 4.5 for the case of a more general transition polarizability.
4.8 SIMPLIFICATION OF ars e fv f :e ivi BY STAGES

Throughout the various subsequent stages of simplification we shall, as already indicated,
take the initial electronic state to be the ground electronic state, so that ei D eg . We shall
also assume the ground electronic state to be non-degenerate.
4.8.1 Introduction of HerzbergTeller vibronic coupling

For convenience we first rewrite the expression for the cartesian component of the
transition polarizability ef vf :eg vi given by eq. (4.5.14) in a somewhat more condensed
form:

1
hvf jp ef er jvr ihvr jp er eg jvi i
ef vf :eg vi D
h er vr 6D eg vi ,
ef vf
hvf j ef er jvr ihvr jp er eg jvi i

C
where

p ef er D ef pO p er
4.8.1
4.8.2
and so on. Here p ef er is the component of the pure electronic transition electric
dipole moment associated with the electronic transition jef i
jer i. In order to obtain
eqs. (4.5.14) and (4.8.1) the BornOppenheimer approximation was used and closure
over the states Rr effected. The BornOppenheimer approximation in the form expressed
by eqs. (4.5.3) and (4.5.4) neglects coupling between electronic and nuclear motions.
However, in reality the electronic transition moment has a slight dependence on the
normal coordinates of vibration Q which arises in the following way. The electronic
O e itself has a functional dependence on the normal coordinates of vibration
hamiltonian H
and this may be expressed by an expansion in the nuclear displacements around the
equilibrium configuration Q0 :

H
Oe
Oe
1
2 H
O e 0 C
O e Q D H
Qk C
Qk Ql C . . .
4.8.3
H
Q
2
Q
Q
k
k
l
k
k,l
0
69
where Qk , Ql . . . are the normal coordinates of vibration, that is displacements from the
equilibrium configuration. The subscript zero is used here and subsequently to indicate
that a value is taken at the equilibrium position. The second and higher terms in the
expansion of eq. (4.8.3) constitute a perturbation that can mix electronic states. If we
take as the electronic basis the set of electronic functions e , Q0 at the equilibrium
nuclear configuration, the normal coordinate dependence of the electronic functions may
be considered to arise as a result of the vibrational perturbation mixing the e , Q0 .
0
Thus we may set jer Qi equal to jer Q0 i where the superscript 0 denotes the perturbed
state resulting from the electronic hamiltonian mixing the unperturbed state jer Q0 i with
other unperturbed states jes Q0 i. When the displacements Qk are small, as for example
for transitions for which vf D vi C 1, the perturbation is adequately expressed by just the
second term in eq. (4.8.3). With this assumption, using standard perturbation theory we
have for the perturbed ket state
0
jer Q0 i D jer Q0 i C
1 heks er
Qk jes Q0 i
r
s
h es 6Der k e e
4.8.4
where heks er is a coupling integral defined by

heks er D h
O e /Qk 0 j er Q0 i
es Q0 jH
4.8.5
and the energies h er and h es relate to the energies of the unperturbed states. Similarly
0
for the perturbed bra state her Q0 j, since her j D jer i and heks er D hekr es we have
0
her Q0 j D her Q0 j C
1 s
hekr es
he Q0 j
Qk
h es 6Der k
er es
4.8.6
For real functions hekr es D heks er but it will prove convenient to keep the two forms distinct
as in eqs. (4.8.4) and (4.8.6) in the present development.
To simplify the notation we shall henceforth omit the normal coordinate label Q0 and
0
0
write jer i for jer Q0 i, jer i for jer Q0 i and so on.
We now examine how the introduction of such perturbations by the electric hamiltonian
affects the products of components of electronic transition dipoles that occur in the numerators of the two terms in the expression for ef vf :eg vi given by eq. (4.8.1). The first
0
0
0
0
product now becomes hef jpO jer iher jpO jeg i, where we have reverted to expanded nota0
0
tion. Using eq. (4.8.6) for her j and adapting eq. (4.8.4) for jeg i we obtain for the
component of the electronic transition dipole

1 s hekr es
1 hekt eg
r0
g0
r
g
t
he j
Qk jpO j je i C
Qk je i
he jpO je i D he j C
h es 6Der k
er es
h t g k eg et
e 6De
4.8.7
See page 66 for the distinction between Q, Q0 , Qk , Ql . . . .
70
The Raman Effect
Multiplying out the terms in eq. (4.8.7) and adapting the condensed nomenclature introduced in eq. (4.8.2) we have, to first order,
p er 0 eg0 D p 0er eg C
C
1 hekr es
Qk p 0es eg
h es 6Der k er es
hekt eg
1
p 0er et
Qk
h t g k
eg et
4.8.8
e 6De
where the superscript 0 on electronic transition dipoles indicates that unperturbed states
are involved as for example p 0er eg D her jpO jeg i
We see from eq. (4.8.8) that when perturbation by the electric hamiltonian is operative the transition electric dipole p er 0 eg0 involves not only p 0er eg but also additional contributions from transition electric dipoles of the types p 0es eg and p 0er et .
0
The former, which arises from the perturbed bra state her j, involves states jes i which
must be accessible by dipole transitions from jeg i; the latter, which arises from the
0
perturbed ket state jeg i, involves states jet i which do not have to meet this requirement. The three terms contributing to p er 0 eg0 and defined in eq. (4.8.8) are illustrated
in Fig. 4.5(a).
Proceeding similarly we obtain for the component of the electronic transition dipole
in the numerator of the first term in eq. (4.8.1)
p ef0 er 0 D p 0ef er C
1
heks er
p 0ef es
Qk
h es 6Der k
er es
1 hekf et
C
Qk p 0et er
h t f k ef et
4.8.9
e 6De
The transition dipoles of the type p 0ef es which arise from the perturbed bra state hef j
involve states jes i which must be accessible by dipole transitions from jef i. The transition
0
dipoles of the type p 0et er which arise from the perturbed ket state jer i involve states
jet i which do not have to meet this requirement. The three terms contributing to p ef0 er 0
and defined in eq. (4.8.9) are illustrated in Fig. 4.5(b).
The process of acquiring (or borrowing) additional contributions to transition dipoles
through vibronic coupling was first proposed by Herzberg and Teller (1933) and is
called after them. For the kth vibration the amount of such a borrowed contribution
is proportional to the appropriate coupling integral and the displacement Qk and inversely
proportional to the energy difference between the two states involved. The latter condition means that appreciable coupling requires the states involved to be relatively close in
energy. For this reason terms involving energy denominators like h er es are likely
to be more significant than those involving energy denominators like h eg et or
h ef et . Furthermore, there are symmetry conditions to be met. For example for the
O e /Qk Qk transforms in electron space as
coupling term heks er Qk to be non-zero, since H

|t
71
|t
|s
|t
( p )0eret
|s
|s
hkeres
|r
|r
|r
hketeg
( p )0ereg
( p )0eseg
|f
|g
|f
|f
|g
|g
(a)
Figure 4.5(a) The three contributions to p er0 eg0 which arise from HerzbergTeller
coupling of the states jsi and jri, and jti and jgi via Qk .
|t
|t
|s
|r
|t
(pr )0e te r
|s
hke ser
|s
|r
|r
hke fe t
(pr ) e fe r
(pr ) e fe s
|f
|f
|f
|g
|g
|g
(b)
Figure 4.5(b) The three contributions to p ef0 er0 which arise from HerzbergTeller
coupling of the states jsi and jri, and jti and jfi via Qk .
Qk , the following symmetry condition must be satisfied:

er Qk es 1
4.8.10
where is an irreducible representation or symmetry species. It should be noted that

this condition can be satisfied for both totally symmetric and non-totally symmetric
vibrations.
We use for irreducible representations and for halfwidth of a level.
72
The Raman Effect
When, in eq. (4.8.1), p ef er is replaced by p ef0 er 0 using eq. (4.8.9), p er eg is
replaced by p er 0 eg0 using eq. (4.8.8) and so on, a rather complicated result is obtained
which we shall express in the form
ef vf :eg vi D AI C BI C CI C DI
4.8.11
where
1
A D
h
BI D
1
h 2
er vr 6D eg vi ,
ef vf

er vr 6D eg vi ,
ef vf
hvf j
hvf jp 0ef er jvr ihvr jp 0er eg jvi i hvf jp 0ef er jvr ihvr jp 0er eg jvi i
C
er vr :eg vi 1 ier vr
er vr :ef vf C 1 C ier vr

heks er
0
fj
h
v
p

Qk jvr ihvr jp 0er eg jvi i
f
s

e e
r es
e
s
r
e 6De
p 0ef es
e 6De
er vr 6D eg vi ,
ef vf

es 6Der
1
h 2
heks er
Qk jvr ihvr jp 0er eg jvi i
er es
hkr s
ee
f jp 0 jvr ihvr j
h
v
Qk p 0es eg jvi i

f
r
e e
r es
e
s
r
hvf jp 0ef er jvr ihvr j
CI D
hekr es
Qk p 0es eg jvi i
er es

er vr 6D eg vi ,
ef vf
4.8.12
1
h 2

es 6Der

hekf et
hv j
Qk p 0et er jvr ihvr jp 0er eg jvi i
t
ef e
t
f k
e 6De
er vr :eg vt 1 ier vr
4.8.13

hekf et
0
r
r
0
i
f
hv j
Qk p et er jv ihv jp er eg jv i
t

e

e
t
f
k
e 6De
C
1
h 2

hekt eg
0
hvf jp 0f r jvr ihvr j
Qk jvi i
p

r
t

ee
e e
g et
e
t
g k
e 6De
er vr 6D eg vi ,
ef vf

hvf jp 0ef er jvr ihvr j
et 6Deg
DI D
4.8.14
e 6De
er vr 6D eg vi ,
ef vf
hvf j
hekt eg
i
Qk jv i
g
t
e e
0
k

hkr s0
h
s er
e
0
r
r
e
e
f
hv j
pef es
Qk jv ihv j
Qk 0 p 0es0 eg jvii
0
r
s
r

e
e
e
es 6Der k
es
s0
r k0
p 0er et
1
h 3
73

0
0
es 6Der k
0
p 0ef es0
heks0 er
er es0
Qk 0 jvr ihvr j

es 6Der
hekr es
Qk p 0es eg jvi i
er es
4.8.15
In BI the excited electronic states jes i involved in the HerzbergTeller coupling must
be accessible by allowed dipole transitions from jeg i or jef i. In CI , however, the excited
electronic states involved in the HerzbergTeller coupling do not have to meet this requirement. To emphasize the distinction, the electronic states involved in the HerzbergTeller
coupling have been designated jet i in CI .
It should be noted that vr refers to a vibrational state in the excited electronic state
r
e . Likewise vf is associated with ef which is not necessarily eg . To keep the notation simple we have written AI for AIef vf :eg vi and omitted component labels; similarly
74
The Raman Effect
for BI , CI and DI . We shall continue this practice in subsequent developments except

when it becomes expedient to revert to a fuller notation to emphasize the transitions
involved.
We see from eqs. (4.8.12) to (4.8.15) that the two constituent terms of AI each involve
products of two unperturbed electronic transition dipoles. The four constituent terms of
BI each involve products of one unperturbed electronic transition dipole with electronic
transition dipoles arising from HerzbergTeller coupling of two excited electronic states.
The four constituent terms of CI each involve products of one unperturbed electronic
transition dipole with electronic transition dipoles arising from HerzbergTeller coupling
of one excited electronic state and either the initial or final electronic state. The two
constituent terms of DI each involve products of two electronic transition dipoles, both
of which arise from HerzbergTeller coupling. Here coupling of one excited electronic
0
state jer i or two different excited electronic states jes i and jes i is involved.
We may simplify these equations somewhat because the electronic transition moments
and the integral hekr es do not, under the conditions for which the BornOppenheimer
approximation is valid, operate on the vibrational wavefunctions. Therefore we may
separate the vibrational terms from the electronic terms and write
IIef vf :eg vi D AII C BII C CII C DII
where
1
AII D
h
BII D
1
h 2

er vr 6D eg vi ,
ef vf
p 0ef er p 0er eg

p 0ef er p 0er eg
C

er vr 6D eg vi , es 6Der
ef vf
4.8.16

hvf jvr ihvr jvi i

4.8.17
p 0ef es heks er p 0er eg

er es er vr :eg vi 1 ier vr

p 0ef es heks er p 0er eg
hvf jQk jvr ihvr jvi i
C
r
s
r
r
e e er vr :ef vf C 1 C ie v

p0 ef er hekr es p 0es eg
1
C 2
er es er vr :eg vi 1 ier vr
h er vr 6D eg vi , es 6Der k
ef vf

p 0ef er hekr es p 0es eg
C
hvf jvr ihvr jQk jvi i
er es er vr :ef vf C 1 C ier vr

hekf et p 0et er p 0er eg
1
CII D 2
ef et er vr :eg vi 1 ier vr
h er vr 6D eg vi , t f g k
ef vf
e 6De e
4.8.18

hekf et p 0et er p 0er eg
C
hvf jQk jvr ihvr jvi i
r
r
t
ef e er vr :ef vf C 1 C ie v

p 0ef er p 0er et hekt eg
1
C 2
eg et er vr :eg vi 1 ier vr
h er vr 6D eg vi , t f g k
75
e 6De e
ef vf

p 0ef er p 0er et hekt eg
C
hvf jvr ihvr jQk jvi i
4.8.19
eg et er vr :ef vf C 1 C ier vr

0

p 0ef es heks er hkr s0 p 0s0 g hvf jQk jvr ihvr jQk 0 jvi i
1
II
e e
e e
D D 3
h er vr 6D eg vi , s s0 r k,k 0 er es er es0 er vr :eg vi 1 ier vr
ef vf
e ,e 6De
p 0ef es0 heks0 er hekr es p 0es eg hvf jQk 0 jvr ihvr jQk jvi i
er es0 er es er vr :ef vf C 1 C ier vr
4.8.20
Figure 4.6 illustrates the transition electric dipoles and HerzbergTeller couplings
involved in the numerators of AII , BII , CII and DII .
4.8.2 Identification of non-resonance and resonance situations

Further simplification of eqs. (4.8.17) to (4.8.20) is only possible if we make specific
assumptions regarding the relationship of the vibronic absorption frequencies er vr :eg vi , or
equivalently er eg C vr vi , to 1 the frequency of the exciting radiation.
If the frequency of the exciting radiation is well removed from any electronic
absorption frequencies, that is er eg C vr vi 1 for all er and vr we have nonresonance or normal Raman scattering (see Fig. 4.2a). We shall in due course consider
two special cases of normal Raman scattering: when ef D eg so that only vibrational
transitions in the ground electronic state are involved; and when ef 6D eg so that electronic and vibronic transitions from the ground electronic state are involved. We shall
refer to these cases as normal vibrational Raman scattering and normal electronic (and
vibronic) Raman scattering, respectively, or simply vibrational Raman scattering and
electronic (and vibronic) Raman scattering when it is clear that resonance is excluded.
If the frequency of the exciting radiation lies within the contour of an electronic
absorption band an entirely different situation arises (see Fig. 4.2c). As 1 approaches
some particular vibronic absorption frequency, er eg C vr vi , the excited electronic state
involved will dominate the sum over states in the expression for the transition polarizability. Also, in each of the eqs. (4.8.17) to (4.8.20) those terms which have er vr :eg vi
1 ier vr in their denominator will be most significant and the other terms with
er vr :ef vf C 1 C ier vr in their denominators may be neglected. Scattering under these
conditions is called resonance Raman scattering. We shall in due course consider two
76
The Raman Effect

|t
|t
|s
|t
|s
|s
hkeser
hkeres
|r
(ps )0ereg
|r
( ps)0e re g
( p r ) e fe r
|r
(pr) e fe s
(ps) eseg
(p) e fe r
|f
|f
|f
|g
|g
|g
(a) (pr)0e fer(ps)0e reg

(first term of AII)
(b) (pr)0e feshke ser(ps)0e reg

(first term of BII)
|t
(c) (pr)0e ferhke res(ps)0e seg

(third term of BII)
|t
|t
(ps )0eret
( pr )0e te r
|s
|s
hkeres
|s
hkeser
|r
|r
|r
h e f et
h e teg
(ps )0e re g
(d) hek fet(pr)0e ter(ps)0e reg

(first term of CII)
(pr ) e fer
(ps ) eseg
(pr ) e fe s
|f
|f
|f
|g
|g
|g
(e) (pr)0e fer(ps)0e ret hke teg

(third term of C II)
(f) (pr)0e feshke serhke res (ps)0e seg

(first term of DII)
Figure 4.6 The transition electric dipole products which occur in the numerators of selected
terms of AII , BII , CII and DII i.e. eqs. (4.8.16) to (4.8.20).
special cases of resonance Raman scattering: when ef D eg so that only vibrational

transitions in the ground electronic state are involved; and when ef 6D eg so that electronic and vibronic transitions from the ground electronic state are involved.
We shall refer to these two cases as vibrational resonance Raman scattering and electronic (and vibronic) resonance Raman scattering, respectively.
We now proceed to derive modified forms of eqs. (4.8.17) to (4.8.20) which relate to the
four special cases mentioned above: normal electronic (and vibronic) Raman scattering;
77
normal vibrational Raman scattering; electronic (and vibronic) resonance Raman scattering; and vibrational resonance Raman scattering. In subsequent chapters these cases
and also normal and resonance rotational and vibrationrotation Raman scattering are
treated in more detail.
4.9 NORMAL ELECTRONIC (AND VIBRONIC) RAMAN

SCATTERING
We now consider more fully the considerable simplifications that can be made in eqs. (4.8.17)
to (4.8.20) when 1 , the frequency of the exciting radiation, is far removed from any
vibronic absorption frequency, a situation which we shall designate as non-resonant. This
means that the following inequality holds: er eg C vr vi 1 . As a consequence, er eg C
vr vi 1 is insensitive to the vr vi and may be replaced by er eg 1 . Also provided
1 r f , er ef C vr vf C 1 can be replaced by er ef C 1 ; this latter simplification
does not require that er eg 1 . Further, as er eg 1 and er ef C 1 are now much larger
than the damping factor ier vr this factor may be neglected.
An important consequence of these simplifications in the frequency denominators is
that closure over the set of vibrational states in each electronic state er becomes allowed
because the denominators are not vr dependent.
In contrast to the resonance case which we consider later, a large number of excited
electronic states s will contribute to the BII , CII and DII terms of eqs. (4.8.17) to (4.8.20)
through the matrix elements of the type heks er defined in eq. (4.8.5). The relative magnitudes
of these contributions cannot be ascertained with any facility and it is therefore more
realistic in the non-resonance case to use Taylor series expansions of p ef er etc. which
are less explicit than that used in eq. (4.8.7), and write instead

p ef0 er 0 D p 0ef er C
p kef er Qk
4.9.1
k
where
p kef0 er 0
p ef er
Qk
4.9.2
0
This derivative with respect to the normal coordinate Qk implicitly contains the vibronic
coupling of states jer i to other electronic states but does not define it, explicitly. However
by comparing eqs. (4.9.1) and (4.9.2) with (4.8.9) we see that
k

k

h
1
h
s
r
ee
ef et
p 0ef es
C
p 0et er
4.9.3
p kef0 er 0 D
h s r
er es
ef et
t
f
e 6De
e 6De
Corresponding relationships for p kef0 er 0 , p ker 0 eg0 and p ker 0 eg0 are readily obtained.
We now introduce into eqs. (4.8.17) to (4.8.20) the simplifications we have just
discussed namely neglect of vr vi , vr vf and ier vr , closure over vibrational states, and
use of equations of the type (4.9.1) instead of eqs. (4.8.8) and (4.8.9). We then have
ef vf :eg vi D AIII C BIII C CIII C DIII
4.9.4
78
The Raman Effect
Energy
Final vibronic state
Initial vibronic state
(a)
(b)
Figure 4.7 Energy level diagrams: (a) normal pure electronic Raman scattering; (b) normal
pure vibronic Raman scattering.
where
III
1
D
h r g f
e 6De ,e
III
III
CC
DIII
p 0ef er p 0er eg

p 0ef er p 0er eg
C
er eg 1
er ef C 1
1
D
h r g f k
hvf jvi i
4.9.5
p kef er p 0er eg C p 0ef er p ker eg

er eg 1
e 6De ,e

p 0ef er p ker eg C p kef er p 0er eg
C
vf jQk jvi
4.9.6
er ef C 1

0
0
1 p kef er p ker eg p kef er p ker eg
D
C
hvf jQk Qk 0 jvi i
h r g f k,k 0
er eg 1
er ef C 1
e 6De ,e
4.9.7
Equations (4.9.5) to (4.9.7) relate to normal electronic (and vibronic) Raman scattering
6D eg . Figure 4.7 gives energy level diagrams for these processes which are considered in detail in Chapter 9. It will emerge that the relationship of vf to vi depends on the
properties of the excited electronic states jer i, jeg i and jef i.
ef
4.10 NORMAL PURE VIBRATIONAL RAMAN SCATTERING

For pure vibrational Raman scattering both the initial and the final electronic states are
the ground electronic state and so ef D eg . When this condition is introduced, eq. (4.9.4)
is replaced by
4.10.1
vf vi D AIV C BIV C CIV C DIV
79
where we have simplified the nomenclature by putting

eg vf :eg vi D vf vi
and
AIV
BIV C CIV
DIV
4.10.2

p 0eg er p 0er eg f i
p 0eg er p 0er eg
C
4.10.3
v v
er eg 1
er eg C 1

1 p keg er p 0er eg C p 0eg er p ker eg
D
h er 6Deg k
er eg 1

p 0eg er p ker eg C p keg er p 0er eg f i
C
4.10.4
v Qk v
er eg C 1

0
0

1 p keg er p ker eg
p keg er p ker eg f
D
C
v Qk Qk 0 vi
h er 6Deg k,k 0
er eg 1
er eg C 1
4.10.5
1
D
h er 6Deg
Since at this stage of the development we are taking the wave functions to be real,
it follows from the hermitian properties of the electric dipole interaction hamiltonian
operator that the two numerators in AIV are equal; similarly the two numerators in BIV C
CIV are equal, as are the two numerators in DIV . Thus, as the denominators in AIV ,
BIV C CIV , and DIV are of the form a C b and a b we may use the algebraic identity
1
2a
1
C
D 2
4.10.6
a C b a b
a b2
to factor out the frequency dependence in eqs. (4.10.3) to (4.10.5) and rewrite them in
the alternative forms

1 2er eg
0
0
vf vi
p

p

4.10.7
AIV,R D
eg er
er eg
h er 6Deg e2r eg 12

1 2er eg
k
0
0
k
vf Qk vi
p

p

C
p

p

BIV,R C CIV,R D
g
r
r
g
g
r
r
g
e e
ee
e e
ee
h er 6Deg k e2r eg 12
4.10.8

2er eg

1
0
p keg er p ker eg vf Qk Qk 0 vi
4.10.9
DIV,R D
h er 6Deg kk 0 e2r eg 12
where the superscript label R has been added to emphasize that these formulae apply only
when the wave functions are real.
It will prove useful for later developments to relate the terms AIV , BIV C CIV and DIV
to the adiabatic polarizability and its derivatives at the equilibrium configuration, Q D 0.
If we expand Q as a Taylor series in Q we obtain

1 2
Qk C
Qk Ql C 4.10.10
Q D 0 C
Qk 0
2 k l
Qk Ql
k
0
80
The Raman Effect
where to simplify the notation the explicit dependence of ( on Q has been omitted
on the right-hand side of the equation. Then introducing eq. (4.10.10) into eq. (4.7.12)
we have

f i
f i
vf vi D 0 v jv C
v Qk v
Q
k
0
k

2

f
1
C
4.10.11
v Qk Ql vi C
2 k l
Qk Ql
0
Comparing eq. (4.10.11) with eqs. (4.10.3) to (4.10.5) we see that

1 p 0eg er p 0er eg
p 0eg er p 0er eg
C
0 D
h er 6Deg
er eg 1
er eg C 1

1 p keg er p 0er eg C p 0eg er p ker eg

D
Qk 0 h er 6Deg
er eg 1

p 0eg er p ker eg C p keg er p 0er eg
C
er eg C 1

1 p keg er p ker eg
p keg er p ker eg
2
D
C
h er 6Deg
er eg 1
er eg C 1
Qk2
4.10.12
4.10.13
4.10.14
The expressions for the first and second derivatives in eqs. (4.10.13) and (4.10.14), respectively, are given for only one normal coordinate Qk . They relate to transition polarizability
components for a specific transition in which only one vibrational quantum number, vk changes. For such transition polarizability components we introduce the notation [v]vf :[v]vi
k
k
where [v] represents all the vibrational quantum numbers that remain unchanged.
It should be noted that the expression for the second derivative given in eq. (4.10.14)
is correct only to the first order. If the Taylor series expansion of the electronic transition
dipole moment is taken to the second derivative then eq. (4.9.1) must be replaced by

1
p ef er D p 0ef er C
p kef er Qk C
p kl
4.10.15
e f e r Qk Ql
2
k
k
l
where p kl
implicitly contains the vibronic coupling of jer i to other electronic states
ef er
which involve this second derivative. Additional terms will then appear in eq. (4.10.14).
When these contributions are taken into account, eq. (4.10.14) must be replaced by

kk
1 p keg er p ker eg C p kk
2
eg er p er eg C p eg er p er eg
D
h er 6Deg
er eg 1
Qk2

kk
p keg er p ker eg C kk
eg er p er eg C p eg er p er eg
C
4.10.16
er eg C 1
where again only one normal coordinate Qk has been considered.
81
As we are considering only real wave functions it follows from the hermitian properties
of the electric dipole interaction hamiltonian operator that the terms AIV , BIV C CIV and
DIV in eqs. (4.10.3) to (4.10.5) and AIV,R , BIV,R C CIV,R and DIV,R in eqs. (4.10.7) to
(4.10.9) are symmetric with respect to interchange of the indices and . This is also the
case for 0 defined by eq. (4.10.12) and the derivatives in eqs. (4.10.13), (4.10.14)
and (4.10.16).
The non-resonant pure vibrational transition polarizability determines the properties
of Rayleigh scattering and normal vibrational Raman scattering such as selection rules,
polarization and intensities. These are considered in detail in Chapter 5. The energy level
diagram for this process is given in Fig. 4.2a.
4.11 ELECTRONIC (AND VIBRONIC) RESONANCE RAMAN

SCATTERING
When the excitation frequency lies within the contour of an electronic absorption band
the approximations used in the treatment of normal scattering no longer apply. The
dependence of the frequency denominator on the vibrational quantum numbers cannot
be neglected; the frequency difference er eg C vr vi 1 is very small and comparable
with the damping factor er vr and so this factor cannot be disregarded; and the explicit
dependence of the first-order term in the HerzbergTeller expansion on the excited states
jes i must be taken into account.
The sensitivity of the frequency denominator to changes in the vibrational quantum
numbers vr means that closure over the vibrational states, as was used in going from
the eqs. (4.8.17) to (4.8.20) to the eqs. (4.9.5) to (4.9.7) is not possible. As a result the
transition polarizability is no longer a property of the ground electronic state only.
However, other simplifications are possible. As 1 approaches some particular electronic absorption frequency er eg C vr vi the excited electronic state involved will dominate the sum over states in the expression for the transition polarizability. It is therefore
usually adequate to consider one or at most two electronic manifolds in treatments of
resonance Raman scattering. Also under resonant conditions, only the first term in each
of eqs. (4.8.17) to (4.8.20) will be significant. The second term contributes only a slowly
varying background and may be neglected. Further simplifications are also possible. We
can restrict the sums over vibrational states to one sum over the states jvr i in one excited
electronic state jer i. Where HerzbergTeller coupling is involved we need consider only
a single second excited state jes i in the B term (or jet i in the C term) and only two
0
excited states jes i and jes i (where s0 may be equal to s) in the D term. Similarly, we
can restrict the summation to a single normal coordinate Qk in the B and C terms and
a pair of normal coordinates Qk and Qk 0 (where k and k 0 may be the same) in the D
term.
Introducing these simplifications into eqs. (4.8.17) to (4.8.20) we have, for the resonance situation,
ef vf :eg vi D AV C BV C CV C DV
4.11.1
82
The Raman Effect
where
hvff jvrr ihvrr jvig i
1
AV D p 0ef er p 0er eg
h
vr
BV D
hvff jQk jvrr ihvrr jvig i

1
heks er
0
0
p

p

r eg
f
s

e
e e r s
h 2
e
e
vr
C
CV D
4.11.2
hvff jvrr ihvrr jQk jvig i

1
hekr es
0
0
p

p

s
g
ef er
ee
er es
h 2
vr
4.11.3

1 hekf et
0
0
p

p

t
r
r
g
ee
ee
h 2 ef et
vr
hekt eg hvff jvrr ihvrr jQk jvig i

1
0
0
p

p

r
t
f
r

ee
e e
eg et vr er vr :eg vi 1 ier vr
h 2
4.11.4
k0
hvff jQk jvrr ihvrr jQk 0 jvig i

hes er her es0
1
D D 3 p 0ef es
p 0es0 eg
er es er es0
h
vr
k
4.11.5
0
In these equations the exclusions er 6D eg , ef; es , es 6D er and ef, eg 6D et apply, and the
vibrational quantum numbers, vi , vr and vf have been given additional superscripts as
for example vig to indicate the electronic state to which they refer.
Equations (4.11.1) to (4.11.5) form the basis for the treatment of resonance vibronic
Raman scattering which is considered in some detail in Chapter 9. An energy level
diagram for this process is given in Fig. 4.8.
Energy
Resonant vibronic state
Final vibronic state
Initial vibronic state
Figure 4.8 Energy level diagram for resonance vibronic Raman scattering.
83
4.12 VIBRATIONAL RESONANCE RAMAN SCATTERING

Normal vibrational resonance Raman scattering occurs when both the initial and final
electronic states are the ground electronic state, that is ef D eg . When this restriction on
ef is applied to eqs. (4.11.1) to (4.11.5) we obtain
eg vf :eg ei D AVI C BVI C CVI C DVI
4.12.1
where
rr
hv
1
jvrr
k ihvk jvk i
k
D p 0eg er p 0er eg
r r
h
r er vr :eg vi 1 ie vk
fg
VI
vk
ig
4.12.2
rr
hv
1
heks er
jQk jvrr
0
0
k ihvk jvk i
k
p

p

g
s
r
g
e e
ee
er es
er vr :eg vi 1 ier vrk
h 2
vr
k
k
fg
BVI D
ig
rr
hv
1
jvrr
0
0
k ihvk jQk jvk i
k
p

p

eg er
es eg
er es
er vrk :eg vki 1 ier vrk
h 2
vr
fg
hekr es
ig
4.12.3
CVI D
hekg et
1
p 0et er p 0er eg
h 2 eg et
C
hvfg jQk jvrr ihvrr jvig i

k
vrk
fg
ig
rr
hekt eg hvk jvrr
1
0
0
k ihvk jQk jvk i
p

p

g
r
r
t

e e
ee
eg et r er vr :eg vi 1 ier vrk
h 2
v
k
k
4.12.4
DVI D
0
heks er hekr es0
ig
hvfg jQk jvrr ihvrr
1
0
0
k
k
k 0 jQk 0 jvk 0 i
p

p
eg es
es 0 eg
er es er es0
h 3
vrk ,vrk 0
k
k
4.12.5
0
In these equations the exclusions er 6D eg , es , es 6D er and et 6D eg apply, and the

vibrational quantum numbers have again been given additional superscripts to show the
electronic state to which they refer.
Equations (4.12.1) to (4.12.5) constitute the basis for the detailed treatment of vibrational resonance Raman scattering which is given in Chapter 7. The energy level diagram
for this process is given in Fig. 4.2(c).
4.13 UNITS AND ORDERS OF MAGNITUDE

It follows from eq. (4.2.11), for example, that the essential nature of a general transition
polarizability tensor afi can be expressed by writing
^
afi D
transition dipole2
transition energy
4.13.1
84
The Raman Effect

^
(The symbol D means corresponds to.) As a transition dipole has units of C m, and a
transition energy units of C V, the transition polarizability fi has units of C V1 m2 .
This is in accord with the units for a component of a polarizability tensor derived from
the classical equation for an induced dipole. A detailed treatment of the classical induced
molecular dipole, including alternative units, is given in Chapter A13, Section A13.5.
Some authors prefer to use what may be termed a reduced transition tensor for which
we adopt the symbol a . The essential nature of this may be expressed by writing
fi
a
fi
dipole length2
transition wavenumber
4.13.2
If the dipole length has units of m and the transition wavenumber units of m1 (the SI
unit), then the reduced transition polarizability afi has units of m3 . It follows straightforwardly that the relation between the two forms is
e2
a
4.13.3
hc0 fi
where e is the proton charge and c0 the speed of light in vacuum. The value of the quantity
e2 /hc0 is 1.292 1013 C V1 m1 .
The above considerations may be used to obtain orders of magnitude for transition
polarizabilities and reduced transition polarizabilities. A typical value of a transition dipole
is 1.0 1030 C m (corresponding to a dipole length of 6.2 1012 m as e D 1.602
1019 C); and in a colourless system a typical value of an electronic transition energy
is 8.0 1019 C V. (corresponding to a transition wavenumber of 4.0 106 m1 ). Using
these values we find that a transition polarizability would have a value of 1.25 1042 ,
C V1 m2 . The corresponding value of the reduced transition polarizability is 9.67
1030 m3 . Using the value e2 /hc0 D 1.3 1013 C V1 m1 we find that the two values
are self-consistent.
A typical experimental value of fi obtained from measurements on the Q branch in
the vibration-rotation Raman spectrum of the nitrogen molecule is 5 1042 C V1 m2 .
afi D
REFERENCES
Born, M., and Oppenheimer, J. R. (1927). Ann. Phys. 84, 457.
Buckingham, A. D. and Fischer, P. (2000). Phys. Rev. A, 61, 035801.
Feynman, R. (1962). Quantum Electrodynamics, W.A. Benjamin: New York.
Hassing, S., Mortensen, O. S. and Svendsen, E. N. (2000). Proceedings of XVIIth Conference on
Raman Spectroscopy, Shu-Lin Zhang & Bang-Fen Zhu (eds.), 110. John Wiley: Chichester.
Herzberg, G. and Teller, E. (1933). Z. Phys. Chem. B21, 410.
Klein, O. (1927). Z. Physik 41, 407.
Placzek, G. (1934). Rayleigh-Streuung und Raman-Effekt, in Handbuch der Radiologie, E. Marx,
(ed.), 6, 205-374, Academische Verlag: Leipzig.
85
5
Next when I cast mine eyes and see
That brave vibration each way free;
O how that glittering taketh me!
Robert Herrick
5.1 INTRODUCTION
In this chapter we consider the characteristic properties of pure vibrational normal Raman
scattering. As explained in Chapter 4 these properties are determined by the Placzek pure
vibrational transition polarizability avf vi with isotropic averaging where appropriate.
Normal Rayleigh scattering corresponds to a special case with vf D vi .
For the treatment to be developed in this chapter it will only be necessary to regard
avf vi as a symmetric tensor whose components for a vibrational transition from vi to
vf are a function of the vibrational coordinates. In the absence of resonance the explicit
dependence of the tensor components on the energy levels of the scattering system need
not be considered. We shall therefore only be concerned in this chapter with geometrical
and tensorial relations and the properties of vibrational wave functions.
We begin with a brief recapitulation of the definition and properties of avf vi so that
this chapter can stand on its own. This should prove helpful to those more interested in
the application of the formulae developed here than in their detailed derivations.
Next, we set the scene by defining appropriate illuminationobservation geometries and
establishing the necessary nomenclature. We then obtain formulae for scattered intensities,
scattering cross-sections and polarization properties for various illuminationobservation
geometries and for various states of polarization of the incident radiation. The case of a
general vibrational transition from vi to vf is treated first; and then various special cases,
86
The Raman Effect
based on the harmonic oscillator approximation, are considered. It must be emphasized

that vi and vf represent the sets of vibrational quantum numbers in the initial and final
vibrational states respectively; and a general vibrational transition can involve changes in
one or more of the initial vibrational quantum numbers. A notation for such changes is
developed later in section 5.7.
An alternative method of treating the relationship between the scattered radiation and
the incident radiation is to use the Stokes parameters. General formulae can then be
obtained which cover all illuminationobservation geometries and all states of polarization of the incident radiation. Such formulae are enormously complicated but explicit
expressions for particular geometries can be extracted from them and these have acceptable
forms. The formulation in terms of Stokes parameters is considered later in Section 5.6.
To conclude this chapter we consider the vibrational selection rules and the vibrational
wavenumber patterns to which they give rise.
Before proceeding, we recall that tensors in general are treated in Chapter A10 and the
polarizability tensor is considered in detail in Chapter A14. The concepts and formulae
developed there are used here without detailed comment.
5.2 THE PLACZEK VIBRATIONAL TRANSITION

POLARIZABILITY: RECAPITULATION
5.2.1 Cartesian basis
In the cartesian basis, for a general vibrational transition from vi to vf the relation between
the amplitude of the induced electric dipole moment vector, c p0 and the amplitude of
the electric field vector, c E0 of the incident electromagnetic radiation is given in tensor
form by
c p0 vf vi D c avf vi c E0
5.2.1
p0 vf vi D vf vi E0
5.2.2
and in component form, by
Here, c avf vi is the Placzek vibrational transition polarizability tensor in the cartesian
basis, and vf vi its component which is defined as follows
vf vi D hvf jO Qjvi i
with
1
O Q D
r6Dg
heg jpO jer iher jpO jeg i heg jpO jer iher jpO jeg i
C
er eg 1
er eg C 1
where Q is a vibrational normal coordinate.
Chapter A21, Section A21.2.3 treats Stokes parameters.
5.2.3
5.2.4
87
We recall from Chapter 4 that the definitions given in eqs. (5.2.3) and (5.2.4) are valid
only if the following conditions are fulfilled:
1.
2.
3.
4.
ef D ei D eg ;
er eg 1 vr vi and vr vf ;
O Q is a function of the nuclear coordinate Q only;
the ground electronic state is non-degenerate.
We also note that at this stage we take the wave functions to be real.
We must also recall that to obtain eq. (5.2.3) rotational transitions have been factored
out. Thus the intensity formulae for a system of freely rotating molecules must involve
isotropic averages of the squares of the transition polarizability tensor components as
for example hxx 2vf vi i and hxy 2 ivf vi . The use of such isotropic averages gives results
identical with those that would be obtained by not factoring the rotational transitions and
ultimately summing over the complete set of initial and final rotational states.
For ensuing developments it will be helpful to set out in full the three time-independent
linear equations implicit in eqs. (5.2.1) and (5.2.2):
px0 D xx Ex0 C xy Ey0 C xz Ez0
py0 D yx Ex0 C yy Ey0 C yz Ez0
5.2.5
pz0 D zx Ex0 C zy Ey0 C zz Ez0

We have simplified the notation in the above equations and used (p0 ) and ( ) to
represent p0 vf vi and vf vi , respectively. The retention of the curved brackets serves
as a reminder that transition electric dipole components and transition polarizability
tensor components are involved. The omission of the transition-defining subscripts vi
and vf should not lead to confusion in the context of this section. The corresponding
simplification for the transition polarizability tensor would be to use c (a) or (a) if explicit
specification of the coordinate basis is not needed.
This is an appropriate point at which to set out our future policy in regard to nomenclature for transition polarizability tensors, their components and their invariants. In general
we shall use the simplest nomenclature possible consistent with clarity and unambiguity.
For mathematical developments a fuller labelling is usually necessary, whereas for tables
and many formulae a simpler notation is convenient, making for compactness. In particular when dealing with tensor invariants for Raman scattering we shall usually write a
for avf vi , for vf vi and for vf vi ; and also G 0 for G 0 vf vi , G 1 for G 1 vf vi
and G 2 for G 2 vf vi . We shall normally only add transition labels or coordinate basis
labels when specific transitions or coordinate bases require emphasis. This policy will
also be applied to related quantities such as transition dipoles. For Rayleigh scattering we
shall write a0 , 0 and so on for the tensor invariants.
As we are taking the wave functions to be real at this stage, the Placzek vibrational
transition polarizability tensor is symmetric with respect to exchange of the cartesian
component indices and and thus D . Furthermore, in the decomposition of
88
The Raman Effect
the tensor (c a, (c a1 ) is necessarily zero; and, amongst the tensor invariants, c G
c are necessarily zero.
1
and
5.2.2 The spherical basis

In the spherical basis the analogue of eq. (5.2.1) is
s p0 vf vi D s avf vi s E0
5.2.6
and the corresponding component form is

p0 vf vi D E 0
5.2.7
s avf vi is the Placzek vibrational transition polarizability tensor in the spherical basis
and vf vi its component, which by analogy with eqs. (5.2.3) and (5.2.4) is
defined as follows:
with
5.2.8

1 heg jpO jer iher jpO jeg i heg jpO jer iher jpO jeg i
O Q D
C
5.2.9
r6Dg
er eg 1
er eg C 1
In eqs. (5.2.6) to (5.2.9) we have reverted to the full notation for a transition polarizability
tensor and its components in order to facilitate comparisons with eqs. (5.2.1) to (5.2.4).
Again the definitions given in eqs. (5.2.8) and (5.2.9) are only valid if Q is a
function of the normal coordinate Q only and also the conditions 1, 2 and 4 given on the
previous page are fulfilled.
The intensity formulae for a system of freely rotating molecules will now involve
isotropic averages of the squares of transition polarizability components in the spherical
basis, typical examples of which are hj11 vf vi j2 i and hj11 vf vi j2 i.
The set of independent linear equations implicit in eqs. (5.2.6) and (5.2.7) is as follows:
p10 D 11 E10 C 10 E00 C 11 E10
p00 D 01 E10 C 00 E00 C 01 E10
5.2.10
p10 D 11 E10 C 10 E00 C 11 E10

where we have again introduced a simplified notation and written p0 for p0 and
for vf vi . The are in general complex and the general relationships
between tensor components in the spherical basis, which are dealt with in Chapter A14,
Section 14.3, are somewhat more complicated than in the cartesian basis. However, for
the special case of the Placzek vibrational transition polarizability tensor in the spherical
basis s a1 , s G 1 and s are necessarily zero; and as a result, 11 D 11 , 10 D 01
and 10 D 01 .
Since the vf vi are in general complex we write hj vf vi j2 i
89
5.3 DEFINITION OF ILLUMINATIONOBSERVATION

GEOMETRY
In principle, the illuminationobservation geometry is completely defined if we know the
orientation of the unit vector n0s defining the direction of propagation of the scattered
radiation relative to the unit vector n0i defining the direction of propagation of the incident
radiation. Referring to Fig. 5.1 this relative orientation may be defined by the angle
or
by the complementary angle , where D 180
D cos1 n0i n0s . We may regard n0s
and n0i as defining a plane which we shall call a scatter plane . For plane polarized
radiation we may define two polarization vectors, one for the incident and one for the
scattered radiation, which lie in the scatter plane (i.e. are parallel to the scatter plane) and
are at an angle D 180
to each other; and in addition two orthogonal polarization
vectors, one for the incident and one for the scattered radiation, which are perpendicular
to the scatter plane and mutually parallel. Thus the geometrical relationships of both the
polarization vectors and propagation vectors of the incident and scattered radiation are all
established when we define n0s and n0i .
In practice, however, we need to relate n0s and n0i to a space-fixed or laboratory axis
system. The use of such an axis system is valuable conceptually in obtaining intensity
formulae and also in the choice of illuminationobservation geometry in experimental
| |s
s
ns0
q
ni0
||i
Figure 5.1 Definition of scatter plane in terms of n0i and n0s and
.
Chapter A17 discusses propagation vectors. Here, the propagation vectors are unit vectors because we are
considering scattering from an ideal gas.
But see page 91 for a more general definition.
90
The Raman Effect

z
ez
ex
ey
n i = ez
Figure 5.2 Space-fixed right-handed cartesian axis system x, y, z associated with unit
vectors ex , ey and ez .
situations. However, ultimately, as we are considering isotropically averaged systems, the

intensity formulae will be expressed in terms of tensor invariants which are independent
of the choice of axes.
Without loss of generality we can choose the relationship of the scatter plane to the
space fixed axes so that the geometrical factors have the simplest possible form. We
consider the scattering molecule to be at the origin O of a space-fixed right-handed
cartesian axis system x, y, z associated with unit vectors ex , ey and ez (Fig. 5.2). We
choose the direction of propagation of the incident radiation, which is taken to be a
plane harmonic monochromatic light wave, so that n0i D ez . Initially we shall consider
the incident radiation to be propagating as a non-divergent parallel beam of small crosssection. This can be achieved readily with laser excitation. Although the scattered radiation
is generally propagating in all directions, the experimental conditions can be arranged so
that the scattering which is observed along a particular direction also closely approximates
to a non-divergent beam. Thus, initially, it is realistic to base our treatment of scattering
formulae on narrow non-divergent incident and scattered beams. When subsequently we
consider small departures from these idealized conditions it will transpire that corrections
to the intensity formulae are either zero or small.
With these idealized conditions in mind we consider the case of observation of the
intensity and polarization properties of scattered radiation along a general direction Ov
associated with n0s . Two angles and , defined as in Fig. 5.3, are needed to specify
this direction in relation to the space-fixed coordinate system x, y, z with associated unit
vectors ex , ey and ez , respectively. An observation direction defined in this way leads to
rather complicated geometrical factors involving and .
However, for the idealized case of non-divergent incident and scattered beams, no loss
of generality is involved if the unit vectors ex , ey and ez are chosen so that the observation
direction, which we now label Ow makes an arbitrary angle with the positive z direction
91
z
u
n0s
q
ex
ez
O
ey
f
y
x
ni = ez
0
Figure 5.3 n0i D ez and n0s in a general orientation Ov defined by and .

z
ns0 = ea
a
ea
c
ec
b
eb
q
ez
ex
ey
ni0 = ez
Figure 5.4 n0i D ez and n0s D ea lying in the scatter plane (defined by ex and ez ) with
orientation Ow defined by . The hatched area denotes the scatter plane.
and lies in the plane defined by ex and ez . A considerable simplification in the geometrical
factors then results. This situation is illustrated in Fig. 5.4 and corresponds to a special
case of Fig. 5.3 with D 0. The plane defined by ex and ez is now the scatter plane
and forms a reference plane for the definition of linear polarization states which will be
adopted henceforth.
The definition of the scatter plane by ex and ez overcomes the difficulties that arise for D 0 or if n0i
and n0s are used to define the scatter plane.
92
The Raman Effect
In practice, observations are almost invariably made at particular values of : D /2,

that is n0s D ex (so-called 90 scattering); D 0, that is n0s D ez (so-called forward scattering); or D , that is n0s D ez (so-called back- or retro-scattering). For these cases
still further simplifications in the geometrical factors result.
For the general observation direction Ow it is convenient to define a secondary righthanded cartesian system a, b, c associated with the unit vectors ea , eb and ec . This is
oriented so that n0s D ea , ea and ec lie in the scatter plane defined by ex and ez , and eb is
perpendicular to it (Fig. 5.4). The two sets of unit vectors are then related as follows:
n0s D ea D ex sin C ez cos
eb D ey
5.3.1
ec D ex cos C ez sin

We see that when D /2, ea D ex , eb D ey and ec D ez . However, when D 0, ea D ez ,
eb D ey and ec D ex ; and when D , ea D ez , eb D ey and ec D ex . As we can always
arrange matters so that D 0, then for idealized non-divergent incident and scattered
beams the relationships given in eqs. (5.3.1) are all that are normally required.
However, in an experimental situation the scattered radiation may be observed over a
small solid angle around a chosen observation direction. To make quantitative estimates
of how this affects the observed intensities it will be necessary in due course to consider
the case of a general illuminationobservation geometry for which both and are
non-zero. We therefore give below the relationships between ex , ey and ez and another set
of unit vectors ea , eb and ec , associated with a secondary right-handed cartesian system
oriented so that n0s D ea , ea and ec lie in the scatter plane zOv and eb is perpendicular to
it and directed so that eb D eb D ey when D 0. (Fig. 5.5)
n0s D ea D ex cos sin C ey sin cos C ez cos
eb D ex sin C ey cos
5.3.2
ec D ex cos cos ey sin cos C ez sin

Clearly, when D 0, the set of unit vectors ea , eb and ec becomes the set ea , eb and ec
and eqs. (5.3.2) reduce to eqs. (5.3.1).
When dealing with circular polarized radiation we use spherical coordinates instead
of cartesian coordinates. We therefore need the corresponding relations between the unit
vectors in the spherical basis defined in terms of the unit vectors ex , ey and ez of the
cartesian axis system x, y, z and those defined in terms of a secondary cartesian axis
system appropriate to the chosen observation direction, that is a, b, c or a, b, c.
In the principal cartesian axis system we have

1
e1i D p ex C iey
2
i
e0 D ez

1
i
D p ex iey
e1
2
5.3.3
93
z
u
n0s=ea
eb
ec
ea
f
y
x
n i=ez
0
Figure 5.5 n0i D ez and n0s D ea in a general orientation Ov defined by and . The hatched
area denotes the scatter plane.
where the superscript i on the spherical coordinates identifies these unit vectors as
being associated with incident radiation. We note that when n0i D ez in our convention e1i
i to right circular polarization.
corresponds to left circular and e1
In the secondary cartesian axis system a, b, c we have
1
e1s D p ec C i eb
2
s
e0 D ea
1
s
e1
D p ec i eb
2
5.3.4
where the superscript s on the spherical coordinates identifies these unit vectors as being
associated with scattered radiation. It should be noted that for our chosen orientation of
the secondary axis system a, b, c, when n0s D ea then e1s corresponds to left circular
s
to right circular polarization. Using eqs. (5.3.1) we find, after some algebraic
and e1
See Chapter A21, Section 21.2.
94
The Raman Effect
manipulation, that

1
e1s D p ex cos C iey ez sin
2
e0s D ex sin C ez cos

1
s
e1
D p ex cos iey ez sin
2
5.3.5
For the special case of D 0, eqs. (5.3.5) have the same form as eqs. (5.3.3) with the
s
superscript i replaced by s. For n0i D ez and D 0, e1s corresponds to left circular and e1
to right circular polarization. For D , however, eqs. (5.3.5) reduce to

1
e1s D p ex iey
2
e0s D ez

1
s
e1
D p ex C iey
2
5.3.6
s
For n0i D ez , e1s now corresponds to right circular and e1
to left circular polarization.
We now consider the secondary cartesian axis system a, b, c involving both and . We
first introduce a new set of unit vectors ex , ey and ez associated with the cartesian axis
system, obtained by anti-clockwise rotation of the principal axis system x, y, z through
an angle about the z axis (Fig. 5.5). It follows that
ex D ex cos C ey sin
ey D ex sin C ey cos
5.3.7
ez D ez
s
If we then relate the e1s , e0s and e1
to the ex , ey and ez , the resulting expressions have
the same form as eqs. (5.3.5) with ex , ey and ez replaced by ex , ey and ez . Thus for
circularly polarized light there is no dependence on .
5.4 INTENSITY OF SCATTERED RADIATION: SOME GENERAL

CONSIDERATIONS
5.4.1 Development of a symbol for scattered intensity
We now consider how the states of polarization, propagation vectors and angular dependence of the incident and scattered radiation might be brought together in one intensity
symbol. As the scatter plane is defined by ex and ez it is only necessary, as already
explained, to specify one scattering angle to define the illumination-observation geometry. Thus if we represent the states of polarization of the incident (i) and scattered (s)
radiation by the symbols pi and ps , respectively, we can define a general intensity symbol
95
as follows:
I; ps , pi
5.4.1
The symbols used to designate the various states of polarization pi and ps are summarized
in Reference Table 5.1.
As an example we consider the case of incident radiation propagating along the z
axis (n0i D ez ) and linearly polarized with the electric vector along the y axis (Ey0 6D 0)
and observation of the component of the scattered radiation propagating along the x axis
(n0s D ex ) and polarized with the electric vector along the z axis (pz0 6D 0). The intensity
symbol for this case is then
I /2; jjs , ?i
5.4.2
As this symbol is defined by reference to the scatter plane there is no need to specify
the cartesian axis system. Thus this symbol is well suited for use with scattering systems
consisting of freely orienting molecules because, as we shall see, the process of isotropic
averaging leads to intensities being expressed in terms of tensor invariants which do not
require cartesian axis labels. This is in contrast to what is known as the Porto symbol,
introduced in a paper by Damen, Porto and Tell (1966). This symbol involves cartesian
axis labels and is designed for use with oriented scattering systems like crystals. As this
book is concerned only with freely rotating molecules we use the scatter plane symbol in
the form given above throughout this book.
The scatter plane concept can also be applied to scattered power for which a fully
characterized general symbol would be
; ps , pi
5.4.3
5.4.2 Scattering cross-section

The relationship between the intensity I of Rayleigh and Raman scattered radiation from
a single molecule and the irradiance I of the incident radiation has the general form
I D 0I
5.4.4
This is evident from the general considerations already presented in Chapter 2, Section 2.3
and a perusal of the various specific formulae to be developed in the following sections
of this chapter.
For some purposes, particularly the comparison of scattering efficiencies of different
molecules, it is useful to introduce scattering cross-sections because they are essentially
In an earlier book (Long, 1977), the intensity symbol was defined as P Ips and for this particular case
it would have been ? Ijj /2. The new system avoids superscripts and subscripts on I. Although the order
of writing the polarization states is reversed in the new system, as the labels i and s are now always used
on the polarization symbols, no confusion is likely.
Chapter A20, Section A20.2 deals with intensity.
Chapter A17, Section A17.2.3 deals with irradiance.
96
The Raman Effect
molecular properties that are independent of the irradiance of the incident radiation but
not necessarily of other experimental factors.
We proceed as follows. We first rearrange eq. (5.4.4) and obtain
I
5.4.5
0 D
I
where 0 has the unit m2 sr1 molecule1 . We then introduce the quantity which is
defined by
4
D
0 d
5.4.6
0
where d is an element of solid angle (unit: steradian, which is dimensionless). Note that
has the unit m2 molecule1 . The quantity is termed the total scattering cross-section
per molecule and 0 the first differential scattering cross-section per molecule. If the
scattered intensity varies across a range of wavenumbers around a central wavenumber
we can take this into account by introducing 00 , the second differential cross-section per
molecule, per unit wavenumber interval, defined by
0
2
D
5.4.7
Q
Q d
where 00 has the unit m sr1 molecule1 when Q has the unit m1 .
It can be seen from eqs. (5.4.5) and (5.4.6) that measures the rate at which energy
is removed from the incident beam by scattering over 4 steradians, relative to the rate
at which energy crosses a unit area perpendicular to the direction of propagation of the
incident beam. As has dimensions of m2 molecule1 we may say that the radiation
scattered over 4 steradians by a molecule with a scattering cross-section corresponds
to the radiation incident on the area . We shall see subsequently how small that area is
for normal Raman scattering.
Unlike the intensity I, the several cross-sections , 0 and 00 are independent of the
irradiance I of the exciting radiation. However it must be emphasized that the crosssections remain dependent upon other experimental parameters. We now examine this
dependence using as an example 0 , the most commonly used cross-section.
Comparison of eq. (5.4.5) with typical formulae for the intensity of Rayleigh and
Raman scattering, eq. (5.5.17) for example, shows that for a single scattering molecule
we can write

some function of squares of
0
4
D kQ Q s transition tensor invariants and the
5.4.8
illuminationobservation geometry
Far from resonance, the function in the braces is independent of the wavenumber of
the incident radiation and in this case it is useful to introduce a wavenumber-normalized,
first-differential cross-section of a molecule defined by 0 Q s4 . Thus we have

some function of squares of
0 4
Q s D kQ transition tensor invariants and the
5.4.9
illuminationobservation geometry
00 D
97
where 0 Q s4 has the unit m6 sr1 molecule1 when Q s has the unit m1 . However, in
general the function in the braces does depend on the excitation wavenumber. Thus, when
quoting a wavenumber-normalized cross-section, it is desirable to indicate the excitation
wavenumber used to obtain it. Furthermore, the polarization of the incident and scattered
radiation and the illuminationobservation geometry have to be taken into consideration.
It follows that, strictly, 0 for example should be labelled in the same way as I and
be written as 0 ; ps , pi . Also it may be necessary to indicate the molecular transition
involved.
As it is straightforward using eqs. (5.4.8) and (5.4.9), to obtain expressions for scattering cross-sections from the intensity formulae we are about to develop, we shall not
give explicit formulae for scattering cross-sections in either the ensuing developments or
Tables.
5.5 INTENSITY FORMULAE AND POLARIZATION

CHARACTERISTICS FOR A GENERAL VIBRATIONAL
TRANSITION IN VARIOUS ILLUMINATIONOBSERVATION
GEOMETRIES
We consider the scattered radiation to originate from a small volume element located at
the origin of the axis system and to contain N freely rotating identical molecules of an
ideal gas.
Initially we shall treat the scattering associated with a general vibrational transition
from vi to vf . This may involve changes in the vibrational quantum numbers of one or
more vibrational modes. For such a transition the number of molecules contributing to
the scattering is the number in the initial vibrational state vi which is given by
Nvi D Ngvi
expfEvi /kTg
Qv
5.5.1
where Evi is the energy of a molecule in the initial state vi with vibrational degeneracy
gvi and Qv is the sum over vibrational states or vibrational partition function defined by

gvj expfEvj /kTg
5.5.2
Qv D
j
with the summation over all vibrational states j. We do not specify the relation of Evi
to the vibrational frequencies of the molecule at this stage. Specific types of vibrational
transitions will be considered subsequently.
We shall assume throughout that the scattering is incoherent so that the intensity of
the scattering from N molecules is simply N times that scattered by a single molecule.
We recall that Raman scattering is always incoherent whatever the relationship of the
98
The Raman Effect
direction of observation to the direction of the incident radiation; and that non-forward
( 6D 0) Rayleigh scattering from an ideal gas is also incoherent.
5.5.2 Linearly polarized incident radiation

We begin by noting certain generalizations which emerge when eqs. (5.2.5) are considered
in relation to the illuminationobservation geometry and the various possible combinations of linear polarization states of the incident and scattered radiation.
For incident radiation with n0i D ez , E0 is restricted to Ex0 jji or Ey0 ?i , and for
scattered radiation with n0s D ex (i.e. D /2, p0 is restricted to py0 ?s or pz0 jjs ,
as Fig. 5.6 illustrates. Thus the only diagonal component of (a) that can occur with
this illumination-observation geometry is (yy ), and its isotropic average hyy 2 i will be
involved in I /2; ?s , ?i . The three other possible combinations of incident linear and
scattered linear polarizations, will involve off-diagonal components of (a). Thus (pz0 ) and
Ey0 involve (zy ), (pz0 ) and Ex0 involve (zx ), and (py0 ) and Ex0 involve (yx ). It follows
that I /2; jjs , ?i involves hzy 2 i, I /2; jjs , jji involves hzx 2 i, and I /2; ?s , jji
involves hyx 2 i. Now all isotropic averages of the type hxy 2 i are equal and distinct
from those of the type hxx 2 i and thus I /2; jjs , ?i D I /2; jjs , jji D I /2; ?s , jji 6D
I /2; ?s , ?i . It can be seen from Fig. 5.6 that for I /2; ?s , ?i the electric vectors of
the incident and scattered radiation are parallel to each other, whereas for the other three
intensities they are perpendicular to each other. This distinction does not involve any axis
system, as is to be expected for scattering systems that are isotropically averaged.
z
O
pz (| |s)
0
Ex (||i)
0
py (s)
Ey (i)
0
ns0 = ex
ni0 = ez
Figure 5.6 Components of E0 when n0i D ez , and p0 when n0s D ex .
See discussion in Chapter 3, Section 3.7.

Some authors have used the relative orientations of pi and ps , namely parallel and perpendicular, to
describe these two situations involving hxx 2 i and hxy 2 i respectively, for example. This system must not
be confused with our usage, in which the orientations of pi and ps are related to a reference or scatter plane.
99
It follows that only two appropriately chosen intensity measurements involving linear
polarized radiation are needed to characterize a symmetric transition polarizability tensor
in the cartesian basis. One must be I /2; ?s , ?i because it alone involves isotropic
averages of the type hyy 2 i, the other can be any of the three involving isotropic averages
of the type hzy 2 i, but usually I /2; jjs , ?i is chosen. Other methods of characterizing
the symmetric transition polarizability tensor, which involve different geometries and also
degrees of depolarization will emerge in the more detailed discussions that follow, but in
all cases two appropriate measurements suffice.
We now treat in more detail each of the possible combinations of incident and scattered
linear polarizations for various values of . We consider first the case of linearly polarized radiation incident radiation with n0i D ez and Ey0 6D 0 (that is ?i ) and observation
of scattered radiation with n0s D ex D /2. To calculate the intensity of the radiation
scattered with n0s D ex and polarized either perpendicular or parallel to the scatter plane
we first need the components (py0 ) or (pz0 ), respectively, of the amplitude of the transition electric dipole moment (p0 ). For a single molecule situated at the origin of the
axis system and having a fixed orientation in space, it follows straightforwardly from
eqs. (5.2.5) that
py0 D yy Ey0
5.5.3
pz0 D zy Ey0
5.5.4
and
Then for N freely rotating molecules of an ideal gas, with Nvi molecules in the initial vibrational state vi we can calculate the intensity of the scattered radiation, linearly polarized
with its electric vector in the y direction by forming the square of eq. (5.5.3), averaging
over all orientations of the molecule by replacing yy 2 by its isotropic average hyy 2 i
inserting this into eq. (2.2.1) or eq. (2.2.3) with sin2 D 1 and multiplying by Nvi . In this
way we obtain from eq. (5.5.3)
I /2; ?s , ?i D k0 Nvi s4 hyy 2 iEy20 D k0Q Nvi Q s4 hyy 2 iEy20
5.5.5
where k0 is given by eq. (2.2.2) and k0Q by eq. (2.2.4).

Proceeding similarly we obtain from eq. (5.5.4)
I /2; jjs , ?i D k0 Nvi s4 hzy 2 iEy20 D k0Q Nvi Q s4 hzy 2 iEy20
5.5.6
We now introduce into eqs. (5.5.5) and (5.5.6) the irradiance I of the incident radiation,
which is given by
I D 12 c0 0 Ey20
5.5.7
and express the isotropic averages in terms of the transition polarizability invariants.
Throughout this chapter and in the associated tables we use the traditional invariants a2
Chapter A14 discusses these invariants in detail, and Table A14.9 gives the relations between isotropic
averages such as hyy 2 i and hzy 2 i and a2 and 2 .
100
The Raman Effect
and 2 rather than the alternative Placzek invariants G 0 and G 2 :

2 C 4 2
2 C 4 2
45a
45a
I D kQ Nvi Q s4
I
I /2; ?s , ?i D k Nvi s4
45
45
and
I /2; jjs , ?i D k Nvi s4
2
15

I D kQ Nvi Q s4
2
15
5.5.8

I
5.5.9.
where
k D
2 0
1
k D
2
0 c0
160 c04 2
5.5.10
2 0
2
kQ D 2
0 c0
0
5.5.11
and
kQ D
The total intensity of the radiation scattered with n0i D ez , that is the intensity observed
when no analyser is in the scattered beam, is given by the sum of eqs. (5.5.8) and (5.5.9):

2 C 7 2
45a
I /2; ?s C jjs , ?i D k Nvi s4
I
45

45a2 C 7 2
4
D kQ Nvi Q s
5.5.12
I
45
For the depolarization ratio of the scattered radiation /2; ?i we have

I /2; jjs , ?i
i

/2; ? D
I /2; ?s , ?i
D
3 2
45a2 C 4 2
5.5.13
We may proceed likewise for other states of polarization using incident radiation with
n0i D ez . For the case of linearly polarized radiation with Ex0 6D 0, that is polarized parallel
to the scattering plane and scattered radiation with n0s D ex , so that D /2, it follows
straightforwardly using eqs. (5.2.5) that

2
5.5.14
I /2; ?s , jji D I
; jjs , jji D kQ Nvi Q s4
I
2
15
Sometimes called degree of depolarization. Chapter A21 discusses the polarization of electromagnetic
radiation.
101
and

s
I /2; jj C ? , jj D
kQ Nvi Q s4
2 2
15
5.5.15
Here and subsequently we give formulae for I; ps , pi in terms of Q only. When

needed, the change to formulae in terms of is readily made. For the depolarization
ratio /2; jji we have

I /2; ?s , jji
i
D1
/2; jj D
5.5.16
I /2; jjs , jji
provided 6D 0.
We see that a symmetric transition polarizability tensor can be completely characterized, not only by two intensity measurements, namely I /2; ?s , ?i and any one of
I /2; jjs , ?i , I /2; ?s , jji or I /2; jjs , jji , but also by measuring one intensity namely
I /2; ?s , ?i and the depolarization ratio, /2; ?i .
The corresponding formulae for n s D ez D 0 and n s D e z D can be obtained
by following the same procedures. For n0s D ez or ez the components of (p0 ) required
are (py0 ) for perpendicular polarization ?s and (px0 ) for parallel polarization jjs . The
following intensity formulae are readily obtained for incident radiation with polarization
state ?i :

2 C 4 2
45a
5.5.17
I0 or ; ?s , ?i D kQ Nvi Q s4
I
45

2
s
i
4
I0 or ; jj , ? D kQ Nvi Q s
5.5.18
I
15

2 C 7 2
45a
I0 or ; ?s C jjs , ?i D kQ Nvi Q s4
5.5.19
I
45
The depolarization ratios are
0 or ; ?i D
3 2
45a2 C 4 2
For the jji case the intensity formulae are

s
I0 or ; ? , jj D
kQ Nvi Q s4

I0 or ; jjs , jji D kQ Nvi Q s4

I0 or ; ?s C jjs , jji D kQ Nvi Q s4
2
15
5.5.20
5.5.21
45a2 C 4 2
45
45a2 C 7 2
45

I
5.5.22
5.5.23
102
The Raman Effect
We note that the intensity formulae given by eqs. (5.5.17), (5.5.18), (5.5.21) and (5.5.22)
again fall into two categories: those with the electric vector of the incident and scattered
radiation parallel to each other (and so involving a and ) and those with these vectors
perpendicular to each other (and so involving only ).
The depolarization ratios for the jji case are
0 or ; jji D
3 2
45a2 C 4 2
5.5.24
5.5.3 Natural incident radiation

In light-scattering experiments illumination with natural radiation has now been
completely replaced by radiation from a laser source which has a defined polarization.
However for completeness, and also to provide the formulae used in the earlier literature,
we include the case of incident natural radiation. Natural radiation of irradiance I is
equivalent to the superposition of two independent, linearly polarized waves each of
irradiance I /2 with their electric vectors perpendicular to each other and to the direction
of propagation. Thus we can obtain the intensity of the scattered radiation produced
by natural incident radiation with n0i D ez by adding the intensity produced by linearly
polarized incident radiation with Ex0 6D 0, (that is jji ) and irradiance I /2, and the intensity
produced by linearly polarized incident radiation with Ey0 6D 0 (that is ?i ) and irradiance
I /2. For scattered radiation with n0s D ex , that is D /2,

45a2 C 7 2 I
s
i
4
I /2; ? , n D kQ Nvi Q s
5.5.25
45
2

2
I
2
I /2; jjs , ni D kQ Nvi Q s4
5.5.26
15
2
and
I /2; ?s C jjs , ni D kQ Nvi Q s4
45a2 C 13 2
45

I
2
5.5.27
The depolarization ratio is

/2; ni D
6 2
45a2 C 7 2
5.5.28
An alternative procedure for obtaining /2; ni should be noted. It is also given

by the ratio fI /2; ?s , jji C I /2; jjs , jji g/fI /2; ?s , ?i C I /2; jjs , ?i g as is easily
verified from eqs. (5.5.14), (5.5.8) and (5.5.9). The numerator of this ratio is the total scattered intensity with n0s D ex when the incident radiation with n0i D ez is linearly polarized
See footnote on page 98.
103
with Ex 6D 0jji ; the denominator is the total scattered intensity with n0s D ex when the
incident radiation with n0i D ez is linearly polarized with Ey 6D 0?i . Experimentally this
avoids the use of an analyser in the scattered beam.
Similarly, for scattered radiation with n0s D ez D 0 or n0s D ez D ,

2 C 7 2
45a
I
5.5.29
I0 or ; ?s , ni D kQ Nvi Q s4
45
2

2 C 7 2
45a
I
I0 or ; jjs , ni D kQ Nvi Q s4
5.5.30
45
2
and
I0 or ; ?s C jjs , ni D kQ Nvi Q s4
45a2 C 7 2
45

I
5.5.31
The depolarization ratios are

0 or ; ni D 1
5.5.32
The formulae for intensities and depolarization ratios for incident radiation which
is linearly polarized or natural are given in Reference Tables 5.2(a)(c). The detailed
preamble to these tables includes the relationships between the intensity formulae as
presented there and both scattering cross-sections and Placzek invariants. The method
of presentation used in Reference Tables 5.2(a)(c) enables the scalar part involving a2 ,
the anisotropic part involving 2 and the scalar plus anisotropic parts involving a2 and
2 to be extracted if required. These tables also include the antisymmetric contributions
which involve 2 , but these apply only when the transition tensor is not symmetric as in
resonance Raman scattering.
It should be noted that for normal Raman scattering the isotropic and anisotropic contributions to the total intensity can be determined experimentally from appropriate intensity
measurements. Thus for example Iiso is given by I /2; ?s , ?i 43 I /2; jjs , ?i and
Ianiso directly by I /2; jjs , ?i .
5.5.4 Angular dependence of scattered intensity

We now consider the general dependence of the intensity of the scattered radiation on
the angle and . We treat first the case when D 0 so that the scattered radiation is
confined to the xz plane and has n0s D ea The incident radiation has n0i D ez (Fig. 5.4).
The general procedure for obtaining expressions for the scattered intensities as a function of scattering angle is as follows. We first obtain expressions for transition dipole
moment amplitudes referred to the appropriate subsidiary cartesian axis system, in this
case a, b, c (see Fig. 5.4) and then make the transformation to the axis system x, y, z.
Chapter 7 considers vibrational resonance Raman scattering.
104
The Raman Effect
We will illustrate these procedures by considering I; jjs , ?i and I; jjs , jji . For
I; jjs , ?i we have
pjj0 D pc0
5.5.33
and using the set of eqs. (5.3.1), we obtain

pc0 D px0 cos C pz0 sin
5.5.34
Then, noting that E?0 D Ey0 , using eq. (5.2.5) for the relationship between (px0 ) or (pz0 )
and Ey0 , squaring and introducing isotropic averages we find that
pjj0 2 D pc0 2 D fhxy 2 i cos2 C hzy 2 i sin2 g Ey20
Now hxy 2 i D hzy 2 i D 2 /15 and hence

pjj0 2 D
2
15
5.5.35
Ey20
5.5.36
Thus pjj0 2 is independent of . Using eqs. (5.5.36) and (2.2.1) and introducing the
irradiance of the incident radiation, I , we obtain

2
s
i
4
5.5.37
I; jj , ? D kO Nvi Q s
I
15
For I; jjs , jji eqs. (5.5.33) and (5.5.34) also apply but, as Ejj0 D Ex0 , instead of
eq. (5.5.35) we now have

pjj0 2 D hxx 2 i cos2 C hzx 2 i sin2 Ex20
5.5.38
However, as hxx 2 i 6D hzx 2 i, pjj0 2 is now dependent on . Using eqs. (5.5.38) and
(2.2.1) and introducing the irradiance of the incident radiation, I , we obtain

2 C 4 2 cos2 C 3 2 sin2
45a
I; jjs , jji D kQ Nvi Q s4
5.5.39
I
45
The formulae for I; ps , pi for all combinations of ps and pi with pi D jji , ?i or ni
and ps D jjs , ?s or jjs C ?s are given in Reference Tables 5.2(a)(c). It can be seen that
the following cases are functions of : I; jjs jji , I; jjs C ?s , jji , I; jjs , ni and I; jjs C
?s , ni . Reference Tables 5.2(a)(c) also include the formulae for ; jji , , ?i and
, ni . It can be seen that ; jji and ; ni are functions of , but ; ?i is not.
These results enable us to estimate the so-called convergence errors in intensity
measurements that arise when there is a small but finite angle of collection around the
direction of observation. Consider for example observation of scattering for which ideally
n0s D ex , that is D 90 . Now for D 80 , I80 ; jjs , jji involves 0.03f45a2 C 4 2 g C
0.97f3 2 g. This is very close to I90 ; jjs , jji unless f45a2 C 4 2 g 3 2 .
105
We now consider the dependence of the intensity of the scattered radiation on both
and (see Fig. 5.5). The scattered radiation under consideration is along Ov with n0s D ea
and the incident radiation has n0i D ez . The first step is to obtain expressions for transition
dipole moment amplitudes referred to the subsidiary cartesian axis system, a, b, c and
then to make the transformation to the axis system x, y, z. By way of illustration we shall
consider three cases: I, ; jjs , ?i , I, ; ?s , ?i and I, ; jjs , jji .
For I, ; jjs , ?i we have
pjj0 D pc0
5.5.40
pjj0 D px0 cos cos py0 sin cos C pz0 sin
5.5.41
and using eqs. (5.3.2) we obtain
Then noting that E?0 D Ey0 , using eqs. (5.2.5) for the relations between px0 (or py0 , pz0 )
and Ey0 , squaring and introducing isotropic averages, we find that
pjj0 2 D fh2xy i cos2 cos2 C h2yy i sin2 cos2 C h2zy i sin2 gEy20
5.5.42
This may be rewritten in the form

pjj0 2 D fh2zy i sin2 C hxy 2 i cos2 C [hyy 2 i hxy 2 i] cos2 sin2 gEy20
5.5.43
Comparison with eq. (5.5.35) shows that the third term in eq. (5.5.43) gives the additional
contribution to pjj0 2 when 6D 0. We see that there is an azimuthal contribution to pjj0 2
for all values of except D /2. This contribution will be small for small values of .
For I, ; ?s , ?i we have
p?0 D pb0
5.5.44
p?0 D px0 sin C py0 cos
5.5.45
and using eqs. (5.3.2) we obtain
Proceeding as previously with E?0 D Ey0 we find that

p?0 2 D fh2xy i sin2 C h2yy i cos2 gEy20
5.5.46
which we may rewrite as

p?0 2 D fh2yy i [h2yy i h2xy i] sin2 gEy20
5.5.47
Thus the first term corresponds to (p?0 ) when D 0 and the second term gives the
additional contribution when 6D 0. We see that there is always an azimuthal contribution
to (p?0 ) but this is small for small values of .
In the case of I, ; jjs , jji eqs. (5.5.40) and (5.5.41) also apply, but we now have
Ejj0 D Ex0 . Thus
pjj0 2 D fh2xx i cos2 cos2 C h2yx i sin2 cos2 C h2zx i sin2 gEx20
5.5.48
106
The Raman Effect
which we may rewrite as

pjj0 2 D fhzx 2 i sin2 C hyx 2 i cos2 C [hxx 2 i hxy 2 i] cos2 cos2 gEx20
5.5.49
This reduces to eq. (5.5.38) when D 0. For 6D 0 there is an azimuthal contribution
[hxx 2 i hxy 2 i] cos2 cos2 to pjj0 2 for all values of cos2 , except D /2.
For actual experimental conditions the situation can be more complicated. If a relatively
wide angle lens is used to collect the scattered radiation it is necessary to integrate over the
solid angle of collection, and if a short focal length lens is used to focus the exciting laser
radiation into the sample this can change the polarization characteristics of the exciting
radiation. We shall not pursue such practical considerations here.
5.5.5 Circularly polarized incident radiation

We begin with the case of incident radiation with n0i D ez which is right circularly polarized, so that E10 6D 0 and E10 D 0. To calculate the intensity of the scattered radiation
with n0s D ez , that is for D 0, with right or left circular polarization we first need the
components p10 and p10 , respectively, of the amplitude of the transition electric
dipole. For a single molecule situated at the origin of the axis system and having a fixed
orientation in space, we have from the set of eqs. (5.2.10),
p
L 0 D p10 D 11 E10 D 11 E
R 0
5.5.50
p
R 0 D p10 D 11 E10 D 11 E
R 0
5.5.51
and
Then, for N freely rotating molecules of an ideal gas with Nvi in the initial vibrational
state vi , it follows that
2
R s,
R i D k0Q Nvi Q s4 hj11 j2 ijE
I0;
R 0j
5.5.52
2
L s;
R i D kQ Nvi Q s4 hj11 j2 ijE
I0;
R 0j
5.5.53
and
If we now introduce into eqs. (5.5.52) and (5.5.53) the irradiance I of the incident radiation using the relationship
2I
2
jE
5.5.54
R 0j D
c0 0
and express the isotropic averages in terms of the transition tensor invariants a2 and 2 ,
we obtain

2 C 2
45a
R s,
R i D kQ Nvi Q s4
5.5.55
I
I0;
45
The E 0 , p0 and are in general complex and so we write jE 0 j2 . . ., jp0 j2 . . . j j2 . . . .
See Chapter A14, Section A14.7.4 and Table A14.7.
107
and

L s,
R i D kQ Nvi Q s4
I0;
2 2
15

I
5.5.56
Proceeding similarly it is straightforward to show that

L s,
L i D I0;
R s,
R i
I0;
5.5.57
R s,
L i D I0;
L s,
R i
I0;
5.5.58
and
These relationships stem from the fact that only two distinct types of isotropic averages
are ever involved: those of the type hj11 j2 i which arise when there is no reversal of the
circular polarization and those of the type hj11 j2 i which arise when there is reversal
of the circular polarization.
For observation in the back-scattering geometry, that is with n0s D e z so that D ,
we have the general relationships
R s , i D I0;
L s , i
I ;
5.5.59
L s , i D I0;
R s , i
I ;
5.5.60
and
i
R i
L i
where
can be either
or
on both sides of these equations. These relations
are a consequence of the fact that for D 0 the circular polarizations of the incident and
scattered radiation are defined in relation to an observer looking at radiation with n0i D ezi ,
whereas for D the circular polarization of scattered radiation is defined in relation to
an observer looking at radiation with n0i D e iz0 .
We may use the above results to obtain expressions for the reversal coefficients P 0
and P and the degrees of circularity C , pi :
6 2
D P 1
45a2 C 2
45a2 5 2
R i D
C 0;
45a2 C 7 2
P 0 D
5.5.61
5.5.62
with
C ; pi D C C , pi
5.5.63
R i D C ;
L i
C ;
5.5.64
and
Further manipulation will show that

2 /2; ?i

P 0 D
1 /2; ?i
See Chapter A21, Section A21.4.
5.5.65
108
The Raman Effect
or
/2; ?i D
and that
P 0
2 C P 0

/2; ni

P 0 D
1 /2; ni
5.5.66
5.5.67
or
P 0
1 C P 0
5.5.68
1 3 /2; ?i
1 C /2; ?i
5.5.69
R i
1 C 0;
R i
3 C C 0;
5.5.70
/2; ni D
and that
R i D
C 0;
or
/2; ?i D
So far we have considered forward and backward scattering of circularly polarized

radiation produced by incident radiation which is itself circularly polarized. The formulae
we have derived have involved in effect just four of the possible isotropic averages, namely
hj11 j2 i D hj11 j2 i for non-reversal and hj11 j2 i D hj11 j2 i for reversal of the
circular polarization. We see that in fact, only two distinct isotropic averages are involved.
However there are other possibilities involving the other five space averages which we
now consider.
First, scattered radiation associated with p00 which implies linearly polarized radiation
with the electric vector along the z axis. It follows from eqs. (5.2.10) that this can arise
from incident radiation whose electric vector is E10 , E00 or E10 . When the incident
electric vector is E00 we have, recalling that p00 , 00 and E00 are real,
p00 2 D h00 2 iE020
5.5.71
with
45a2 C 4 2
5.5.72
45
This is most readily interpreted by making the trivial transformation to the cartesian basis
so that eq. (5.5.71) becomes
h00 2 i D
pz20 D hzz 2 iEz20
5.5.73
We see that the scattered radiation is linearly polarized with pz0 6D 0, and that like the
incident radiation it propagates in the xy plane.
When the electric vector of the incident irradiation is E10 or E10 we have
p00 2 D hj01 j2 ijE10 j2
5.5.74
109
and
p00 2 D hj01 j2 ijE10 j2
5.5.75
with
hj01 j2 i D hj01 j2 i D
2
15
5.5.76
In each of these cases circularly polarized incident radiation with n0i D ez gives rise to
linearly polarized radiation with Ez0 6D 0, which propagates in the xy plane. As hj01 j2 i D
hj01 j2 i, then if jE10 j2 D jE10 j2 the intensity of the linearly polarized scattered radiation calculated using eq. (5.5.74) is equal to that calculated using eq. (5.5.75).
Finally, we consider the case of scattered radiation associated with p10 and p10
produced by linearly polarized incident radiation with electric vector E00 and propagating
in the xy plane. We then have
jp10 j2 D hj10 j2 iE020
5.5.77
jp10 j2 D hj10 j2 iE020
5.5.78
and
with
2
.
5.5.79
15
Thus the scattered radiation consists of left handed and right handed circularly polarized
radiation of the same intensity with n0s D ez . This is equivalent to linearly polarized
radiation with n0s D ez with its electric vector along the x direction. The last five cases
all give rise to linearly polarized scattered radiation in contrast to the first four cases
which were associated with circularly polarized scattered radiation produced by circularly
polarized incident radiation.
The formulae for intensities, depolarization ratios, reversal coefficients and degrees of
circularity of scattered radiation produced by circularly polarized incident radiation are
given in Reference Table 5.2(d).
hj10 j2 i D hj10 j2 i D
5.5.6 Symmetry and depolarization ratios, reversal coefficients

and degrees of circularity
Comparison of the expressions for depolarization ratios, reversal coefficients and degrees
of circularity for Rayleigh and Raman scattering reveals some important differences. We
recall that a0 , the mean value of the equilibrium polarizability tensor a0 , can never be
zero, because all molecular systems are polarizable to some extent; the anisotropy 0 will,
however, be zero for isotropic systems. Thus, for Rayleigh scattering
0 /2; ?i < 3/4.
5.5.80
110
and
The Raman Effect

0 /2; ni < 6/7
5.5.81
0 P 0 < 6
5.5.82
R 1
5/7 < C 0;
5.5.83
By contrast for Raman scattering, either the mean value a or the anisotropy of the
transition polarizability tensor can be zero. Thus, for Raman scattering
0 /2; ?i 3/4
0 /2; ni 6/7
and
0 P 0 6
R i 1
5/7 C 0;
5.5.84
5.5.85
5.5.86
5.5.87
Conventionally, a Raman line is said to be depolarized (dp) when /2; ?i D 3/4 or

/2; ni D 6/7; polarized (p) when 0 < /2; ?i < 3/4 (or 0 < /2; ni < 6/7);
and completely polarized when /2; ?i D 0 (or /2; ni D 0). The circular polarR i D 1 or
ization is said to be completely unreversed when P 0 D 0 (or C 0;
i
L D 1) and partly reversed when 0 < P 0 6 (or 5/7 C 0;
R i < 1 or
C 0;
i
L 5/7)
1 < C 0;
R i
It is interesting to note that, in Raman scattering, if /2; ?i > 1/3, C 0;
s
i
s
i
R ,
R I0;
L
R will be negative. Thus if the
will be negative and hence I0;
R s,
R i I0;
L s,
R i an elegant
Raman spectrum is recorded in the form of I0;
i
distinction may be made between strongly polarized bands ( /2; ? < 1/3) for which
R s,
R i I0;
L s,
R i is positive and weakly polarized and depolarized bands
I0;
i
R s,
R i I0;
L s,
R i is negative. For D
/2, ? > 1/3 for which I0;
(back-scattering configuration) it must be remembered that C , pi D C 0, pi so
R s,
R i I0,
L s,
R i are reversed.
that the signs of all the statements given for I0,
i
The depolarization ratio /2; jj defined in eq. (5.5.16) has not been included in
the above discussion. It is always unity for Rayleigh and Raman scattering, provided
0 6D 0 or 6D 0, respectively, and yields no useful information. In the special case of
Rayleigh scattering from systems with 0 D 0 and Raman scattering from systems with
D 0 eq. (5.5.16) does not apply as eq. (5.5.14) shows there is no scattered radiation
along the x axis in the jji case.
We see that, as well as intensity measurements, depolarization ratios, reversal coefficients and degrees of circularity also provide information on the polarizability tensor
invariants a0 and 0 or a and , but of course these are not all independent quantities.
The choice between two appropriate intensity measurements and one appropriate intensity
measurement and one appropriate polarization measurement to determine the two tensor
invariants will usually depend on the experimental situation.
The values of depolarization ratios, reversal coefficients, and degrees of circularity
in Raman scattering are determined by symmetry properties. However, the symmetry
information that can be obtained from measurement of these quantities is necessarily
somewhat circumscribed because isotropic averaging has been invoked for freely rotating
molecules.

r p/2; i =
111
I(p/2; ||si)
r p/2; ni =
I(p/2; si )
CCl4
n1(a1)
459 cm1
I(p/2; | |s ni)
I(p/2; i ni )
r(p/2; i )
0.01
0.01
r(p/2; ni )
30
30
CCl4
n4( f 2 )
315 cm1
CH2Cl2
n4(a1)
283 cm1
0.75
r(p/2; i )
0.86
r(p/2; ni )
0.30
r(p/2; i )
0.46
r(p/2; ni )
z
Analyser
Analyser
Broken curves
Ex
y
I(p/2; ||s,i)
Ey
Ey
I(p/2; ||s,n i)
z
Analyser
Analyser
Full curves
Ex
y
Ey
I(p/2; s,i)
x
I(p/2; s,n i)
Ey
x
Figure 5.7 Measurement of /2; ?i and /2; ni for some bands of CCl4 and CH2 Cl2 .
112
The Raman Effect
We conclude this section by considering some examples of the measurement of polarization properties of Raman scattering. Figure 5.7 illustrates the measurement of /2, ?i
and /2; ni for selected fundamental vibrations of CCl4 and CH2 Cl2 . For a discussion of these results we use a0 k and 0 k , the invariants of the derived polarizability
tensor a0 k associated with the kth vibrational mode. We see that in the case of CCl4 for
the spherically symmetric Q 1 a1 vibration for which 0 1 D 0, the depolarization ratio
tends to zero, whereas for the degenerate vibration 4 f2 for which a0 4 D 0, /2; ?i
D 3/4 and /2; ni D 6/7. For the Q 4 a1 vibration of CH2 Cl2 for which a0 4 6D 0 and
0 4 6D 0, 0 < /2; ?i < 3/4 and 0 < /2; ni < 6/7. In these Raman spectra the
intensities have been measured using peak heights. An alternative method would have
been to use integrated areas. It should be appreciated that in some cases the two methods
I(p; R s, R i)
L s , R i)
+I(p;
z
700
500
300
100
1
n/cm
CCl4
I(p; R s, R i)
L s , R i)
I(p;
(a)
I(p; R s, R i)
z
800
x
600
400
1
n/cm
CHCl3
I(p; L s, R i)
(b)
R s,
R i C I ;
L s
R i and
Figure 5.8 (a) Stokes Raman spectra of CCl4 : I ;
s
i
s
i
R ,
R I ;
L ,
R . (b) Stokes Raman spectra of CHCl3 : (Q, 200800 cm1 ):
I ;
s
i
R ,
R and I ;
L s
R i .
I ;
See Chapter 3, Section 3.2 where the derived polarizability tensor is introduced and discussed.
113
of measuring intensities may give somewhat different results because depolarization ratios
can vary across the line width.
Figures 5.8(a) and (b) illustrate Raman spectra obtained with incident right circularly
polarized radiation. These spectra were obtained in the back-scattering configuration ( D ).
R s,
R i I ;
L s,
R i is negative for 1 a1
From Fig. 5.8(a) we see that for CCl4 , I ;
0
0
for which a 1 6D 0 and 1 D 0 and positive for the other vibrations for which a0 D 0
L s,
R i > I ;
R s,
R i for
and 0 6D 0. From Fig. 5.8(b) we see that for CHCl3 , I ;
0
0
1
1
the a1 vibrations for which a and 6D 0 (Q3 D 364 cm and Q 2 D 667 cm ), whereas
R s,
R i > I ;
L s,
R i for the e vibrations (Q6 D 260 cm1 and Q 5 D 760 cm1 )
I ;
0
0
for which a D 0 and 6D 0.
5.6 STOKES PARAMETERS FOR SCATTERED RADIATION

For completeness we now give expressions for the Stokes parameters of the scattered
radiation in terms of the invariants of the transition polarizability tensor and the Stokes
parameters of the incident radiation. In all cases the incident radiation is taken to have
n0i D ez and an irradiance I D 12 c0 0 E0i 2 where E0i is the amplitude of the electric
field strength of the incident radiation. The formulae below refer to scattering by N freely
rotating molecules with Nvi in the initial vibrational state vi and situated at the origin O
of the cartesian axis system.
For scattered radiation with n0s D ex , ( D /2), the Stokes parameters of the scattered
radiation at a distance x from the origin O along the direction defined by ex may be
shown to be

2 C 13 2
2 C 2
i
k
N
45a
45a

Q
v
Q 4

P cos 2 cos 2
E0i 2
S0s /2 D
x2 s
90
90
5.6.1

2
2
kQ Nvi 4
45a C
S1s /2 D
Q s
5.6.2
1 P cos 2 cos 2 E0i 2
2
x
90
S3s /2 D 0
5.6.3
S4s /2 D 0
5.6.4
where kQ is given by eq. (5.5.11).

For scattered radiation with n0s D ez ( D 0) or n0s D ez D the Stokes parameters
of the scattered radiation at a distance z from the origin O along the direction defined by
ez (or ez ) may be shown to be

kQ Nvi Q s4 45a2 C 7 2
s
s
E0i 2
5.6.5
S0 0 D S0 D
z2
45

4
2 C 2
i
k
N

Q
45a

Q
v s
S1s 0 D S1s D
P cos 2 cos 2
E0i 2
5.6.6
z2
45
114
The Raman Effect

45a2 C 2
E0i 2
P cos 2 sin 2
45

4
2 5 2
i
k
N

Q
45a
Q v s
S3s 0 D S3s D
P sin 2E0i 2
z2
45
S2s 0
S2s
kQ Nvi Q s4
D
z2
5.6.7
5.6.8
These Stokes parameters may be used to generate the intensities, depolarization ratios,
reversal coefficients, and degrees of circularity of the scattered radiation for incident
radiation of various states of polarization. We recall that for natural radiation P D 0. For
linearly polarized radiation P D 1, D 0 with D 0 if the incident radiation has Exi0 6D 0
(in this case, polarized parallel to the scatter plane) and D /2 if Eyi 0 6D 0 (in this case,
polarized perpendicular to the scatter plane). For circularly polarized incident radiation
P D 1, D 0, with D /4 for right circularly polarized radiation and D /4 for
left circularly polarized radiation.
We consider first intensities, and as examples we treat two cases, namely linearly
polarized scattered radiation and circularly polarized scattered radiation. Using the definitions of Stokes parameters given in Chapter A21, eqs. (A21.2.27) and (A21.2.28) and
adapting them to the geometry in use here, the Stokes parameters for scattered radiation
with n0s D ex D /2 and linearly polarized with Eys0 6D 0 ps D?s or linearly polarized
with Ezs0 6D 0 ps D jjs are found to be
S0s /2; ?s , pi D 12 fS0s /2 C S1s /2g
5.6.9
S1s /2; jjs , pi D 12 fS0s /2 S1s /2g
5.6.10
and
where S0s /2 and S1s /2 are given by eqs. (5.6.1) and (5.6.2), respectively. These
Stokes parameters relate squares of the amplitudes of the electric field strengths of the
scattered radiation at a distance x from the scattering molecule, to E0i 2 . To convert
the Stokes parameters of the scattered radiation to intensities we multiply by the factor
1
i 2
2
2 c0 0 x , and to convert E0 to I , the irradiance of the incident radiation, we use the
i 2
relationship E0 D 2I /c0 0 . We then obtain for the intensities

2 C 7 2
2 C 2
45a
45a
I /2; ?s , pi D kQ Nvi Q s4

P cos 2 cos 2
I 5.6.11
90
90
and

s
I /2; jj , p D
kQ Nvi Q s4
2
15

I
5.6.12
Using the definitions of Stokes parameters given by eqs. (A21.2.32) and (A21.2.35) in
Chapter A21 the Stokes parameters for scattered radiation with n0s D ez D 0 which is
We use the same labelling for Stokes parameters as that developed for intensities (Section 5.4.2).
This procedure is given in detail in Chapter A21, Section A21.2.4.
115
R s or
L s ) are found to be
right or left circularly polarized (ps D
R s , pi D 12 fS0s 0 C S3s 0g
S0s 0;
5.6.13
L s , pi D 12 fS0s 0 S3s 0g

S1s 0;
5.6.14
and
where S0s 0 and S3s 0 are given by eqs. (5.6.5) and (5.6.8), respectively. Converting to
intensities and introducing the irradiance of the incident radiation we obtain

45a2 C 7 2
45a2 5 2
s
i
4
R , p D kQ Nvi Q s
I0;
C
P sin 2 I 5.6.15
90
90

45a2 C 7 2
45a2 5 2
s
i
4
L , p D kQ Nvi Q s
I0;

P sin 2 I 5.6.16
90
90
Turning now to depolarization ratios, the general expression for /2; pi is (see
eq. (A21.4.6) in Chapter A21)
/2; pi D
I /2; jjs , pi
S0s /2 S1s /2
D
I /2; ?s , pi
S0s /2 C S1s /2
5.6.17
Introducing the expressions for the Stokes parameters given by eqs. (5.6.1) and (5.6.2)
we obtain
/2; pi D
6 2
45a2 C 7 2 45a2 C 2 P cos 2 cos 2
5.6.18
from which it is readily found that

3 2
45a2 C 4 2
/2; jji D 1
/2; ?i D
5.6.19
5.6.20
and
6 2
5.6.21
45a2 C 7 2
Similarly, using the following general expression for the reversal coefficient (see
eq. (A21.4.7) in Chapter A21),
S0 0 jS3 0j
D P 1
P 0 D
5.6.22
S0 0 C jS3 0j
we find that for either right or left circularly polarized incident radiation the reversal
coefficients are given by
/2; ni D
P 0 D
6 2
DP
45a2 C 2
1

5.6.23
116
The Raman Effect
Again, using the following general expression for the degree of circularity (see
eq. (A21.4.8) in Chapter A21)
C 0; pi D
S3 0
D C ; pi
S0 0
5.6.24
the degrees of circularity are given by

2 5 2
45a
C 0; pi D
P sin 2 D C ; pi
45a2 C 7 2
5.6.25
The Stokes parameters for various situations are summarized in Reference Table 5.2(e).
General formulae for the intensities and polarization properties of the scattered radiation
derived from the Stokes parameters are given in Reference Table 5.2(f).
5.7 SPECIFIC VIBRATIONAL TRANSITIONS

So far we have treated only a general vibrational transition from vi to vf . We now consider
specific vibrational transitions.
We recall that in the approximation considered in this chapter Q is a function of
the normal coordinates Q. We can therefore expand o Q as a Taylor series in Q, as
follows:

1
2
Q D 0 C
Qk C
Qk Ql C . . . 5.7.1
Qk 0
2 k l
Qk Ql
k
0
Introducing this expansion into eq. (5.2.3) we obtain

vf vi D 0 hvf jvi i

hvf jQk jvi i
C
Q
k
0
k

1 2
hvf jQk Ql jvi i C . . .
C
2 k l
Qk Ql
5.7.2
In the harmonic oscillator approximation, the total vibrational wave function is the
product of harmonic oscillator wave functions for each of the normal modes of vibration.
Thus, for vi and vf , we may write

vi D
vki Qk
5.7.3
k
vk D
vkf Qk
To simplify the nomenclature we write Q for O Q and so on.
5.7.4
117
where vki Qk and vkf Qk are harmonic oscillator wave functions associated with the
f
normal coordinate Qk , and have vibrational quantum numbers vik and vk in the initial and
final states, respectively. The product is taken over all normal coordinates. When these
expansions are introduced into eq. (5.7.2), and we neglect derivatives of higher than
the first, we obtain

vf vi D 0
vf Qk
vi Qk
k

Qk
vf Qk jQk j
k
vi Qk
5.7.5
This corresponds to the assumption of electrical harmonicity and leaves only the first two
terms in eq. (5.7.5) to be considered.
We now invoke the following properties of harmonic oscillator functions:

f
0 for vk 6D vik
5.7.6
hvf Qk jvi Qk i D
f
k
k
1 for vk D vik
f
0
for vk D vik
i
5.7.7
hvf Qk jQk jvi Qk i D vik C 11/2 bvk for vf
D
v
C
1
k
k
k
k
f
i 1/2
i
vk bvk
for vk D vk 1
where

bvk D

2k
1/2
h
8 2 c0 Q k
1/2
5.7.8
With this definition bvk has dimensions of (mass)1/2 (length)1 corresponding to massweighted normal coordinates Qk with dimensions of (mass)1/2 (length). Thus Qk bvk , is
dimensionless.
We can now determine the conditions that have to be satisfied if the first two terms in
eq. (5.7.5) are to be non-zero. The first term on the right-hand side of eq. (5.7.5) involves
0 the component of the equilibrium polarizability tensor and relates to Rayleigh
scattering. Because of the orthogonality of the harmonic oscillator functions, this term will
be non-zero if none of the vibrational quantum numbers change in going from the initial
f
to the final state, i.e. if vik D vk for all k. When this condition is satisfied, because of the
normalization properties of the harmonic oscillator functions, the integral in this term is
unity and the term reduces simply to 0 . As not all of the components of the equilibrium polarizability tensor of a system can be zero, Rayleigh scattering will always occur.
The second term on the right-hand side of eq. (5.7.5) involves components of a derived
polarizability tensor, and relates to Raman scattering. Consider the kth summand. This will
vanish unless every term in the product is non-vanishing, and to achieve this the following
conditions must be satisfied: for all modes except the kth, the vibrational quantum numbers
f
must be the same in the initial and final states (i.e. vj D vij except for j D k) and, for the
118
The Raman Effect

f
kth mode, the vibrational quantum number must change by unity (i.e. vk D vik 1). The
transition tensor component is associated with Stokes Raman scattering for vk D 1 (i.e.
f
f
vk D vik C 1), and with anti-Stokes Raman scattering for vk D 1 (vk D vik 1). These
conditions are a consequence of the properties of harmonic oscillator wave functions given
above. These arguments apply to each normal mode and to each of the components of
the derived polarizability tensor. Thus, the kth summand in the second term of eq. (5.7.5)
takes the values

f
f
i
1/2
vk C 1 bvk
if vk D vik C 1 and vj D vij for all other Qj 5.7.9
Qk 0
and
vik 1/2 bvk

Qk

f
if vk D vik 1 and vj D vij for all other Qj
5.7.10
where in the latter case vik 6D 0. It follows that in the approximation of electrical and
mechanical harmonicity

i
1/2
[v],vi C1:[v],vi D vk C 1 bvk
D vik C 11/2 bvk 0 k 5.7.11
k
k
Qk 0
and
[v],vi 1:[v],vi D

k
vik 1/2 bvk

Qk
D vik 1/2 bvk 0 k
5.7.12
where we use [v] to represent all those vibrational quantum numbers which are unchanged
in a transition and introduce the notation for derived polarizability components.
These arguments show that, in the approximation of electrical and mechanical harmonicity, only those transitions in which just one vibrational quantum number changes by
unity can be observed in the Raman effect. Usually, such transitions are fundamentals
for which in the case of Stokes Raman scattering vik D 0, but they can also be hot bands
for which vik D 1, 2 . . .. It must be noted, however, that the restriction vk D 1 is a
necessary, but not a sufficient, condition for the occurrence of Raman scattering associated
with the kth vibrational mode. It is also necessary that at least one of the elements of the
derived polarizability tensor be non-zero. This latter condition is exactly the same as that
obtained from the classical treatment.
If we do not assume electrical anharmonicity the third term in eq. (5.7.2) which involves
second derivatives of ( ) must be taken into account. Arguments similar to those used
above based on the properties of harmonic oscillator wave functions show that there are
other types of components of the vibration transition polarizability which can be non-zero
provided the appropriate second derivatives are non-zero. Two examples associated with
Stokes transitions are
119
[v],vi C2:[v],vi
k
which correspond to first overtones, and

[v],vi C1:vi C1:[v],vi ,vi
k
which correspond to binary combination tones. Expressions for a number of vibration

transition polarizability components associated with Stokes and anti-Stokes transitions
are given in Reference Table 5.2(g).
For the developments in the rest of this section we return to the case of mechanical
and electrical harmonicity and restrict our considerations to the kth vibrational mode.
Because of the factors vik C 1 and vik generated by the quantum mechanical treatment of
an harmonic oscillator, the intensity of the scattering from an assembly of N molecules is
not simply N times that from a single molecule. Although at normal temperatures most
of the molecules will be in the lowest vibrational state vik D 0, there will also be some
population of the higher vibrational states vik D 1, 2 etc.. Thus, in an assembly of N
molecules, the scattering associated with a given Stokes vibrational transition from vik
to vik C 1 depends on Nvik C 1fvi where fvi is the fraction of molecules in the state
k
k
vik . Since, in the harmonic approximation, the wavenumbers of the vibrational transitions
1
0, 2
1, 3
2 are identical, the calculation of the total Stokes scattered intensity
involves summing Nvik C 1fvi over all vik for a given mode k. Now, from the Bolzmann
k
distribution law,
expfvik C 1/2hc0 Q k /kTg
fvi D
k
expfvik C 1/2hc0 Q k /kTg
5.7.13
where Q k is the wavenumber associated with the 1

0 transition of the kth non-degenerate
vibrational mode and T is the absolute temperature. Utilizing this expression, it is found
that

N
vik C 1fvi D
5.7.14
N
k
1 expfhc0 Q k /kTg
i
For the associated anti-Stokes vibrational transitions from vik to vik 1 the calculation
of the total anti-Stokes scattered intensity involves summing Nvik fvi over all vik for a
k
given mode k. It is found that

N
vik fvi D
N
5.7.15
k
expfhc0 Q k /kTg 1
i
Thus in the harmonic oscillator approximation, two changes are necessary to the general
formulae developed earlier in this Chapter and given in Reference Tables 5.2(a)(f) for
the intensity of vibrational Raman scattering for various illuminationobservation geometries. For Raman scattering associated with the vibrational wavenumber Q k we must
replace the general polarizability tensor invariants a and by a0 k and 0 k the invariants of the derived polarizability tensor for the kth vibrational mode; or, alternatively,
120
The Raman Effect
replace the general Placzek invariants G 0 and G 2 by the specific invariants G 0
vk and
2
G vk . In addition for Stokes scattering we must replace the population factor Nvi by the
weighted population factor given by eq. (5.7.14). We give two examples below by way
of illustration. The wavenumber forms of eqs. (5.5.8) and (5.5.9) for I /2; ?s , ?i and
I /2; jjs , ?i are now replaced by

kQ hNQs4
45a0 2k C 4 0 2k
s
i
I /2; ? , ? D
5.7.16
I
8 2 c0 Q k f1 exphc0 Q k /kTg
45
and
kQ hNQs4
I /2; jjs , ?i D
8 2 c0 Q k f1 exphc0 Q k /kTg
0 2k
15

I
5.7.17
For anti-Stokes Raman scattering the intensity formulae corresponding to eqs. (5.7.16)
and (5.7.17) are obtained by substituting expfhc0 Q k /kTg 1 for 1 expfhc0 vQ k /kTg
and recognising that Q s D Q l C Q k . It follows that the ratio of Stokes Raman intensity at
Q s D Q l C Q k to anti-Stokes intensity at Q s D Q l C Q k is given by
I(Stokes)
Ql Q k 4
expfhc0 Q k /kTg
5.7.18
D
I(anti-Stokes)
Ql C Q k 4
for a given vibrational wavenumber Q k . It is evident from eq. (5.7.18) that the ratio
I (Stokes)/I (anti-Stokes) increases rapidly as Q k increases.
In the case of vibrational degeneracy if there are gvk vibrational modes each with the
same wavenumber Q k , the factor gvk must be included in the right hand side of intensity
expressions like those given by eqs. (5.7.16) and (5.7.17).
For depolarization ratios, reversal coefficients and degrees of circularity it is only necessary to make the replacements of a2 and 2 , by a0 2k and 0 2k , because these formulae
all involve ratios of intensities and so weighted population factors (and vibrational degeneracies) cancel.
5.8 VIBRATIONAL SELECTION RULES

It follows from the developments in Section 5.7 that for a general vibrational transition
from vi to vf to be Raman active at least one component of the transition polarizability
tensor xy vf vi must be non-zero. Thus, as
xy vf vi D hvf jO xy jvi i
5.8.1
the general rule for the vibrational transition vi to vf to be Raman active is that at least
one of the triple products vf xy vi belongs to a representation that contains the totally
symmetric species, that is
1 vf xy vi
5.8.2
where 1 is the totally symmetric representation. This general rule makes no statement
as to the intensity with which a permitted transition will appear in the Raman spectrum.
121
In some cases the intensity may be so low as to be unobservable experimentally. It is

even possible that the intensity may be fortuitously zero. Thus the general rule expresses
a necessary condition but not a sufficient one for Raman activity.
It should be noted that the general selection rule makes no assumptions as to whether
the vibrational functions are harmonic or anharmonic, nor does it involve the Taylor series
expansion of O xy . Thus the validity of the general selection rule transcends that of the
more selective selection rules which we consider subsequently.
The general selection rule itself is not completely rigorous because it depends on
the BornOppenheimer approximation and the assumption that all rotationvibration
interactions are negligible. The BornOppenheimer approximation is such a good one
that its use is very unlikely to affect the validity of the general rule. However,
vibrationrotation interactions and especially Coriolis forces are not always negligible and
may in certain cases render transitions Raman active that, according to the general rule,
would be inactive. The possibility of Coriolis interaction between different vibrational
modes is also subject to symmetry considerations and this restricts Coriolis-induced
breakdown of the general selection rule. The consideration of terms other than the dipole
term in the interaction hamiltonian is only important in certain special cases, as for
example, for scattering by chiral molecules.
The application of the general selection rule requires a knowledge of the symmetry
properties of the vibrational wave functions vi and vf , and of the polarizability tensor
components xy . We have shown in Chapter 4, Section 4 that xy transforms as xy so
that xy D xy. Thus the symmetry species of the transition polarizability tensor
components depend only on the point group to which a molecule belongs and so can be
determined once and for all for each point group. We shall not concern ourselves with
their determination. We give the symmetry species of the polarizability tensor components
in Reference Table 5.3 for all the point groups.
For meaningful discussions of the symmetry properties of the vibrational wave functions
it is necessary to consider specific situations. We begin with the selection rules for cases
where the total vibrational wave functions vi and vf are products of harmonic oscillator
wave functions for each of the normal modes of vibration and are defined by eqs. (5.7.3)
and (5.7.4).
We first consider a non-degenerate fundamental vibrational transition where in the
initial state all the vibrational quantum numbers are zero and in the final state only the
kth vibrational number has changed to unity. Using the notation introduced on page 118
f
we can write [v], vik D 0 for the initial state and [v], vk D 1 for the final state. We now
introduce eqs. (5.7.3) and (5.7.4) into eq. (5.8.1) and invoke the conditions contained in
eqs. (5.7.6) and (5.7.7). We then obtain for the kth mode
xy [v],vk D1:[v],vkD0 D h1 Qk jO xy j0 Qk i
5.8.3
It follows that the integral in eq. (5.8.3) is non-zero if the triple product 1 Qk xy0 Qk
contains the totally symmetric species, that is, if
1 1 Qk xy 0 Qk
5.8.4
122
The Raman Effect
As 0 Qk is totally symmetric the integral is non-zero if the product 1 Qk xy contains
the totally symmetric species, that is if
1 1 Qk xy
5.8.5
Now harmonic wave functions of the type k Qk always belong to the same symmetry
species as the vibrational mode Qk . Hence a fundamental vibrational mode Qk will be
Raman active if Qk and at least one of the products of the type xy belong to the same
symmetry species of the point group of the molecule because then Qk xy D 1 .
The conditions for Raman activity of doubly and triply degenerate vibrations are the
same. Consider for example the case of a doubly degenerate vibration Qk with normal
coordinates Qka and Qkb . This pair of normal coordinates will transform according to a
doubly degenerate species of the point group. If a pair of products of the type xy transform
in the same way, that is they belong to the same doubly degenerate representation, then
the mode will be Raman active. As we are dealing with a fundamental transition, the
energy level with vk D 1 can be realized in two ways: with vka D 1 and vkb D 0, or with
vka D 0 and vkb D 1. The case of a triply degenerate vibration Ql with normal coordinates
Qla , Qlb and Qlc is similar. The triple of normal coordinates Qla , Qlb and Qlc and a triple
of products of the type xy must belong to the same triply degenerate representation for the
mode to be Raman active. There are three ways of realizing the energy level vl D 1 for
a fundamental transition: (i) vla D 1, vlb D 0, vlc D 0; (ii) vla D 0, vlb D 1, vlc D 0; and
(iii) vla D 0, vlb D 0, vlc D 1.
We now consider transitions of the type [v], vk C 1
[v], vk where vk D 1, 2 . . . .
These arise when the vibrational levels vk D 1, vk D 2 are significantly populated and
give rise to so-called hot bands. In the harmonic approximation the wavenumber of each
of the hot bands associated with the kth mode is the same as that of the fundamental. For
such cases, the transition polarizability is given by
xy [v],vk C1:[v],vk D hvk C1 Qk jO xy jvk Qk i
5.8.6
The symmetry properties of the harmonic wave functions are such that the product
vk C1 Qk vk Qk transforms as Qk and thus the selection rule for a hot band is the
same as for the fundamental.
In reality, molecules display anharmonicity. We now need to explore how anharmonicity
affects the vibrational wave functions and the frequencies associated with transitions, and
the symmetry selection rules.
As far as wavenumbers are concerned we postpone a quantitative treatment until
Section 5.9 which deals with vibrational wavenumber patterns in Raman spectra. At
this stage we shall confine ourselves to a few qualitative observations. The effect of
anharmonicity is to modify the relative energies of the vibrational levels and hence
the wavenumbers associated with transitions are not quite the same as those expected
from the harmonic treatment. For example the wavenumbers of the hot bands associated
with the kth mode are all slightly different from each other and from the fundamental
wavenumbers. This gives rise to band broadening due to overlapping of bands of similar
wavenumbers, or in some cases to a set of distinct bands relatively closely spaced.
123
In contrast to the situation for the energy, the symmetry properties of anharmonic
wave functions are exactly the same as those of the corresponding harmonic wave
functions. Thus the symmetry requirements for Raman activity of a vibrational mode
which we have just deduced on the basis of harmonic wave functions apply when the
wavefunctions are anharmonic. Mathematically this is a consequence of the fact that anharmonic oscillator functions can always be expressed as linear combinations of harmonic
oscillator functions. Physically, we can recognize that this is because the presence of
anharmonicity, while modifying the classical form of vibration and its energy, cannot
change the symmetry of the potential energy so that the symmetry of the anharmonic
mode is still that of the harmonic mode. Now, the harmonic oscillator functions have
known mathematical forms whereas, in general, the anharmonic oscillator functions do
not, so it is a most fortunate circumstance that we can make deductions about symmetry
on the basis of the harmonic functions and apply them with equal validity to the anharmonic case.
As we have still not introduced any assumptions about electrical harmonicity the selection rule stated above is a truly general one for fundamental vibrations and related hot
bands. It is completely independent of any assumptions about mechanical or electrical
harmonicity.
We now consider how anharmonicity affects the selection rules. The selection rule
vk D 1 and the restriction of Raman activity to fundamental transitions and related
hot bands only, is a direct consequence of the properties of harmonic oscillator functions
given in eqs. (5.7.6) and (5.7.7). If the wave functions are anharmonic the selection rule
is much less restrictive. Transitions of the type vk D 2, 3 etc. are allowed, giving
rise to overtone bands. Combination and difference bands also arise from selection rules
of the types vk D 1 and vl D 1.
The effect of anharmonicity on the transition wavenumber means that the wavenumber
of an overtone is not an exact multiple of the fundamental wavenumber. Similarly, the
wavenumbers of combination and difference bands are not the algebraic sums of the fundamental wavenumbers involved. The intensities of overtones, combination and difference
bands are usually very small in the Raman effect because the second derivatives, such as
2 xy /Qk2 0 are small relative to the corresponding first derivatives xy /Qk 0 . This
means that the variation of xy with Qk around the equilibrium position is nearly linear
(see Fig. 3.3).
5.9 PATTERNS OF VIBRATIONAL SPECTRA

Now that the selection rules have been established, it is a straightforward matter to obtain
expressions for the magnitudes of the wavenumber shifts jQj (unit: cm1 ) observed in the
Raman effect as a result of permitted vibrational transitions. The absolute wavenumbers of
the Stokes and anti-Stokes Raman lines are, of course, given by Q 1 jQj and Q 1 C jQj
respectively where Q 1 is the wavenumber of the incident exciting radiation.
The energy of a harmonic oscillator Evib is given by
Evib D v C 1/2hc0 Q
5.9.1
124
The Raman Effect
G(v )
(a)
Harmonic
(c)
Anharmonic
=5
Stokes
=4
anti-Stokes
=1
=3
=2
= 1
(b)
(d)
n1
=1
n
increasing
=0
Figure 5.9 Vibrational terms (a) and Stokes and anti-Stokes vibration Raman spectra (b) for
a harmonic oscillator. Vibrational terms (c) and Stokes and anti-Stokes vibration Raman
spectra (d) for an anharmonic oscillator (xe positive). For clarity of presentation the effect
of anharmonicity has been magnified in (c) and (d).
where D cQ0 is the classical frequency of the oscillator. For spectroscopic purposes, it
Q v, where G
Q v D Evib /hco
is convenient to introduce the vibrational term denoted by G
and is in wavenumbers (unit cm1 ). Thus, in the harmonic approximation,
Q v D v C 1/2Q
G
5.9.2
The energy levels of a harmonic oscillator are therefore equally spaced (see Fig. 5.9(a)).
The vibrational selection rule v D 1 will thus lead to the appearance of a Raman line
associated with the vibrational transition vk 1
vk with a wavenumber shift equal to
Q k , the wavenumber of the kth normal mode, provided the symmetry properties of the kth
normal coordinate Qk and at least one of the polarizability tensor components satisfy the
necessary conditions. Thus, in the harmonic oscillator approximation, the Stokes and antivk
Stokes Raman lines associated with all transitions of the oscillator of the type vk 1
will have the same wavenumber shift irrespective of the value of vk (see Fig. 5.9(b)).
When anharmonicity is taken into account, the vibrational term for a diatomic molecule
has the form
Q v D v C 1/2Qe v C 1/22 Q e xe
G
5.9.3
where higher-order terms are neglected. Here Q e is the harmonic vibration wavenumber
and Q e xe is the vibrational anharmonicity constant, also in wavenumbers. The term Q e xe
is very much smaller than Q e and, for the choice of signs in eq. (5.9.3), is practically
always positive. In the anharmonic case, the energy levels are not equally spaced (see
Q e is the wavenumber associated with infinitely small vibrations about the equilibrium separation in the
actual potential field.
125
Fig. 5.9(c)). Applying the selection rule v D C1, the wavenumber shift of a Raman line
v will be given by
associated with the transition v C 1
jQj D Q e 2v C 1Qe xe
5.9.4
Thus, in the anharmonic oscillator approximation, the wavenumber shift of a Stokes or

anti-Stokes vibrational line associated with the kth normal mode now depends on the
vibrational quantum number v. Consequently, Raman wavenumber shifts arising from the
v will not coincide exactly (see Fig. 5.9(d)). For
various transitions of the type v C 1
the special case of the transitions 1
0 eq. (5.9.3) becomes
jQj D Q e 2Qe xe
5.9.5
The above formulae may be expressed in an alternative form. As the lower energy level
of the anharmonic oscillator corresponds to v D 0 we have from eq. (5.9.3)
Q
G0
D 12 Q e 14 Q e xe
5.9.6
Q
Then, if the vibrational energy levels are related to G0
Q 0 v D vQ0 v2 Q 0 x0
G
5.9.7
Q 0 D Q e Q e xe
5.9.8
Q 0 x0 D Q e xe
5.9.9
jQj D Q 0 Q 0 x0
5.9.10
where
and
In terms of Q 0 and x0
If the wavenumber shifts of, say, a fundamental transition v D 1 $ v D 0 and a first

overtone v D 2 $ v D 0 are observed, then it is possible to determine Q e and xe (or Q 0
and x0 ).
For a diatomic molecule, the above considerations suffice to determine the pattern of
the vibrational Raman spectrum, at least qualitatively. However, the actual wavenumber
shifts and their intensities depend on the mechanical and electrical properties of the
molecule, i.e. upon the masses and force constants and the invariants ak0 and k0 of the
derived polarizability tensor. In the case of polyatomic molecules, similar considerations
determine only the wavenumber pattern associated with a particular normal mode which
is Raman active. The number of such Raman active modes depends on the number of
atoms in the molecule and the symmetry of their spatial arrangement, and their relative
intensities depend on the relative values of ak0 and k0 . Thus, we can make no useful general
predictions about the pattern of a vibrational Raman spectrum. However, in specific cases
the number of vibrational Raman lines can be predicted and experience enables estimates
to be made of the likely wavenumbers and intensities of particular types of vibrations.
126
The Raman Effect
5.10 ORDERS OF MAGNITUDE

It follows from eq. (5.4.9) that we may write the expression for a wavenumber-normalized
scattering cross-section in the form
0 Qs 4 D kQ 2fi
5.10.1
where fi , is a general transition polarizability. For simplicity all the quantities involved
are treated as scalars. This practice will be followed throughout this section. The constant
kQ has the value 1.259 1023 C2 V2 m2 . Thus we can rewrite eq. (5.10.1) as
f 0 /m2 gfQs 4 /m4 g D f1.259 1023 /C2 V2 m2 gf2fi /C2 V2 m4 g 5.10.2
Choosing as a typical experimental value, fi D 5.0 1042 C V1 m2 , eq. (5.10.2)
leads to a value of 0 Qs 4 D 3.2 1060 m6 sr1 molecule1 . If we take Q s to have a
value of 2 106 m1 2 104 cm1 then 0 D 5.1 1035 m2 sr1 molecule1 .
If a reduced transition polarizability is used then we replace eq. (5.10.1) by
0 Qs 4 D
2 e4
2
20 h2 c02 fi
5.10.3
We now introduce the dimensionless fine structure constant fsc which is defined by
fsc D
e2
20 c0 h
5.10.4
and has the value 7.297 103 . We can now rewrite eq. (5.10.3) as
0 Qs 4 D 4 2 2fsc 2fi
or alternatively
0 /m2

Qs 4 /m4 D 2.10 103 2fi /m6
5.10.5
5.10.6
Using 0 Q s4 D 3.2 1060 m6 sr1 molecule1 , eq. (5.10.6) yields 2fi D 3.9
1029 m3 .
Adapting eq. (5.7.11) we may write for the fundamental vibration of a diatomic
molecule A2

1/2

h
1/2 1/2
vf D1:vi D0 D
mA Q
5.10.7
2
8 H c0
Q 0
where vf D1:vi D0 has units of C V1 m2 , H is the mass of the hydrogen atom in kg, mA
is the reduced molecular mass for the molecule A2 , Q is the wavenumber of the vibration
with units m1 and /Q0 is the derived polarizability at the equilibrium position
associated with the vibration. We may rewrite eq. (5.10.7) in the convenient form

The standard symbol for the fine structure constant is , but we have felt it expedient to use fsc in this
situation.
127
1/2
vf D0:vi D0 /C V1 m2 D f4.11 109 /m1/2 gmA
fQ1/2 /m1/2 gf/Q0 /C V1 mg
5.10.8
In the case of the diatomic molecule N2 , for which mA D 7 and Q D 2.331 105 m1
(i.e. 2331 cm1 ) eq. (5.10.8) yields

1
2
12
vf D1:vi D0 /C V m D 3.22 10 /m
C V1 m
5.10.9
Q 0
Thus if vf D1:vi D0 is taken to be of the order of 5.0 1042 C V1 m2 (see above),
/Q0 would be of the order of 1.6 1030 C V1 m. These values may be converted
to cgs units by dividing by 4 0 . We then obtain vf D1:vi D0 D 4.5 1032 m3 or 4.5
1026 cm3 and /Q0 D 1.4 1020 m2 or 1.4 1016 cm2 .
It is instructive to consider the scattered intensities associated with these transition
polarizabilities. We may write the general relation between p0 fi the amplitude of the
induced dipole and E0 the amplitude of the electric field strength of the incident electromagnetic radiation in the form
p0 fi D fi E0
5.10.10
We see that when fi is of the order of 5 1042 C V1 m2 , p0 fi will be of the order
of 5 1036 C m when E0 D 106 V m1 . This electric field strength would correspond
to an irradiance I of the order of 1010 W m2 which can be achieved with a laser of
power 1 W focused to a spot with an area of 1010 m2 , corresponding to a spot diameter
of 105 m.
It follows from eq. (2.2.3) that we may write for the scattered intensity I
I D k0Q Q s4 p0 2fi
5.10.11
where k0Q D 1.671 1020 C1 V m2 s1 and Q s has units of m1 . Thus if p0 fi D
5 1036 C m and Q s D 2 106 m1 we find that I D 6.7 1026 W sr1 molecule1 .
As the energy of the scattered photon is of the order of 4.0 1019 J, one photon per
steradian per second would require about 107 molecules in the focal volume. A typical
focal volume is 106 cm3 which for a gas at STP contains 3 1013 molecules. Thus we
could expect scattering of the order of 106 photons per steradian per second. For a liquid
such as CCl4 the same focal volume would contain about 6 1015 molecules, and the
scattering would be of the order of 2 109 photons per steradian per second.
5.11 EPILOGUE
It is evident from the treatments developed in this chapter that vibrational Raman spectra
are rich sources of molecular information. They provide vibrational wavenumbers which,
together with those obtained from infrared spectroscopy, can be used to calculate molecular
The reading list for Chapter 5 in Further Reading provides a selection of references in which the applications of vibrational Raman spectroscopy mentioned here are treated in detail.
128
The Raman Effect
force constants. Vibrational Raman intensities can also be analysed using bond polarizability theories to provide electro-optic parameters of bonds. The polarization properties
of vibrational bands give information on molecular symmetry. The ratio of Stokes and
anti-Stokes intensities of vibrational bonds can be used to measure temperature. The linear
dependence of intensity on concentration facilitates quantitative analysis.
The shape of a rovibrational Raman band can also yield important information. Although
the theory is not treated here, in essence the band shape is determined by the reorientational
motion of the molecules and by the relaxation of that particular vibrational mode. In totally
symmetric vibrations these two processes may be separated by determining Iiso and Ianiso
from appropriate Raman intensity measurements.
An application that is now assuming considerable importance is the use of vibrational
Raman spectra solely for the identification of molecular species. Until relatively recently
Raman spectra could only be obtained from appreciable amounts of materials and these
had to be free from fluorescent contaminants. Thus analytical applications were very
limited. However the situation has now been transformed completely by quite dramatic
advances in the techniques available for exciting, detecting and recording Raman spectra.
As a result vibrational Raman spectra can now be obtained from minute amounts of
material in almost any condition or environment. Raman spectra can be obtained from
particular points in a sample and it is possible, for example, to construct maps of the
amount of a particular species across a surface. Furthermore, vibrational Raman spectroscopy is now very effective for many aspects of process control including in-line
probing.
To analyse the Raman spectra obtained, good databases of vibrational Raman spectra are
needed. These are expanding rapidly and many are now available on line. They can be used
with search algorithms of various degrees of sophistication to identify unknown spectra.
From the wealth of examples available to illustrate the power of vibrational Raman
spectroscopy as an analytical tool we have chosen two examples.
Plate 5.1 shows two illuminated capitals (capital letters) from a manuscript produced in
the second half of the eleventh century. This manuscript consists of some 204 parchment
pages, of size 31 cm 22 cm. It is a copy of the treatise of St. Augustine on the Trinity
and was made for LAbbaye de la Trinite, Fecamp, in Normandy, France. It is now in
the Bibliothèque Nationale in Paris (reference BN latin 2088).
Raman spectroscopy has been used to analyse the colours used in these illuminated
capitals. The Raman spectra was obtained using a Raman microscope and minute samples
carefully removed from the manuscript. The colours lapis-lazuli blue, white lead (in a
superposition with lapis-lazuli blue) and vermilion were identified in the areas indicated
by arrows in Plate 5.1. The Raman spectra characteristic of these colours are shown in
Fig. 5.10. It should be noted that in some studies of early manuscripts the Raman spectra
can be recorded in situ when it is considered there is no possibility of damage to the
manuscript.
Plate 5.2 shows a cross-section (enlarged approximately 10 times) of a sandstone sample
collected from an area near East Beacon Ridge, South Victoria Land, Antarctica by a
See page 103 in this chapter.
129
Intensity
1500
1000
Wavenumber/cm1
500
Intensity
(a)
1300
1200
1100
1000
900
Wavenumber/cm1
800
Intensity
(b)
450
(c)
400
350
300
250
200
150
100
Wavenumber/cm1
Figure 5.10 Characteristic Raman spectra of (a) lapis-lazuli blue (b) white lead (basic lead
carbonate) (c) vermilion (mercury (II) sulphide).
130
The Raman Effect
British Antarctic Survey expedition in the summer of 199596. This region is characterized by rocky outcrops, snow cover and ambient temperatures of ca 25 C. Strong
katabatic winds blow off the polar plateau and the relative humidity in the dry valleys is
0%. The sample was stored at 5 C from the time of collection and was only allowed to
attain room temperature when it was analysed. Such samples of sandstone are inhabited
by cryptoendolithic lichen communities. Their study is of considerable interest because of
the information they could yield about the strategy adopted by such lichen communities
in order to survive in such a hostile environment.
Distinct layers of the sandstone sample can be identified as shown in Plate 5.2. These
are: a, iron stained crust; b, black lichen zone; c, hyaline lichen zone, mainly fungal; d,
microalgal zone; e, inorganic accumulation zone at ca 10 mm from the surface; f, bedrock.
Using Raman microscopy it was possible to obtain Raman spectra from each layer. These
are presented in Fig. 5.11 which also shows the assignment of bands to particular species.
These Raman spectroscopic studies can be used to characterize the pigments and
biomolecules produced by cyanobacteria in order to survive in the hostile environment
of the Antarctic. These compounds include those which can protect against harmful
Qua
Sil
COx
Chl
b
Car
Car
c
Intensity
CWa
Ami
d
f
3000
2500
2000
1500
1000
500
Wavenumber/cm1
Figure 5.11 Raman spectra of the six layers, a, b, c, d, e and f of the sandstone sample from
East Beacon Ridge (The spectra for different layers are displaced vertically). Characteristic
vibrational bands are labelled as follows: Car, carotene; Chl, chlorophyll; COx, calcium
oxalate; Ami, amide; CWa, cell wall materials; Qua, quartz and Sil, silica.
131
ultraviolet radiation, desiccation, low temperature and hypersalinity. The knowledge

gained from such studies could be used to determine whether there was or is microbial
life on Mars and other planets. To this end a miniature Raman microscope is under
development for use on a future Mars lander.
REFERENCES
Damen, T. C., Porto, S. P. S. and Tell, B. (1966). Phys. Rev. 142, 570.
Long, D. A. (1977). Raman Spectroscopy, McGraw Hill: New York.
132
The Raman Effect

REFERENCE TABLE 5.1: Definitions for I q; p s , p i
Symbol(s)
Definition
intensity of scattered radiation (unit: W sr1 )
angle between direction of observation in the scatter plane and

nzi D ez , the direction of propagation of the incident radiation
?i , ?s
radiation, linearly polarized with electric vector perpendicular to the

scatter plane (i D incident, s D scattered)
jji , jjs
radiation, linearly polarized with electric vector parallel to the scatter

plane (i D incident, s D scattered)
R i,
R s

L i,
L s
radiation right circularly polarized (i D incident, s D scattered)
i , s
radiation, of unspecified circular polarization (i D incident,

s D scattered)
ni
natural (unpolarized) incident radiation
total D?s C jjs

i
p ,p
radiation left circularly polarized (i D incident, s D scattered)
total scattered radiation, observed without an analyser in the beam

radiation, of unspecified state of polarization (i D incident,
s D scattered)
See Fig. 5.4.

Linear polarization of monochromatic radiation is defined in Section A21.2.1, right and left circular polarization in Section A21.2.2 and natural radiation in Section A21.3 of Chapter A21.
REFERENCE TABLE 5.2(a) TO 5.2(g): Intensities, polarization

properties and Stokes parameters for vibrational Raman (and
Rayleigh) scattering
(i) Introduction
This set of related tables gives the functions of the invariants of the several scattering
tensors necessary for calculating scattered intensities, depolarization ratios, reversal and
circularity coefficients, and Stokes parameters for vibrational transitions in an assembly of
randomly oriented molecules for various illuminationobservation geometries and states
of polarization of the incident radiation.
The treatment is quite comprehensive. It includes not only the case of a symmetric
transition polarizability tensor (involving only two invariants a2 and 2 ) applicable to the
non-absorbing systems considered in this chapter, but also the more general case where
133
the tensor is not a symmetric tensor (involving three invariants a2 , 2 and 2 ) applicable
to resonance scattering considered in Chapters 7 and 9.
Rayleigh and Raman scattering by chiral molecules is considered in Chapter 10, and
corresponding tables are given there.
(ii) Illuminationobservation geometry

For all the tables the illumination and observation geometry is defined as follows:
z
w
q
O
y
The scattering sample is located at the origin O and the direction of illumination is always
such that n0i D ez . The scatter plane is the xz plane, and a general direction of observation
in this plane is Ow, which makes an angle with the z axis; important special cases of
the observation direction are D 0 n0s D ez , D /2 n0s D ex , and D n0s D ez .
(iii) Vibrational Raman (and Rayleigh) scattering and the functions

f a 2 , g2 , d2 , q and F a 2 , g2 , d2 , q
Consider an assembly of N randomly oriented molecules irradiated with monochromatic
radiation of wavenumber Q 1 , irradiance I and polarization state pi . Then, as shown in
this chapter, I; ps , pi , the intensity of the radiation of polarization state ps scattered
along a general direction Ow defined by which arises from transitions from an initial
state defined by the set of vibrational quantum numbers vi to a final state defined by the
set of vibrational quantum numbers vf may be written in the general form

1
I; ps , pi D kvQ Nvi Q s4
fa2 , 2 , 2 , I
45
where kQ D 2 /20 and Q s D Q 1 jQvf vi j. Here jQvf vi j is the magnitude of the wavenumber
associated with the vibrational transition from vi to vf ; the C and signs refer to antiStokes and Stokes Raman scattering, respectively; Nvi is the number of molecules in
the initial vibrational state vi ; and fa2 , 2 , 2 , is a function of the invariants a2 ,
2 and 2 of the transition polarizability tensor avf vi and the angle , provided the
The simplified notation for invariants using a2 for a2vf vi and so on, is introduced in Section 5.2.1 of this
chapter.
134
The Raman Effect
molecules are non-chiral. The intensity of Rayleigh scattering may be obtained by putting
jQvf vi j D 0 and using the invariants of the equilibrium polarizability tensor a0 in place
of the transition polarizability tensor, since the initial and final states are the same and
vf D vi . Where vibrational states are degenerate more than one transition of the same
type will be associated with the same wavenumber shift. The observed intensity is then
the sum of the intensities associated with each transition and the vibrational degeneracy
must be included in the intensity formulae.
For illumination with linearly polarized radiation having the electric vector perpendicular to the scatter plane pi D?i Ey0 6D 0, Ex0 D 0, I D 12 c0 0 Ey20 , the function fa2 , 2 ,
2 , is given in Reference Table 5.2(a) for various observation geometries and polarization states of the scattered radiation; the general formulae for depolarization ratios
; pi are also included. The corresponding information for incident radiation linearly
polarized with the electric vector parallel to the scatter plane pi D jji , Ex0 6D 0, Ey0 D 0,
I D 12 c0 0 Ex20 is given in Reference Table 5.2(b), and for incident natural radiation
pi D ni , Ex0 D Ey0 6D 0, I D 12 c0 0 [Ex20 C Ey20 ] in Reference Table 5.2(c). The case of
R i , Ey0 D iEx0 or pi D
L i , Ey0 D iEx0
incident circularly polarized radiation pi D
1
2
2
and I D 2 c0 0 [Ex0 C Ey0 ] is dealt with in Reference Table 5.2(d) which includes general
formulae for P and C ; pi
The properties of the radiation scattered along Ow may be expressed alternatively in
terms of the four Stokes parameters, S0s , S1s , S2s and S3s , each of which may be
expressed in the general form

kQ Nvi Q s4 1
s
2
2 2
Fa , , , , P, , E0i 2
S D
2
r
45
where r is the distance along the observation direction at which the observation is made, E0i
refers to the incident radiation, and Fa2 , 2 , 2 , , P, , is a function of the invariants
a2 , 2 , and 2 , the angle and the polarization characteristics P, and of the incident
radiation, provided the molecules are non-chiral.
The function Fa2 , 2 , 2 , , P, , is given in Reference Table 5.2(e) for the observation angles D 0, /2 and , when the incident radiation has the general polarization state
P, , ; or is natural P D 0; or has some of the following special polarization states:
pi
jji
?i
1
1
1
1
0
0
/4
/4
R i

L i
0
/2
0
0
The Stokes parameters of the scattered radiation may be used to express any desired
property of the scattered radiation, as a function of a2 , 2 , 2 , , P, and . The appropriate
We use F for a function which relates Stokes parameters of the scattered radiation to E0i in contrast to
f which is used for functions which relate intensities and polarization properties of the scattered radiation
to I .
135
R i , pi
forms of the general function fa2 , 2 , 2 , P, , for I; jjs , pi , I; ?s , pi , I;
s
i
i
i
L p together with expressions for ; p , P and C , p for the special
and I;
cases of D 0, /2 and are given in Reference Table 5.2(f) for a general polarization
state of the incident radiation P, , . When the values of P, and appropriate to the
R i or
L i are introduced into the general expressions
special polarization states jji , ?i , ni ,
i
i
for intensities, ; p , P and C ; p given in Reference Table 5.2(f), they reduce to
those in Tables 5.2 (a)(d).
When the polarizability theory is applicable (see Chapter 4, Section 4.2 for the conditions to be satisfied), the transition polarizability tensor is always symmetric and individual
components may be related to derivatives of the polarizability tensor with respect to
normal coordinates of vibration, the order of the derivatives, and the number of normal
coordinates involved depending on the vibrational transitions involved. The relationships
between the transition polarizability tensor and the derivatives of the polarizability tensor
are summarized in Reference Table 5.2(g) for the following transitions:
vk D 1; vk D 2; vk D C1, vl D C1; vk D 1, vl D C1; and v D 0

The general results in Reference Tables 5.2(a) to 5.2(f) may be used together with
Reference Table 5.2(g) to obtain intensity formulae in the polarizability theory approximation. For example, for the single fundamental transition vk D C1, intensity formulae
may be obtained from the general formulae, if 2 is put equal to zero and a2 and 2 are
regarded as invariants of the symmetric tensor a0 k with elements xy /Qk 0 and are
multiplied by bv2k vk C 1. Similar procedures apply to other transitions.
In the simple harmonic approximation, all transitions of the type vk D C1 have the
vk D 0 cannot be separated from
same wavenumber and thus the transition vk D 1
vk D 2
vk D 1, and so on. The observed intensity is thus obtained by summing the
intensities associated with each transition of the type vk D C1, taking into account the
population and the factor bv2k vk C 1 for each initial state. Again, the required intensity
formulae can be obtained from the general formulae if 2 is put equal to zero and a2
and 2 are regarded as invariants of the symmetric tensor a0 k and are multiplied by
bv2k f1 exphc0 Q k /kTg1 , and Nvi is replaced by N. For the transitions vk D 1 the
multiplying factor is bv2k fexphcQk /kT 1g1 . If the kth mode has a degeneracy of gvk
the observed intensity will be gvk times greater. In formulae for quantities such as , P
and C which involve intensity ratios, the multiplying factors such as bv2k vk C 1 cancel
and the only changes needed in the general formulae are to put 2 equal to zero and to
redefine a2 and 2 as invariants of a0 k .
(iv) Wavenumber-normalized first-differential cross-sections

The function fa2 , 2 , 2 , is related to the wavenumber-normalized first-differential
cross-section of a single molecule, 0 Qs 4 (see Section 5.4.2) as follows:

1
I; ps , pi
0
4
2
2 2
Qs D
fa , , ,
D kQ
Nvi Qs 4 I
45
Thus the function fa2 , 2 , 2 , given in these tables is directly proportional to 0 Qs 4 .
136
The Raman Effect
(v) Tensor decomposition

The method used in Reference Tables 5.2(a) to 5.2(d) for presenting the function
fa2 , 2 , 2 , enables the scalar part fa2 , , the anisotropic part f 2 , and the scalar
plus anisotropic parts fa2 , 2 , to be extracted readily if required. The depolarization
ratios, reversal coefficients and degrees of circularity can be decomposed similarly.
(vi) Placzek invariants

As explained in Chapter A14, Section A14, the Placzek invariants G j may be used
instead of a2 , 2 and 2 . The formulae in Reference Tables 5.2(a) to 5.2(g) may be
changed to those based on the Placzek invariants by using the following replacements:
a2 D 13 G
0
2 D 32 G
1
2 D 32 G
2
(vii) Units
Throughout the tables intensities have units of W sr1 and fa2 , 2 , 2 , units of
C2 V2 m4 sr1 . The first differential cross-section has units of m2 sr1 molecule1 ,
the frequency-normalized first-differential cross-section has units of m6 sr1 molecule1 ,
when the wavenumber Q has SI units of m1 . As kQ D 1.259 1023 C2 V2 m2 we can
write for one scattering molecule
0 Q s4 /m6 sr1 D f1.259 1023 /C2 V2 m2 gffa2 , 2 , 2 , g C2 V2 m4 sr1
I/W sr1 D f1.259 1023 /C2 V2 m2 gfQs 4 /m4 g
ffa2 , 2 , 2 , C2 V2 m4 sr1 gI /W m2
Conversions to photon counts are dealt with in Chapter A17, Section A17.2.3.
137
Reference Table 5.2(a) Intensities and depolarization ratios for incident linear polarized radiation
?i ; D , D /2, D 0 and D .
General formulae:

s
I; p , ? D
kQ Nvi Q s4

1
2
2 2
fa , , , I
45
where fa2 , 2 , 2 , is a linear combination of the stated invariants whose coefficients are
given below and include any explicit dependence on :
; ?i D
I; jjs , ?i
I; ?s , ?i
Depolarization ratios are not given explicitly but follow immediately using the above formula
and the appropriate intensity formulae I; ps , ?i .
I; ps , ?i
Illuminationobservation
geometry
Coefficients of
a2
2
45
I; ? C jj , ?
45
I /2; ?s , ?i
45
I /2; jjs , ?i
I /2; ?s C jjs , ?i
45
I0; ?s , ?i D I ; ?s , ?i
45
I0; jjs , ?i D I ; jjs , ?i
45
Ow D 0
z
w
I; ?s , ?i
q
I; jj , ?
O
y
Ox D /2
z
O
x
Oz D 0 and Oz D
z
O
x
I0; ?s C jjs , ?i D I ; ?s C jjs , ?i
138
The Raman Effect
Reference Table 5.2(b) Intensities and depolarization ratios for incident linear polarized radiation jji ; D , D /2, D 0 and D .
General formulae:

1
s
i
4
2
2
2
fa , , , I
I; p , jj D kQ Nvi Q s
45
; jji D
I; jjs , jji
I; ?s , jji
Depolarization ratios use not given explicitly but follow immediately using the above formula
and the appropriate intensity formulae I; ps , jji .
I; ps , jji
geometry
Coefficients of
a2
2
Ow D 0
z
w
I; ?s , jji
q
O
y
I; jj , jj
45c
(3 C c )
5s2
I; ?s C jjs , jji
45c2
(6 C c2 )
5(1 C s2 )
I /2; ?s , jji
I /2; jjs , jji
I /2; ?s C jjs , jji
10
I0; ?s , jji D I ; ?s , jji
I0; jjs , jji D I ; jjs , jji
45
I0; ?s C jjs , jji D I ; ?s C jjs , jji
45
Ox D /2
z
O
x
Oz D 0 and Oz D
z
O
x
c D cos , s D sin .
139
Reference Table 5.2(c) Intensities and depolarization ratios for incident natural (unpolarized)
radiation ni ; D , D /2, D 0 and D .
General formulae:

1
s
i
4
2
2
2
fa , , , I
I; p , n D KQ Nvi Q s
45
I; jjs , ni
I; ?s , ni
Depolarization ratios are not given explicitly but follow immediately using the above formula
and the appropriate intensity formulae I; ps , ni .
; ni D
geometry
I; ps , ni
Coefficients of
a2
2
45
2
45c2
2
451 C c2
2
7
2
6 C c2
2
13 C c2
2
5
2
51 C s2
2
52 C s2
2
I /2; ?s , ni
45
2
7
2
5
2
I /2; jjs , ni
I /2; ?s C jjs , ni
45
2
13
2
15
2
I0; ?s , ni D I ; ?s , ni
45
2
7
2
5
2
I0; jjs , ni D I ; jjs , ni
45
2
7
2
5
2
I0; ?s C jjs , ni
45
Ow D 0
z
w
I; ?s , ni
q
I; jjs , ni
O
y
I; ?s C jjs , ni
Ox D /2
z
O
x
Oz D 0 and Oz D
D I ; ?s C jjs , ni
c D cos , s D sin .
140
The Raman Effect
Reference Table 5.2(d) Intensities, reversal coefficients and degrees of circularity for incident
circularly polarized radiation i ; D , D /2, D 0 and D .
General formulae:
I; s , i D kQ Nvi Q s4

1
fa2 , 2 , 2 , I
45
given below and include any explicit dependence on .
L s,
R i
R s,
L i
I;
I;
D
DP
R s,
R i
L s,
L i
I;
I;
P D
R i D
C ;
1
C
R s,
R i I;
L s,
R i
I;
R i
D C C ,
R i
I; totals ,
R i D C ;
L i
C ;
Illumination observation
geometry
I; s , i
Coefficients of
a2
2
451 C 2c C c2
4
13 10c C c2
4
53 C 2c c2
4
451 2c C c2
4
13 C 10c C c2
4
53 2c c2
4
45
4
13
4
15
4
45
4
13
4
15
4
Ow D
z
R i
R s,
I;
L
L ,
D I;
s
O
y
L s,
R i
I;
R ,
L
D I;
s
Ox D /2
z
R s,
R i
I /2;
L
L ,
D I /2;
s
O
x
L s,
R i
I /2;
L i
R s,
D I /2;
141
Reference Table 5.2(d) (continued )

Illumination observation
geometry
I; s , i
Coefficients of
a2
2
45
Oz D 0 and
Oz D
R s,
R i
I0;
L i
L s,
D I0;
s
R ,
L i
D I ;
s
R i
L ,
D I ;
O
x
c D cos , s D sin .
L s,
R i
I0;
R s,
L i
D I0;
s
R ,
R i
D I ;
s
L ,
L i
D I ;
Oz D 0 and
Oz D
0
0
y x
S3 0 D S3
O
y
45a2 C 2 52
45a2 C 2 52
45a2 C 7 2 C 52 45a2 C 7 2 C 52
0
0
45a2 C 2 52
45a2 C 7 2 C 52
6 2 C 102
P D 1, D 0,
D /2
(i.e. Ey 6D 0)
P D 1, D 0,
D0
(i.e. Ex 6D 0)
S2 0 D S2
S1 0 D S1
S0 0 D S0
S2 /2
S3 /2
S1 /2
S0 /2
Ox D /2
Stokes
parameter
Illumination
observation
geometry
P D 1, D /4,
D0
(i.e. Ey D iEx
P D P, D , D
(general polarization)
O
y
45a2 C 7 2 C 52
0
0
45a2 C 2 52
2
45a2 5 2 C 52 P sin 2
45a2 5 2 C 52
45a2 C 2 52 P cos 2 sin 2
45a2 C 2 52 P cos 2 cos 2
45a2 C 7 2 C 52
0
0
45a2 C 2 52 1 P cos 2 cos 2

2
45a2 C 7 2 C 52
0
0
45a2 C 2 52
2
45a2 C 13 2 C 152 45a2 C 13 2 C 152 45a2 C 13 2 C 152 45a2 C 2 52 P cos 2 cos 2
2
2
2
PD0
(i.e. Ex D Ey 6D 0)
where r is the distance along the observation direction at which the observation is made, Ei0 refers to the incident radiation, and Fa2 , 2 , 2 , , P, , is a function
of the invariants a2 , 2 , and 2 , the angle and the polarization characteristics P, and of the incident radiation.
The function Fa2 , 2 , 2 ., P, , is given below for the observation angles D 0, /2 and , when the incident radiaton has the general polarization state P,
and ; or is natural P D 0; or has certain special polarization states.
Each of the four Stokes parameters, S0s , S1s , S2s and S3s , can be expressed in the general form

kQ Nvi Q s4 1
2 , 2 , 2 , , P, , Ei 2
Fa
Ss D
0
r2
45
Reference Table 5.2(e) Fa2 , 2 , 2 , , P, , for Stokes parameters for D /2, D 0 and D ; general and special polarization
states of incident radiation.
142
The Raman Effect
143
R s , pi ;
Reference Table 5.2(f) fa2 , 2 , 2 , , P, , for I; jjs , pi ; I; ?s , pi ; I,
s
i
i
i
L , p ; ; p ; P ; C ; p ; D 0, /2 and .
I;
General formulae

1
fa2 , 2 , 2 , P, ,
I; ps , pi D kQ Nvi Q s4
45
The function fa2 , 2 , 2 , P, , is given below for various ps and D 0, /2 and . General
formulae for ; pi , p; pi and C ; pi are also given for D 0, /2 and .
fa2 , 2 , 2 , , P, , , P or C
Illumination
observation
geometry
Quantity observed
Oz D 0 and
I0; jjs , pi
D I ; jjs , pi
45a2 C 7 2 C 52 C 45a2 C 2 52 P cos 2 cos 2

2
I0; ?s , pi
D I ; ?s , pi
45a2 C 7 2 C 52 45a2 C 2 52 P cos 2 cos 2

2
Oz D
R s , pi
I0;
L s , pi
D I ;
45a2 C 7 2 C 52 C 45a2 5 2 C 52 P sin 2

2
L s , pi
I0;
R s , pi
D I ,
O
x
I0; jjs , pi
I0; ?s , pi
P 0; pi D
P 1 ; pi
45a2 C 7 2 C 52 45a2 5 2 C 52 P sin 2

2
45a2 C 7 2 C 52 C 45a2 C 2 52 P cos 2 cos 2
45a2 C 7 2 C 52 45a2 C 2 52 P cos 2 cos 2
45a2 C 7 2 C 52 j45a2 5 2 C 52 P sin 2j

45a2 C 7 2 C 52 C j45a2 5 2 C 52 P sin 2j

C 0; pi D
C , pi
I /2; jjs , pi
Ox D /2
45a2 5 2 C 52
45a2 C 7 2 C 52
P sin 2
3 2 C 52
I /2; ?s , pi
O
x
45a2 C 7 2 C 52 45a2 C 2 52 P cos 2 cos 2

2
(continued overleaf )
144
The Raman Effect
Reference Table 5.2(f) (continued )

R s , pi
I /2;
45a2 C 13 2 C 152 45a2 C 2 52 P cos 2 cos 2

4
L s , pi
I /2;
45a2 C 13 2 C 152 45a2 C 2 52 P cos 2 cos 2

4
I /2; jjs , pi
I /2; ?s , pi
6 2 C 102
45a2 C 7 2 C 52 45a2 C 2 52 P cos 2 cos 2
P /2; pi
C /2; pi
This yields 0; ?i , 0; ni and f0; jji g1 ; 0; pi D ; pi

R i D P 0;
L i .
P 0;
This yields /2; ?i , /2; ni and f /2; jji g1 .

General formulae relating polarization properties to Stokes parameters and scattered intensities to Stokes
parameters will be found on pages 113 116.
Reference Table 5.2(g) Vibration transition polarizability components for the transitions
vk D 1; vk D 2; vk D C1, vl D C1; vk D 1, vl D 1, and v D 0, (assuming
the Placzek polarizability theory is applicable).
Final vibration
state
Initial vibration
state
Section rule
[v], vk C 1
[v], vk
vk D C1
[v], vk 1
[v], vk
vk D 1
[v], vk C 2
[v], vk
vk D C2
[v], vk 2
[v], vk
vk D 2
[v], vk C 1, vl C 1
[v], vk , vl
vk D C1,
vl D C1
[v], vk 1, vl 1
[v], vk , vl
vk D 1,
vl D 1
v D 0
[v]
[v]

[xy ]vf vi

1
xy
vk C 1 2
bv
Qk 0 k

1
xy
vk 2
bv
Qk 0 k
2

1
1
xy
fvk C 1vk C 2g 2
b2
2
Qk2 0 vk
2

1
1
xy
fvk 1vk g 2
b2
2
Qk2 0 vk
2

1
1
xy
bv bv
fvk C 1vl C 1g 2
2
Qk Ql 0 k l
1
1
fvk vl g 2
2
2 xy
Qk Ql
xy 0
1

1

2
2
h
h
1/2
bvk D
D
D
8 2 c0 Q k
4 k
2k
[v] denotes those vibrational quantum numbers which are unchanged in the transition.
Derivatives higher than the second are not considered. c0 = speed of light in vacuum.
bvk bvl
0
145
REFERENCE TABLE 5.3: Symmetry classes for x , y, z , the rotations

Rx , Ry and Rz , and the components of the cartesian basis tensor c a.
The following tables give, for all the common point groups, the symmetry classes to which
belong the cartesian displacements x, y, z, the rotations Rx , Ry and Rz , and the components
of c a. To keep the tables reasonably compact the corresponding information is not
given explicitly for either the components of the s a, the tensors in the spherical
j
basis, or the components m of the a, the tensors in the irreducible basis. The symmetry
properties of the may be obtained quite readily from those of the by using
Table A14.1 which gives the matrices B and B relating the components of c a and
j
of s a. The symmetry properties of the m may be obtained quite readily from those
of the by using Table A14.3 in Chapter A14 which gives the matrices A and A
j
relating the components of c a and m of IR a.
In general these tensors are not symmetric, but as they can become symmetric in a
number of situations, the symmetric and anti-symmetric parts are listed separately: the
symmetric tensor components of a are denoted by , and the anti-symmetric tensor
j
components by . The symmetric components of m will be those with j D 0 and 2
and the anti-symmetric components those with j D 1.
Note the use of superscripts a, b and c in these tables:
b D x axis along C02 :
a D x axis along C2 :
c D x axis in v .
The groups Cs , Ci and Cn n D 2, 3, 4, 5, 6, 7, 8

Cs
A0
A00
Ci
Ag
x, y
z
Rz
R x , Ry
xx , yy , zz , xy
yz , xz
xy
yz , xz
R x , Ry , Rz
xx , yy , zz
xy , yz , zx
xy , yz , zx
Au
x,y,z
C2
A
B
z
x,y
Rz
Rx , R y
xx , yy , zz , xy
yz , zx
xy
yz , zx
C3
A
E
z
(x, y)
Rz
Rx , Ry
xx C yy , zz
xx yy , xy ,
yz , zx
xy
yz , zx
C4
A
B
E
Rz
xy
(x, y)
Rx , Ry
xx C yy , zz
xx yy , xy
yz , zx
A
E1
E2
z
(x, y)
Rz
Rx , Ry
xx C yy , zz
yz , zx
xx yy , xy
xy
yz , zx
C5
yz , zx
Chapter A14 treats the polarizability tensor in the cartesian, spherical and irreducible bases in detail.
146
The Raman Effect

Reference Table 5.3
C6
(continued )
A
B
E1
E2
Rz
xx C yy , zz
xy
(x, y)
Rx , Ry
yz , zx
xx yy , xy
yz , zx
C7
A
E1
E2
E3
z
(x, y)
Rz
Rx , Ry
xx C yy , zz
yz , zx
xx yy , xy
xy
yz , zx
C8
A
B
E1
E2
E3
Rz
xx C yy , zz
xy
(x, y)
Rx , Ry
yz , zx
xx yy , xy
yz , zx
The groups Dn n D 2, 3, 4, 5, 6
D2
D3
D4
D5
D6
A
B1
B2
B3
z
y
x
Rz
Ry
Rx
Aa1
Aa2
E
z
(x, y)
Rz
Rx , Ry
Rz
A1
A2
B1
B2
E
A1
A2
E1
E2
A1
A2
B1b
B2b
E1
E2
xx , yy , zz
xy
zx
yz
xy
zx
yz
xx C yy , zz
xx yy , xy ,
yz , zx
xy
yz , zx
xx C yy , zz
(x, y)
Rx , Ry
z
(x, y)
Rz
Rx , Ry
Rz
(x, y)
Rx , Ry
xy
xx yy
xy
yz , zx
yz , zx
xx C yy , zz
yz , zx
xx yy , xy
xy
yz , zx
xx C yy , zz
xy
yz , zx
xx yy , xy
yz , zx

Reference Table 5.3
147
(continued )
The groups Cnv n D 2, 3, 4, 5, 6

C2v
C3v
C4v
C5v
C6v
A1
A2
B1
B2
Ac1
Ac2
E
A1
A2
B1
B2
E
x
y
x, y
Rz
Ry
Rx
xx , yy , zz
xy
zx
yz
xx C yy , zz
Rz
Rx , Ry
xx yy , xy ,
yz , zx
A1
A2
E1
E2
A1
A2
B1c
B2c
E1
E2
(x, y)
Rx , Ry
xy
xx yy
xy
yz , zx
yz , zx
xx C yy , zz
Rz
Rx , Ry
yz , zx
xx yy , xy
xy
yz , zx
xx C yy , zz
Rz
(x, y)
xy
yz , zx
xx C yy , zz
Rz
(x, y)
xy
zx
yz
Rx , Ry
xy
yz , zx
xx yy , xy
yz , zx
Rz
Rx , R y
xx , yy , zz , xy
yz , zx
xy
yz , zx
Rz
xx C yy , zz
xx yy , xy
xy
Rx , Ry
yz , zx
yz , zx
Rz
xx C yy , zz
xx yy , xy
yz , zx
xy
The groups Cnh n D 2, 3, 4, 5, 6

C2h
C3h
C4h
Ag
Bg
Au
Bu
A0
E0
A00
E00
Ag
Bg
Eg
Au
Bu
Eu
z
x,y
(x, y)
z
Rx , Ry
yz , zx
z
(x, y)
148
The Raman Effect

Reference Table 5.3
C5h
C6h
(continued )
A
E01
E02
A00
E001
E002
Ag
Bg
E1g
E2g
Au
Bu
E1u
E2u
Rz
xx C yy , zz
xy
(x, y)
xx yy , xy
z
Rx , Ry
yz , zx
yz , zx
Rz
xx C yy , zz
xy
Rx , Ry
yz , zx
xx yy , xy
yz , zx
z
(x, y)
The groups Dnh n D 2, 3, 4, 5, 6

D2h
D3h
D4h
Ag
B1g
B2g
B3g
Au
B1u
B2u
B3u
Rz
Ry
Rx
A1g
A2g
B1g
B2g
Eg
A1u
A2u
B1u
B2u
Eu
xy
zx
yz
z
y
x
A1a
0
A2a
E0
A001
A002
E00
xx , yy , zz
xy
zx
yz
xx C yy , zz
Rz
(x, y)
xy
xx yy , xy
z
Rx , Ry
yz , zx
yz , zx
xx C yy , zz
Rz
Rx , Ry
z
(x, y)
xy
xx yy
xy
yz , zx
yz , zx

Reference Table 5.3
D5h
D6h
A01
A02
E01
E02
A001
A002
E001
E002
A1g
A2g
B1g
B2g
E1g
E2g
A1u
A2u
b
B1u
b
B2u
E1u
E2u
149
(continued )
xx C yy , zz
Rz
xy
(x, y)
xx yy , xy
z
Rx , Ry
yz , zx
yz , zx
xx C yy , zz
Rz
Rx , Ry
xy
yz , zx
xx yy , xy
yz , xz
(x, y)
The groups Dnd n D 2, 3, 4, 5, 6

D2d
D3d
A1
A2
B1
B2
E
z
(x, y)
A1g
A2g
Eg
Aa1u
Aa2u
Eu
D4d
xx C yy , zz
Rz
A1
A2
B1
B2
E1
E2
E3
Rx , Ry
xy
xx yy
xy
yz , zx
yz , zx
xx C yy , zz
Rz
Rx , Ry
xx yy , xy ,
yz , zx
xy
yz , zx
z
(x, y)
xx C yy , zz
Rz
xy
z
(x, y)
Rx , Ry
xx yy , xy
yz , zx
yz , zx
(continued overleaf)
150
The Raman Effect

Reference Table 5.3
D5d
D6d
A1g
A2g
E1g
E2g
A1u
A2u
E1u
E2u
A1
A2
B1
B2
E1
E2
E3
E4
E5
(continued )
xx C yy , zz
Rz
Rx , Ry
yz , zx
xx yy , xy
xy
yz , zx
z
(x, y)
xx C yy , zz
Rz
xy
z
(x, y)
xx yy , xy
Rx , Ry
yz , zx
yz , zx
Rz
xx C yy , zz
xx yy , xy
yz , zx
xy
The groups Sn n D 4, 6, 8
S4
S6
A
B
E
Ag
Eg
Au
Eu
S8
z
(x, y)
A
B
E1
E2
E3
Rx , Ry
yz , zx
Rz
Rx , Ry
xx C yy , zz
xx yy , xy ,
yz , zx
xy
yz , zx
Rz
xx C yy , zz
xy
Rx , Ry
xx yy , xy
yz , zx
yz , zx
z
(x, y)
z
(x, y)

Reference Table 5.3
151
(continued)
The groups T, Th , Td , O and Oh

T
A
E
F
Th
Rx , Ry , Rz
xx C yy C zz
xx C yy 2zz ,
xx yy
xy , yz , zx
xy , yz , zx
(x, y, z)
xx C yy C zz
xx C yy 2zz ,
xx yy
Rx , Ry , Rz
(x, y, z)
xx C yy C zz
xx C yy 2zz ,
xx yy
(x, y, z)
Rx , Ry , Rz
xy , yz , zx
xy , yz , zx
A1g
A2g
Eg
F1g
F2g
A1u
A2u
Eu
F1u
F2u
xy , yz , zx
xy , yz , zx
A1
A2
E
F1
F2
Oh
xy , yz , zx
A1
A2
E
F1
F2
Rx , Ry , Rz
Ag
Eg
Fg
Au
Eu
Fu
Td
(x, y, z)
xx C yy C zz
xx C yy 2zz ,
xx yy
xy , yz , zx
xx C yy C zz
(xx C yy 2zz ,
xx yy
Rx , Ry , Rz
xy , yz , zx
xy , yz , zx
(x, y, z)
(continued overleaf)
152
The Raman Effect
Reference Table 5.3
(continued)
The groups I and Ih

I
Ih
A
T1
T2
G
H
xx C yy C zz
x, y, z
yz , zx , xy
2zz xx yy , xx yy ,
xy , yz , zx
Ag
T1g
T2g
Gg
Hg
Au
T1u
T2u
Gu
Hu
Rx , R y , R z
xx C yy C zz
Rx , R y , R z
yz , zx , xy
2zz xx yy , xx yy ,
xy , yz , zx
x, y, z
The groups C1v and D1h

C 1v
D1h
C
C
g

g
g
g
g
C
u

u
u
u
u
z
(x, y)
xx C yy , zz
Rz
Rx , Ry
yz , zx
xx yy , xy
xy
yz , zx
xx C yy , zz
Rz
Rx , Ry
z
(x, y)
yz , zx
xx yy , xy
xy
yz , xz
153
6
Rotational and
VibrationRotation Raman
Scattering
I have ever been of the opinion that
revolutions are not to be evaded.
Benjamin Disraeli
6.1 INTRODUCTION
In this chapter we consider normal (non-resonance) rotational and vibrationrotation
Raman scattering with all transitions confined to the ground electronic state which is
taken to be non-degenerate. Under such conditions the cartesian-based component of
the appropriate vibrationrotation transition polarizability is given by setting ef D eg in
eqs. (4.5.10) and (4.5.11). This yields
eg vf Rf :eg vi Ri D hRf jhvf jheg jO er vr jeg ijvi ijRi i
6.1.1
where

1
C
O e v D
h er 6Deg er eg C vr vi 1 ier vr
er eg C vr vf C 1 C ier vr
r r
vr 6Dvt ,vi
6.1.2
We recall that these equations involve the following assumptions: the BornOppenheimer
approximation applies; and the frequency conditions necessary to effect closure over the
state Rr are satisfied.
154
The Raman Effect
As indicated in general terms in Chapter 4, the right-hand side of eq. (6.1.1) can be
separated into the product of two terms, one purely rotational and one purely vibrational.
The assumption of complete separation of rotational and vibrational states which underlies
this might appear to be very idealized because rotationvibration interactions are almost
invariably significant in the interpretation of rotational and rotationvibration Raman
spectra. It will transpire however that, although the treatment of energies and hence
transition frequencies requires consideration of vibration interactions, valid selection rules
and intensity expressions for many systems can be obtained from an analysis based on the
complete separation of the rotational and vibrational wave functions. Thus encouraged,
we now proceed to a detailed consideration of selection rules and intensities on this basis.
The conditions under which the ensuing idealized treatment can be applied more widely
are considered subsequently.
6.2 IRREDUCIBLE TRANSITION POLARIZABILITY

COMPONENTS
Before proceeding we recall that tensors in general are treated in Chapter A10 and the
polarizability tensor is considered in detail in Chapter A14. The concepts and formulae
developed there will be used here without detailed comment.
Equations (6.1.1) and (6.1.2) involve cartesian components of the transition polarizability tensors. Although the cartesian axis system is appropriate when dealing with an
experimental situation in which the incident radiation and scattered intensities have to be
related to laboratory-fixed axes, it is not well-suited to the problem we now have in hand.
It is much more convenient to work with the irreducible transition polarizability tensor
j
components m which have especially simple transformation properties under rotation.
The experimental situation can be recovered subsequently.
j
j
For this reason we now introduce m eg vf Rf :eg vi Ri and O m er , vr in place of
eg vf Rf :eg vi Ri and O er , vr , respectively. By analogy with eq. (6.1.1) we have
f
f
g j r r
j
O m e , v jeg ijvi ijRi i
m eg vf Rf :eg vi Ri D hR jhv jhe j
6.2.1
j
where O m er , vr is the equivalent, in the space-fixed irreducible system, of O er , vr .

The irreducible tensor and operator components are related to the cartesian components
by the transformation given by eq. (A14.5.4), but we do not need to consider these
inter-relationships at this point.
It is useful to express eq. (6.2.1) in an alternative form, namely
f
f j i
i
j
m vf Rf :vi Ri D hR jhv jm jv ijR i
6.2.2
g j r r
O m e , v jeg i
j
m D he j
6.2.3
where
Chapter A14 treats irreducible tensor components in detail.
155
In eq. (6.2.2), in order to simplify the nomenclature, we have omitted the electronic state
j
labels eg , a practice we shall continue in this chapter. The m may be regarded as
components of the polarizability in the vibrationless and rotationless ground electronic
j
j
state. As the operator O m er , vr is involved in eq. (6.2.3) this definition of m is a rather
general one and no index symmetry for and is assumed. The only simplification
involved is closure over the states Rr so that rotational resonances are excluded.
If further approximations are made to enable closure over the states vr to be effected,
j
then the operator will involve only the states er and may be represented as O m er .
j
The corresponding m then has the properties of a Placzek polarizability. Resonance
situations are excluded and and have index symmetry.
j
We shall consider m to have the general definition with no index symmetry for
and unless otherwise stated.
To explore further the rotational transitions we must use an expanded notation for the
rotational wavefunctions in which we replace the general label R, used hitherto, by JKM,
which represents a rotational wavefunction associated with the three quantum numbers J,
K and M, characteristic of a symmetric top. Equation (6.2.2) then becomes
f f f
f j i
j
O m jv ijJi Ki Mi i
m vf Jf Kf Mf :vi Ji Ki Mi D hJ K M jhv j
6.2.4
j
with O m still defined by eq. (6.2.3).

j
We next relate the space-fixed m to the molecule-fixed irreducible tensor components
j
m0 using the transformation
j j
m 0 Dm 0 m
6.2.5
j
m D
m0
j
j
where Dm0 m is a shorthand notation for Dm0 m ,

, an element of the angular momentum
transformation matrices for a rotation of the molecule-fixed axis system relative to the
space-fixed axis system, this rotation being defined by the Euler angles ,
and . The
j
transformation properties of the Dm0 m under rotations are treated in detail in Chapter A14,
j
Section A14.6. However, it is worth recalling here the special virtue of the Dm0 m , namely
that they mix only irreducible components with the same j value.
j
The m0 are independent of the rotational coordinates and so we can effect a separation
of the rotation and vibrational parts of eq. (6.2.4) when we introduce eq. (6.2.5). Thus
we obtain
j
m vf Jf Kf Mf :vi Ji Ki Mi
0 Dj
m
j
j
hJf Kf Mf jDm0 m jJi Ki Mi ihvf jm0 jvi i
6.2.6
m0 Dj
We see that eq. (6.2.6) consists of a sum of 2j C 1 terms, each of which is characterized
by a different value of m0 . Each term in the sum is the product of a rotational matrix
Chapter A14 treats the index symmetry of (Section A14.2) and (Section A14.3).
J, K and M are defined in detail in Section 6.7.2, page 187 of this Chapter.
Chapter A5 treats the Euler angles and the rotation operator.
156
The Raman Effect
element and a vibrational matrix element. The vibrational matrix element may be evaluated
independently of the rotational matrix element so we may defer detailed consideration
of this.
j
We note that the choice of a particular vibrational matrix element m0 defines m0 ,
but m0 plays a role in both the rotational and vibrational matrix elements. Also for
j
m vf Jf Kf Mf :vi Ji Ki Mi to be non-zero both the rotational and vibrational matrix elements
must be non-zero.
The rotational matrix element can be evaluated exactly for a number of specific molecular systems. In the next section we shall consider in detail the symmetric top molecule.
Linear molecules and spherical top molecules can be treated as special cases of the
symmetric top.
6.3 SYMMETRIC TOP

6.3.1 Selection rules
For a symmetric top molecule the value of a rotational matrix element using Wigner 3-j
symbols is as follows:
j
hJf Kf Mf jDm0 m jJi Ki Mi i D pf[2Ji C 12Jf C 1]1/2

f
f
J
j Ji
J
Mf m Mi
Kf
j
m0
Ji
Ki
6.3.1
The symbol (pf ) is used to denote the cumbersome phase factors which arise in eq. (6.3.1).
When, shortly, we form quadratic products of the type
2
j
j
jj
m fi j D m fi m fi
to obtain Placzek invariants and intensity formulae, the quadratic products of the phase
factors which we may represent symbolically as jpfj2 D pfpf are unity. There is
therefore no need to give them explicitly for these discussions.
Introducing eq. (6.3.1) into eq. (6.2.6) we have
f

J
j Ji
j
i
f
1/2
m vf Jf Kf Mf :vi Ji Ki Mi D pf[2J C 12J C 1]
Mf m Mi
0 Dj

m

Jf
j Ji
f j i
6.3.2
f m0 Ki hv jm0 jv i
K
0
m Dj
The selection rules for the rotational quantum numbers J, K and M follow straightforwardly from the properties of the two sets of Wigner 3-j symbols in eq. (6.3.2). These are
Kf D Ki C m 0
f
M DM Cm
Chapter A19 treats Wigner 3-j symbols in detail.
6.3.3
6.3.4

and
157
jJi jj Jf Ji C j
6.3.5
Unless these conditions are satisfied, the rotational matrix element given by eq. (6.3.1)
is zero. As neither jm0 j or jmj can exceed j, we can combine these restrictions with
eqs. (6.3.3) to (6.3.5) and write
jKj D jKf Ki j j
jMj D jMf Mi j j
6.3.6
jJj D jJf Ji j j
6.3.8
Ji C Jf j
6.3.9
and
6.3.2 Placzek invariants
j
6.3.7
fi
In order to treat intensities of rotation and vibrationrotation Raman scattering we must

evaluate the Placzek invariants G j fi which, for a transition from an initial state i to
a final state f are defined in terms of irreducible transition tensor components referred to
space-fixed axes by

2
jj
6.3.10
G j fi D
m fi j
m
with j D 0, 1, 2, and then obtain the Placzek invariant G fi which is given by

G j fi
6.3.11
G fi D
j
These Placzek invariants can then be used to obtain formulae for observed intensities.
It follows straightforwardly from eqs. (6.3.2) and (6.3.10) that
2
mDj
Jf
j Ji
j
i
f
G vf Jf Kf Mf :vi Ji Ki Mi D 2J C 12J C 1
Mf m Mi
mDj
0 Dj
m
m0 Dj
Jf
Kf
j
m0
Ji
K
2
j
jhvf jm0 jvi ij2
6.3.12
with the quantum number restrictions of eqs. (6.3.3) to (6.3.9) operative. In forming the
quadratic terms required by eq. (6.3.10) to obtain eq. (6.3.12) all cross-terms are zero
because of the orthogonality properties of the 3-j symbols, and jpfj2 D 1.
Now for a given j eq. (6.3.12) relates to a specified transition from vi Ji Ki Mi to
f
v Jf Kf Mf . This means that M D Mf Mi is specified (subject to the restriction
jMj j) and hence m; and also that K D Kf Ki has been specified (subject to the
Chapter A14 treats the Placzek invariants in detail.
158
The Raman Effect

j
restriction jKj j) and hence m0 and with it, m0 , the component of the molecule-fixed
irreducible transition tensor. Thus each of the sums over m and m0 contains only one
term, so that for a specified transition eq. (6.3.12) reduces to
G
j
vf Jf Kf Mf :vi Ji Ki Mi
Jf
j
Ji
D 2J C 12J C 1
Mf M Mi
f

2
J
j
Ji
j
jhvf jK jvi ij2

Kf K Ki
i
2
6.3.13
where we have replaced m by M and m0 by K to emphasize that m and m0 are

determined when the transition is specified. An expression of this form exists for each
J permitted by the operation of the triangular rule relating Jf , j and Ji , and for each
K and M allowed by the operation of the rules jKj j and jMj j, provided
j
always that the vibrational matrix element hvf jK jvi i is non-zero. These considerations
assume that the M levels are not degenerate. They apply to each of the G j fi .
Using eq. (6.3.13) and the quantum number restrictions of eqs. (6.3.3) to (6.3.9) we now
j
j
consider those m0 D K which contribute to each G j fi and the rotational transitions
with which they are associated. We recall that j can have the values 0, 1 and 2 corresponding
to isotropic, antisymmetric and anisotropic invariants. For completeness we shall consider
1 will be
all three invariants G 0 fi , G 1 fi and G 2
fi
fi . The antisymmetric invariant G
j
zero in the Placzek approximation for m0 .
We consider first the vibrational matrix element in eq. (6.3.13) which involves 0
K .
Since j D 0, Jf D Ji ; since jKj j, Kf D Ki ; and since jMj j, Mf D Mi . Thus
vibrational matrix elements with 0
K involved in the isotropic invariant are associated
only with transitions of the type J D 0, K D 0 and M D 0.
We next consider the three vibrational matrix elements which involve j D 1. Since
j D 1, we can have Jf D Ji C 1, Jf D Ji and Jf D Ji 1. Also since jKj j we can
have K D 1 and K D 0. Similarly, since jMj j we can have M D 1 and
M D 0. Thus vibrational matrix elements with 1
K , involved in the antisymmetric
invariant, are associated with transitions of the type J D 1 and 0, with both K and
M taking the values 1 or 0 for each of the three J transitions.
Finally we consider the vibrational matrix element which involves j D 2. Since
j D 2, we can have Jf D Ji 2, Jf D Ji 1 and Jf D Ji . Also, since jKj j, we
can have K D 2, K D 1 and K D 0. Similarly, since jMj j, we can have
2
, involved
M D 2, M D 1 and M D 0. Thus vibrational matrix elements with K
in the anisotropic invariant, are associated with transitions of the type J D 2, 1 or 0,
with both K and M taking the values 2, 1 or 0 for each J transition. Table 6.1
j
summarizes the m associated with the permitted J and K values.
We now consider the evaluation of eq. (6.3.13) which we write in the form
G
j
j
j
vf Jf Kf Mf :vi Ji Ki Mi D 2Ji C 1bJf Kf Mf :Ji Mi Ki jhvf jK jvi ij2
6.3.14
159
Table 6.1 The ranks j (j D 0, 1 or 2) of the irreducible tensors j

m which contribute to the branches
defined by the permitted J and K values.
J
2
1
2
1
0
1
2
2
2
2
2
2
2
2, 1
2, 1
2, 1
2
2
2, 1
2, 1, 0
2, 1
2
2
2, 1
2, 1
2, 1
2
2
2
2
2
2
Jf
Kf
j
K
where
j
bJf Kf Mf :Ji Ki Mi
D 2J C 1
Jf
Mf
j
Ji
M Mi
2
Ji
Ki
2
6.3.15
The two Wigner 3-j symbols in eq. (6.3.15) can be evaluated for the permitted
K, M and J values for each value of j. We do not present the results here because
they are rarely needed, as we now explain.
We have stated above that although eq. (6.3.12) contains two sums, one over m and
one over m0 , when K and M are specified (subject to jKj j and jMj j) only
one term remains in each sum. However this assumes that the M levels are not degenerate
which is only true in the presence of, for example, an appropriate electric or magnetic field.
Normally, the frequency associated with the transition from vi Ji Ki Mi to vf Jf Kf Mf is
independent of the values of Mi and Mf as the M levels are degenerate. In this situation it
is necessary to sum over all values of Mi and Mf in order to obtain all the contributions to
G j vf Jf Ki :vi Ji Ki at the frequency associated with the transition from vi Ji Ki to vf Jf Kf .
This summation follows readily from the following property of Wigner 3-j symbols:
2
Jf
1
j
Ji
D f
6.3.16
f M Mi
M
2J C 1
i
M
for a fixed value of Mf . This result is independent of Mi , and as there are 2Jf C 1 values
of Mf we have the simple result that
2
Jf
j
Ji
D1
6.3.17
Mf M Mi
i
f
M
Thus when the M levels are degenerate, as is normally the case, we can replace the
sum of the G j vf Jf Kf Mf :vi Ji Ki Mi over all Mi and Mf by G j vf Jf Kf :vi Ji Ki where
G
j
j
j
vf Jf Kf :vi Ji Ki D 2Ji C 1bJf Kf :Ji Ki jhvf jK jvi ij2
6.3.18
160
The Raman Effect
and

j
bJf Kf :Ji Ki
D 2J C 1
Jf
Kf
j
Ji
K Ki
2
6.3.19
Equations (6.3.18) and (6.3.19) are of fundamental importance. Together with the selection rules for J and K derived earlier they determine the properties of rotation and
vibrationrotation Raman scattering.
j
The factor bJf Kf :Ji Ki was first introduced by Placzek and Teller (1933) in their classic
work on rotational and vibrationrotation Raman scattering and it has been used subsequently in a number of key papers and reviews. More recent papers and reviews have
tended to use the ClebschGordan coefficients. Their relationship to the Placzek-Teller b
factors is straightforward and is as follows
j
bJf Kf :Ji Ki D fCJi j Jf ; Ki K Kf g2 D fhJi Ki , jK j Jf Kf ig2
6.3.20
The nomenclature for the Clebsch-Gordan coefficient is defined by eq. (A19.2.19) in

Chapter A19.
We now proceed to evaluate eq. (6.3.19) and hence eq. (6.3.18) for j D 0, 1 and 2 When
j D 0, K D 0 (that is Kf D Ki ) and also J D 0 there is just one PlaczekTeller factor.
Using eq. (A19.3.11) this is defined as follows
bJ0
i Ki :Ji Ki D 1
and thus
G
0
i 2
vf Ji Ki :vi Ji Ki D 2Ji C 1jhvf j0
0 jv ij
6.3.21
6.3.22
Using eqs. (A14.7.6) and (A14.7.22) in Chapter A14 we have

2
2
j0
0 j D 3a
6.3.23
so that finally we arrive at the simple result

G
0
vf Ji Ki :vi Ji Ki D 32Ji C 1jhvf jajvi ij2
6.3.24
When j D 2 the situation is somewhat less simple. Equation (6.3.18) then yields
G
2
f 2 i 2
vf Jf Kf :vi Ji Ki D 2Ji C 1bJ2
f Kf :Ji Ki jhv jK jv ij
6.3.25
Since the permitted J and K values, are J D 0, 1, 2 and K D 0, 1, 2 there

are now 25 PlaczekTeller factors. These are given in Table 6.2 (pages 162 and 163).
These factors are not all independent. They obey the relationships
and
2
f
2Ji C 1bJ2
i Ki :Jf Kf D 2J C 1bJf Kf :Ji Ki
6.3.26
2
bJ2
f Kf :Ji Ki D bJf ,Kf :Ji ,Ki
6.3.27
In fact, Placzek and Teller used the symbol bJf Kf :Ji Ki for bJ2f Kf :Ji Ki . They did not need a superscript label
because they did not consider bJ1f Kf :Ji Ki ; and bJ0f Kf :Ji Ki D 1. See eq. (6.3.21).

Also, using the summation properties of the 3-j coefficients, it follows that
2
Jf
1
2
Ji
D
f
i
K
K K
5
i
f
161
6.3.28
for a given K. This leads to the useful result that

2
2Jf C 1
bJf Kf :Ji Ki D
5
K
6.3.29
It will be useful to set out the forms of eq. (6.3.25) for the five permitted K values.
f 2 i 2
vf Jf Ki :vi Ji Ki D 2Ji C 1bJ2
f Ki :Ji Ki jhv j0 jv ij
6.3.30
2
f 2 i 2
vf Jf Ki 1:vi Ji Ki D 2Ji C 1bJ2
f Ki 1:Ji Ki jhv j1 jv ij
6.3.31
2
f 2 i 2
1bJ2
6.3.32
G
G
G
2
vf Jf Ki 2:vi Ji Ki D 2Ji C
For completeness we now consider the case when j D 1, which only arises when the
transition tensor is not symmetric. This situation which occurs in resonance rotational and
vibrationrotation Raman scattering is treated in detail in Chapter 8. Equation (6.3.18)
now becomes
G
1
f 1 i 2
vf Jf Kf :vi Ji Ki D 2Ji C 1bJ1
f Kf :Ji Ki jhv jK jv ij
6.3.33
Here there are nine PlaczekTeller factors for the permitted J and K values, namely
J D 0, 1 and K D 0, 1. These are given in Table 6.3 (page 164).
These factors are not all independent. They obey the following relationships
1
f
2Ji C 1bJ1
i Ki :Jf Kf D 2J C 1bJf Kf :Ji Ki
bJ1
f Kf :Ji Ki D bJf ,Kf :Ji ,Ki
bJ1
f Kf :Ji Ki D
2Jf
C 1
3
6.3.34
6.3.35
6.3.36
The forms of eq. (6.3.33) for the three permitted K values are
f 1 i 2
vf Jf Ki :vi Ji Ki D 2Ji C 1bJ1
f Kf :Ji Ki jhv j0 jv ij
G
1
G
f 1 i 2
vf Jf Ki 1:vi Ji Ki D 2J C 1bJ1
arising from the anti-symmetric invariant G 1
fi will not
1
6.3.37
6.3.38
Raman scattering
be considered
further in this chapter.
j
It should be noted carefully that all the definitions of G vf Jf Kf :Ji Ki vi which we have just
developed include the factor 2Ji C 1 which arises from the degeneracy of the 2Ji C 1
levels M for a given Ji . When dealing with intensities it is usual to include all the
degeneracies of the initial state in one factor. It is then necessary to use
1
G
C 1
2Ji
j
vf Jf Kf :vi Ji Ki
to avoid counting twice the rotational degeneracy factor 2Ji C 1.
2
1
J C 2K C 12 J K 1J K
2JJ 1J C 12J C 1
J 2K C 12 J C K 1J C K
2JJ 1J C 12J C 1
[J2 K C 12 ]J K 2J K
2JJ 1J C 12J C 1
[J2 K 12 ]J C K 2J C K
2JJ 1J C 12J C 1
(Continued opposite)
[J 12 K 12 ]J C K 1J C K

JJ 12J 12J C 1
J K 3J K 2J K 1J K
4JJ 12J 12J C 1
J C K 3J C K 2J C K 1J C K
4JJ 12J 12J C 1
K1
KC2
K2
1
2
[J 12 K C 12 ][J K 1J K]

JJ 12J 12J C 1
KC1
J1
3K2 [J2 K2 ]
JJ 1J C 12J C 1
J2
3[J 12 K2 ][J2 K2 ]

2JJ 12J 12J C 1
K Kf
Jf
Table 6.2 The PlaczekTeller factors bJ2

f Kf :Ji Ki for the symmetric top rotor.
162
The Raman Effect
J 2K2 J C K C 1J C K C 2
2JJ C 1J C 22J C 1
J C 2K2 J K C 1J K C 2
2JJ C 1J C 22J C 1
[J C 12 K C 12 ]J C K C 1J C K C 3
2JJ C 1J C 22J C 1
[J C 12 K 12 ]J K C 1J K C 3
2JJ C 1J C 22J C 1
32K 12 J C KJ K C 1

2JJ C 12J 12J C 3
3[J2 K C 12 ][J C 12 K C 12 ]

2JJ C 12J 12J C 3
3[J2 K 12 ][J C 12 K 12 ]

2JJ C 12J 12J C 3
K1
KC2
K2
1
C2
2
J K C 1J K C 2J K C 3J K C 4

4J C 1J C 22J C 12J C 3
J C K C 1J C K C 2J C K C 3J C K C 4

4J C 1J C 22J C 12J C 3
[J C 12 K2 ]J K C 2J K C 3

J C 1J C 22J C 12J C 3
[J C 12 K2 ]J C K C 2J C K C 3

J C 1J C 22J C 12J C 3
3[J C 12 K2 ][J C 22 K2 ]

2J C 1J C 22J C 12J C 3
JC2
(2Ji C 1bJ2i Ki :Jf Kf D 2Jf C 1bJ2f Kf :Ji Ki . bJ2f Kf :Ji Ki D bJ2f Kf :Ji Ki
For compactness, in this Table the initial states are labelled J and K (i.e. Ji D J; Ki D K)
Thus the final J states are J, J 1, J 2 (i.e. Jf D Ji J D 0; Jf D Ji 1J D 1; Jf D Ji 2J D 2).
and the final K states are K, K 1, K 2 (i.e. Kf D Ki K D 0; Kf D Ki 1K D 1; Kf D Ki 2K D 2).
32K C 12 J KJ C K C 1

2JJ C 12J 12J C 3
KC1
JC1
3K2 [J C 12 K2 ]

JJ C 1J C 22J C 1
[JJ C 1 3K2 ]2
JJ C 12J 12J C 3
K Kf
Jf
Table 6.2 (Continued )

163
164
The Raman Effect
Table 6.3 The PlaczekTeller factors bJ1

f Kf :Ji Ki for the symmetric top.
J
f
1
J1
JC1
K2
JJ C 1
J KJ C K C 1
2JJ C 1
J K C 1J C K
2J2J C 1
J C K C 1J K C 1
J C 12J C 1
J C K C 1J C K C 2
2J C 12J C 2
J K C 1J K C 2
2J C 12J C 2
Kf
0
1
1
J KJ C K
J2J C 1
J K 1J K
KC1
2J2J C 1
J C KJ C K 1
K1
2J2J C 1
K
For compactness, in this Table the initial states are labelled J and K [i.e. Ji D J; Ki D K]. Thus the
final J states are J, J 1, [i.e. Jf D Ji J D 0; Jf D Ji 1J D 1], and the final K states are K,
K 1, [i.e. Kf D Ki K D 0; Kf D Ki 1K D 1]
The results we have obtained for the Placzek invariants G j fi are formulated in terms
j
j
of the irreducible tensor components of m D K . They can also be expressed in terms of
the corresponding spherical tensor components or cartesian tensor components
j
by using the relations between the m and the and which are given in Table 6.4.
These have been obtained by using eqs. (A14.5.2) and (A14.5.4) and Tables A14.2 and
A14.3 in Chapter A14.
We now examine the implications of eqs. (6.3.24) and (6.3.25) for Raman activity.
Throughout the following discussion we shall take the initial and final electronic states to
be the ground state, which is non-degenerate, and vi D 0. We see from eq. (6.3.24) that
isotropic scattering which is controlled by G 0 fi can only occur if J D 0 and K D 0
and hvf jajvi i is non-zero. When vf D vi D 0, which corresponds to Rayleigh scattering,
hvf jajvi i reduces to a0 , the equilibrium mean polarizability which is always non-zero
irrespective of the molecular symmetry. Thus Rayleigh scattering always occurs. When
vf 6D vi , G 0 fi contributes to Raman scattering associated with vibrationrotation
transitions for which v 6D 0, J D 0 and K D 0 provided that hvf jajvi i 6D 0. When
vi D 0 and vf D 1v C 1 the condition on the vibrational matrix element becomes
hvf D 1jajvi D 0i 6D 0. As a is totally symmetric this is satisfied for totally symmetric
vibrations and so transitions with J D 0 and K D 0 are always observed for such
vibrations.
Turning now to eq. (6.3.25) we see in contrast that anisotropic scattering which is
controlled by G 2 fi can contribute to all allowed transitions, J D 0, 1, 2 and
2 i
K D 0, 1, 2, provided the associated transition polarizability hvf jK
jv i is nonzero. In general this depends on both the equilibrium symmetry of the molecule and
the symmetries of its vibrations. In the special case of pure rotational transitions, for
which vf D vi D 0, it is the equilibrium symmetry of the molecule alone that determines
A fuller discussion of the symmetry properties of vibrational transitions is given in Chapter 5, Section 5.8.
2
1
2
2
1
2
02
2
22
1
1
11
1
p f200 11 C 11 g
6
1
p 01 10
2
1
p 10 01
2
11
11 D 11
0
p
2
p
3
0
210 D 201
p
2
p 00 11
3
p
p
201 D 210
0
11 D 11
Diagonalb
tensor
p
3a
1
p 211 C 00
3
p
D 3a
1
p 11 C 11 C 00
3
p
D 3a
1
p 10 C 01
2
1
p 11 11
2
1
p 01 C 10
2
Symmetric tensora

General tensor
1
p xx C yy C zz
3
p
D 3a
General tensor
6D
1
fzx xz C izy yz g
2
i
p xy yx
2
1
fzx xz C iyz zy g
2
1
fxx yy C ixy C yx g
2
1
fxz C zx C iyz C zy g
2
1
p f2zz xx C yy g
6
1
fxz C zx izy C yz g
2
1
fxx yy ixy C yx g
2
11 D 11 ; 10 D 01 ; 10 D 01 ; 11 D 11

For cylindrical symmetry about z axis additionally: 01 D 10 D 0; 10 D 01 D 0
11 D 11 D 0; D 00 11
c
xy D yx ; yz D zy ; xz D zx
d
For cylindrical symmetry about z axis additionally: xx D yy ; xy D yx D 0
zy D yz D 0; xz D zx D 0; D zz xx
2
1
0
1
1
1
0
11
0
j
0
0
zx izy
1
fxx yy 2ixy g
2
1
p f2zz xx C yy g
6
0
p
2
p
3
1
fxx yy C 2ixy g
2
zx C izy
Diagonald
tensor
p
3a
Symmetric tensorc
D
p
3a
Table 6.4 The relationships of the j

m to the and for a general tensor, a symmetric tensor and a diagonal tensor.

165
166
The Raman Effect
which polarizability components are non-zero. However if vi D 0 and vf D 1 then we

2
i
require hvf D 1j2
K jv D 0i 6D 0. This is satisfied if K and the normal coordinate of
the vibrational mode in question belong to the same symmetry class.
From Table 6.4 which relates 2
K to we see that, in the case of a symmetric
tensor, K D 0 transitions involve xx C yy and zz , that K D 1 transitions involve
xz and yz , and that K D 2 transitions involve xx yy and xy . The following
examples illustrate these selection rules. Reference Table 5.3 shows that for a symmetric
top molecule with D3h symmetry, vibrations of a0 , symmetry will have K D 0, vibrations
of e00 symmetry will have K D 1 and vibrations of e0 symmetry will have K D 2.
For a molecule with D3d symmetry vibrations of ag symmetry will have K D 0, and
vibrations of eg symmetry will have K D 1 and K D 2. For a molecule with C4h
symmetry, vibrations of ag symmetry will have K D 0, vibrations of bg symmetry will
have K D 2, and vibrations of eg symmetry will have K D 1.
The tensor components involved in a particular K transition can be related to the
anisotropies of the Raman tensor for the relevant symmetry class. Specific examples are
given later.
Generalizations of the J and K rules can be made and these are presented in
Table 6.5 for the symmetric top and also for linear molecules and the spherical top which
will be considered later.
We have so far concentrated on the symmetry properties of vibrational transition tensor
components which involve the symmetry properties of the vibrational wave functions.
We now look briefly at properties of rotational wave functions. In general rotational
wave functions can be classified as positive C or negative depending on whether
the total wave function is unaltered or changes its sign by reflection of all particles,
electrons and nuclei at the origin. For molecules with a centre of symmetry, there is
an additional symmetry property, namely, that for an exchange of the nuclei the total
eigenfunction either remains unchanged or changes only in sign; and in consequence
the state in question is said to be symmetric (s) or antisymmetric (a) in the nuclei. It
can be shown that in a given electronic state either the positive rotational levels are
symmetric and the negative levels are antisymmetric throughout, or, the positive levels
are antisymmetric and the negative levels symmetric throughout. Strict symmetry-based
selection rules govern transitions between rotational states which give rise to Raman
scattering. They are
C ! C;
! ;
C ;
a ! a;
C
and
! s;
a;
a s
These rules can be shown to be completely compatible with the selection rules for J
and K already obtained. In some instances, however, parity considerations can play an
additional role, as for example in the vibrationrotation transitions associated with the
degenerate vibrations of linear molecules.
167
Table 6.5 Rotation and vibrationrotation Raman spectra: selection rules.

Vibrational
speciesb
Linear
totally symmetric
degenerate
xx C yy ; zz
yz , zx
polarized
depolarized
J D 0, 2
J D 0, 1, 2
always C $ C,
$ ,
C$
/
J0 C J00 0)
Symmetric top totally symmetric
xx C yy ; zz
polarized
xx yy ; xy
depolarized
xx yy ; xy
depolarized
yz , zx
depolarized
J D 0, 1, 2 :
K D 0
J D 0, 1, 2 :
K D 2
J D 0, 1, 2 :
K D 2
J D 0, 1, 2 :
K D 1
J0 C J00 2)c
totally symmetric
xx C yy C zz
J D 0c
doubly degenerate
xx C yy 2zz ,
xx yy
xy , yz , xz
completely
polarized
depolarized
depolarized
J D 0, 1, 2 :
J0 C J00 2)c
non-totally symmetric
non-degenerated
degenerate
Spherical top
triply degenerate
Non-zero component State of

of polarizability
polarization
Selection rulesa
Rotor type
J D 0, 1, 2
a 0
J refers to upper (higher energy) states and J00 to lower (lower energy) states (see section 6.4)
Pure rotation transitions have the same selection rules as totally symmetric vibration rotation transitions.
c
For pure rotation transitions J0 C J00 0.
d
Only in molecules with fourfold axes.
b
6.3.3 Intensities
By analogy with the treatment developed in Chapter 5 and Reference Tables 5.2(a) to
5.2(g) for the intensity of vibrational Raman scattering, formulae for the intensity of
rotational or vibrationrotation Raman scattering by non-chiral molecules will have the
following general form, provided the transition polarizability tensor is symmetric:
I
; ps , pi D kQ Qs 4 gi Ni a2 , 2 ,
I
I
; ps , pi
6.3.39
The symbol
is defined in Chapter 5, Section 5.4.1, kQ is defined by eq. (5.5.11),
Q s is the wavenumber of the scattered radiation, gi is the overall statistical weight of the
initial state, Ni is a population factor whose explicit definition depends on the transition involved and I is the irradiance of the incident radiation. The function a2 , 2 ,

involves the combinations of the transition tensor invariants a2 and 2 which define the
isotropic averages of the transition tensor components appropriate for the incident radiation polarization pi and the scattered radiation polarization ps and any geometric factor
involving the angle
between the direction of the incident radiation and the direction of
observation.
168
The Raman Effect
Only two of the isotropic averages of transition polarizability components, appropriately

chosen, are required to determine the intensity and polarization characteristics of the
scattered radiation for all combinations of pi and ps . For linearly polarized incident
and scattered radiation two cartesian isotropic averages, of the types hxx 2 i and hxy 2 i
suffice; and for circularly polarized incident and scattered radiation, two spherical isotropic
averages of the types h11 i2 and h11 2 i are sufficient. Such isotropic averages are
related to the Placzek invariants as follows

1
2
1 0
2
2
G fi C G fi
hxx i D
6.3.40
2Ji C 1 3
15

1
1
hxy 2 i D
G 2 fi
6.3.41
i
2J C 1 10

1
1
1 0
h11 2 i D
G

C
6.3.42

G 2 fi
fi
2Ji C 1 3
30

1
1 2
2
h11 i D
G fi
6.3.43
2Ji C 1 5
In the above equations we have removed the M degeneracy factor 2Ji C 1 implicit in
G j fi as it will be included in the overall statistical weight gi in intensity formulae.
The Placzek invariants and the population factor depend on the symmetry of the
molecule being considered. Also, both the Placzek invariants and the population factor
depend on the quantum numbers characterising the initial state but in different ways.
Thus, formulae for I
; ps , pi cannot be generalized and must be developed for each
type of molecule. This is undertaken in the sections that follow.
However, it is possible to make generalizations about the depolarization ratios, and
reversal coefficients of the scattered radiation, because these quantities involve ratios of
scattered intensities and thus all population factors cancel. We deal with these quantities next.
We consider first the depolarization ratios /2; ?i ) which can be expressed in terms
of the Placzek invariants as follows
3G 2 fi
/2; ?i D
6.3.44
10G 0 fi C 4G 2 fi
For J D 1, 2 transitions which give rise to pure rotational Raman spectra and
rotational lines in vibrationrotation Raman spectra,
/2; ?i D
3
4
6.3.45
since this scattering is entirely anisotropic in nature and has G 0 fi D 0. For J D 0
transitions, both G 0 fi and G 2 fi can be independently greater than, or equal to,
zero; and the relative contributions of G 0 fi and G 1 fi will depend on the symmetry
of the transition tensor. For Raman scattering with vf 6D vi and J D 0, G 2 fi D 0 for
See Chapter 5 pages 99 and 107.

Chapter A14, Tables A14.5 and A14.7.
See this Chapter page 161.
169
spherically symmetric vibrations and /2; ?i D 0; for totally symmetric vibrations

G 0 fi > 0 and G 2 fi > 0 and so 0 < /2; ?i < 3/4. For all other vibrational
symmetries G 0 fi D 0 and G 2 fi 6D 0 and /2; ?i D 3/4. For Rayleigh scattering
with vf D vi and J D 0, the relative contributions of G 0 fi and G 2 fi will depend
on the symmetry of the equilibrium polarizability tensor. In this case G 0 fi can
never be zero so there is always an isotropic contribution. However for spherically
symmetric molecules G 2 fi D 0 and so /2; ?i D 0; for other molecular symmetries
G 2 fi 6D 0 and hence 0 < /2; ?i < 3/4.
Similar considerations apply to the reversal coefficient which is defined as follows
P 0 D
6G 2 fi

10G 0 fi C G
2
D P 1
6.3.46
fi
For Raman scattering when G 0 fi D 0, P 0 D 6; when G 0 fi 6D 0 and G 2 fi D 0,
P 0 D 0; and when G 0 fi 6D 0 and G 2 fi 6D 0, 0 < P 0 < 6. Thus overall
0 P 0 6 for Raman scattering. For Rayleigh scattering since G 0 fi cannot be zero,
overall 0 P 0 < 6.
6.3.4 Subsequent development

In preparation for the detailed treatment of the rotational and vibrationrotation Raman
spectra of various types of molecules which begins with Section 6.6, we consider in
Section 6.4 molecular rotational and vibrational terms, and in Section 6.5 the statistical
distribution of molecular populations among the various quantum states of a molecule.
6.4 ROTATIONAL AND VIBRATIONAL TERMS

We recall that Q M , the transition wavenumber for a molecule is defined as follows:
Q M D TQ 0 TQ 00
6.4.1
when TQ 0 and TQ 00 are the total terms for the upper (higher energy) and lower (lower
energy) states, respectively; Q M , TQ 0 and TQ 00 are in wavenumbers, (unit: cm1 ) and the
tilde is included to emphasize this.
Qe , a vibrational term
A total term TQ can be written as the sum of an electronic term T
Q
Q
G and a rotational term F so that
Q CF
Q
TQ D TQe C G
6.4.2
This representation of a total term TQ as the sum of three terms is not to be understood as
Q and F
Q are completely independent of each other. Although
necessarily implying that TQe , G
the separation is a reasonable approximation in many cases, interactions between the
terms can be significant. For example the rotational term has a small dependence on
the vibrational quantum number v. When we need to include this dependence it will be
Q v instead of F
Q rather than by introducing a vibrationrotation
expressed by writing F
interaction term explicitly in eq. (6.4.2).
170
The Raman Effect
In this section our purpose is to obtain expressions for Q M for transitions between rotational levels and between vibrationrotation levels which are permitted by the selection
rules derived earlier in this chapter. Throughout we assume that the molecule is in the
ground electronic state and remains there, so that TQ 0e D TQ 00e D 0.
Q 00 , we have in general
Q0 D G
For pure rotational transitions, for which G
Q0 F
Q 00
6.4.3
Q M D F
We shall consider at this stage only rotational transitions where the molecule is in the
ground vibrational state throughout, that is v0 D v00 D 0.
For vibration-rotation transitions we have
Q0G
Q 00 C F
Q0 F
Q 00
Q M D G
6.4.4
when vibrationrotation interaction is neglected, and
Q0G
Q 00 C F
Q 0v F
Q 00v
Q M D G
6.4.5
when vibrationrotation interaction is taken into account.

In many cases the lower vibrational state is the ground state, that is v00 D 0 and the transition is a fundamental one for which v D C1, so that v0 D 1. However, we do not wish
to exclude other possibilities, as for example hot bands, overtones and combination bands.
Q0G
Q 00 ) is just a constant
For this reason, and as for a particular vibrational transition (G
term, we shall not normally specify the vibrational transition; we shall simply write
Q0G
Q 00 D Q vib
G
6.4.6
Specific assumptions relating to v00 , v, harmonicity and so on can then be introduced as
required.
Clearly Q M as defined here is always positive. Thus the Stokes Raman spectrum consists
of lines with wavenumbers of the type Q 1 Q M where Q 1 is the wavenumber of the exciting
radiation and the anti-Stokes Raman spectrum consists of lines of the type Q 1 C Q M .
In the case of vibrationrotation transitions the anti-Stokes Raman spectrum will have
low intensity and its measurement, if practicable, is not usually worthwhile. However
in the case of pure rotation transitions the Stokes and anti-Stokes Raman spectra have
comparable intensities and both can be measured.
We conclude this section by drawing attention to some special aspects of the selection
rules for the quantum number J. For a molecule in which no electronic or vibrational
transitions occur, J D 0 relates simply to Rayleigh scattering at Q 1 ; and J D C2 and
J D C1 relate to both Stokes and anti-Stokes Raman scattering. The selection rules
J D 2 and J D 1, which might at first sight be thought to relate to anti-Stokes
Raman scattering, do not operate for pure rotation Raman scattering. This is because J
which is defined as J0 J00 must be positive since J0 (associated with the higher energy
state) is always greater than J00 (associated with the lower energy state) if no electronic
or vibrational transition is also involved and so by definition J D 1 and J D 2
cannot arise.
In the sections that follow we consider the rotation and vibrationrotation Raman
spectra of the following types of molecules: diatomic molecule, symmetric top molecule,
171
linear molecule, spherical top molecule, and asymmetric top molecule. In these treatments
Q and F
Q v without proof but include brief explanatory notes.
we give the rotational terms F
Q have already been considered in Chapter 5.
The vibrational terms G
In presenting the results it is convenient to use jQj rather than Q M . We recall from
Chapter 3, Section 3.10 that jQj is the magnitude of the wavenumber shift from Q 1 the
wavenumber of the exciting radiation. For Stokes Raman scattering Q is positive and
for anti-Stokes Raman scattering it is negative. The characteristic molecular property Q M
is given by jQj; for Stokes Raman scattering Q M D Q.
Q or F
Q v depends upon the resolution
In practice the level of approximation required for F
available in an experiment. Nowadays a resolution as high as 0.001 cm1 or better can
be achieved and thus rotational terms at a correspondingly high level of approximation
can be needed. However, we begin the treatment for each type of molecule at the lowest
level of approximation for the rotational term and then introduce refinements stage by
stage. This is helpful for the less ambitious reader. Also the simple treatments reveal the
general pattern of the lines in the spectrum. This pattern generally undergoes only modest
qualitative changes as the theory is refined.
6.5 STATISTICAL DISTRIBUTION OF MOLECULAR

POPULATION
For a system of N molecules in thermal equilibrium, the number of molecules Ni in the
initial energy level i, characterized by its energy Ei and its degeneracy gi , is given by
Ngi expfEi /kTg
Ni D
6.5.1
Q
where Q is the molecular partition function. This is given by

gj expfEj /kTg
6.5.2
QD
j
where gj is the statistical weight of the energy level Ej and the summation is over all Ej .
The quotient Ni /N gives the fraction of the total number of molecules with energy Ei .
As already indicated in Section 6.4 it is normally a very good approximation to treat
Q and
Q as the sum of an electronic term TQe , a vibrational term G
the total energy term T
Q
a rotational term F and to regard these as completely independent of each other. On
this basis we can invoke eq. (6.4.2) and factorize eq. (6.5.2) so that the molecular partition function Q, reduces to a product of electronic, vibrational and rotational partition
functions. That is
Q D Qe Qv QR
6.5.3
where
Qe D
gej expfTQej hc0 /kTg
6.5.4
Q j hc0 /kTg
gvj expfG
6.5.5
Qv D

j
172
The Raman Effect
and
QR D
Q j hc0 /kTg
gRj expfF
6.5.6
The summations are over all electronic, vibrational or rotational energy levels; and gej , gvj
and gRj are degeneracy factors for the j th electronic, vibrational and rotational levels,
respectively. These degeneracies may be different for different levels.
As electronic energy separations are usually very large, the exponential terms in
eq. (6.5.4) are all very small except for the electronic ground term. In such cases if the
ground electronic state is non-degenerate, Qe D 1; otherwise Qe is equal to the degeneracy
of the ground electronic level. If excited electronic levels are close in energy to the ground
level then the summation in eq. (6.5.4) must be evaluated using experimentally determined
energy values.
For the case of the vibrational partition function it is usually satisfactory to assume
that the simple harmonic approximation applies. Then, for a diatomic molecule, with a
vibrational wavenumber Q , we have
Q v D v C 1/2Q
G
6.5.7
where v D 0, 1, 2, 3, . . . . If we measure energies from the zero-point energy level then,
as in this case gvj D 1,

Qv D
expfvhc0 Q /kTg
6.5.8
v
If Q is sufficiently large then all the exponential terms in the summation in eq. (6.5.8)
with v > 0 can be neglected and Qv D 1.
For polyatomic molecules with several vibrational modes the total vibrational partition
function is obtained by forming the product of sums of the form given by eq. (6.5.8) for
each vibrational wavenumber Q . If a vibrational level is degenerate this must be taken
into account by using gv D 2 or 3 with the appropriate summation. If the simple harmonic
approximation is not adequate then the summations must be performed explicitly using
data derived experimentally. If a vibrational wavenumber is sufficiently large then the
associated summation Qv is simply the vibrational degeneracy, as explained above.
The case of the rotational partition function is generally more complicated. For a
symmetric top molecule the overall degeneracy gR is given by
gR D gN gJK
6.5.9
where gN is the nuclear spin degeneracy and gJK is the rotational degeneracy of the
symmetric top. The nuclear spin degeneracy, which depends on a number of factors , can
be different for different levels; and in some molecules can be zero for certain levels so
that these levels are not populated. The rotational degeneracy is a function of the rotational
quantum number J and, where appropriate, K so that generalization is not possible.
The most straightforward case is that of a heteronuclear diatomic molecule treated as
a simple rigid rotor, for which

Q
QR D
2J C 1 expfFJhc
6.5.10
0 /kTg
J
See Section 6.7.4, page 203.
173
Here gN D 1 and the rotational degeneracy gR is given by gJ D 2J C 1. This arises
Q
because there are 2J C 1 levels M D 0, 1, 2 . . . J with the same energy FJhc
0,
in the absence of an external field. Other cases will be treated in following sections dealing
with specific types of molecule.
6.6 DIATOMIC MOLECULE

6.6.1 Introduction
Q
For a rigid molecule the rotational term FJ
in the ground vibrational state is given by
Q
FJ
D BJJ C 1
6.6.1
where
BD
h
82 c
0 Ib
6.6.2
and Ib is the moment of inertia of the molecule, about an axis through its centre of gravity
and perpendicular to the internuclear axis.
Equation (6.6.1) ignores any contribution which may arise from the angular momentum
of the electrons, either orbital or spin or both. When orbital angular momentum has to
be taken into account, the diatomic molecule becomes a special case of a symmetric top;
and when the resultant spin quantum number S is not zero a splitting of the rotational
levels arises. These two special cases which arise only in a limited number of molecules
are considered later in this chapter.
For a diatomic molecule whose rotational term is given by eq. (6.6.1) all the rotation is
end-over-end. There is no rotation about the internuclear axis, and the angular momentum
is necessarily perpendicular to the internuclear axis. We can adapt the theoretical treatment,
given in Section 6.3 for a symmetric top, to the case of a diatomic molecule by putting
K D 0 (and hence K D 0). The selection rule associated with G 0 fi is still J D 0
and G 0 fi is given by eq. (6.3.22) with Kf D Ki D 0 and K D 0. For G 2 fi the
selection rules are now J D 0, 2 as can be seen by putting K D 0 in the first row
of Table 6.2; and G 2 fi is given by eq. (6.3.25) with Kf D Ki D 0 and K D 0.
In eqs. (6.3.22) and (6.3.25), vf D vi for pure rotation Raman spectra; and vf 6D vi for
vibration-rotation Raman spectra.
Initially we shall restrict our considerations to diatomic molecules which are heteronuclear, because in such molecules the nuclear spin degeneracy factor gN is the same for
all rotational levels. In the case of homonuclear diatomic molecules gN takes different
values for odd and even J values; and the gN values depend on the nuclear spin of the
atoms involved. Because of these additional factors homonuclear diatomic molecules are
treated separately in Section 6.6.5.
Here and in subsequent sections we first use the selection rules and spectroscopic terms
to obtain formulae for the wavenumbers of the lines in a given branch. We then consider
the intensity distribution within a branch. It will be convenient to use J for Ji and K for
Ki in the rest of this chapter, unless this introduces ambiguity.
174
The Raman Effect
6.6.2 Heteronuclear diatomic molecule: pure rotation

The selection rules permit only one branch, an S branch for which J D C2. With this
selection rule it follows that
Q C 2 FJ
Q
SJ; jQj D FJ
6.6.3
and using eq. (6.6.1) this is readily shown to give
SJ; jQj D 4BJ C 3/2
6.6.4
where J D 0, 1, 2, . . . . It follows that S0; jQj D 6B, S1; jQj D 10B, S2; jQj D
14B, and so on. Thus in an S branch the first rotational line has a wavenumber shift
of 6B from the exciting line and successive rotational lines have equal interline spacings
of 4B. Figure 6.1(a) shows the rotational energy levels of a rigid heteronuclear molecule
and Fig. 6.1(b) the corresponding Stokes and anti-Stokes pure rotational Raman spectra.
For the non-rigid rotor, centrifugal stretching must be taken into account, and the
Q
rotational term FJ
is now given by
O
FJ
D BJJ C 1 DJ2 J C 12 C . . .
6.6.5
where D is the centrifugal stretching constant, which is always positive for the choice of
sign in eq. (6.6.5). The ratio D/B is less than 104 , so that centrifugal stretching effects
F(J )
(a)
Rigid rotor
(c)
Non-rigid rotor
J=6
J=5
J=4
J=3
J=2
J=1
J=0
Stokes
anti-Stokes
(b)
(d)
~
n1
~
n increasing
Figure 6.1 (a) Rigid AB molecule rotational energy levels and (b) the associated Stokes
and anti-Stokes rotational Raman spectra; (c) non-rigid AB molecule rotational energy levels
and (d) the associated Stokes and anti-Stokes rotational Raman spectra.
175
are small. Additional terms may be needed in eq. (6.6.5) for spectra obtained at high
resolution. Such terms are considered briefly later in eqs. (6.6.17) and (6.6.19).
Using eq. (6.6.5) with the selection rule J D C2, eq. (6.6.3) gives
SJ; jQj D 4B 6DJ C 3/2 8DJ C 3/23
6.6.6
where J D 0, 1, 2, . . . . Thus the rotational lines no longer have exactly the same spacing
as in the rigid rotor approximation; the spacing decreases slightly as J increases.
Figure 6.1(c) shows the rotational energy levels of a non-rigid heteronuclear diatomic
molecule and Fig. 6.1(d) the corresponding Stokes and anti-Stokes pure rotation Raman
spectra.
6.6.3 Heteronuclear diatomic molecule: vibrationrotation

The simplest approach for a diatomic molecule undergoing simultaneous vibration and
rotation transitions is to assume that there is no vibrationrotation interaction. The rotational terms are then still given by eq. (6.6.1) for a rigid rotor and by eq. (6.6.5) for a
non-rigid rotor, with B independent of the vibrational state.
Then, for the rigid rotor case we have for the S branch for which J D C 2
Q v0 C FJ
Q v00 FJ
Q C 2 G
Q
SJ; jQj D G
D Q vib C 4BJ C 3/2
6.6.7
Q v0 G
Q v00 .
where J D 0, 1, 2, . . . and Q vib D G
Similarly for the O branch J D 2 we have
Q v00 FJ
Q v0 C FJ
Q 2 G
Q
OJ; jQj D G
D Q vib 4BJ 1/2
6.6.8
where J D 2, 3, 4, . . . and, for the Q branch, J D 0

QJ; jQj D Q vib
6.6.9
Thus for heteronuclear diatomic molecules, the Q branch will be a single line with
jQj D Q vib . For the S branch lines jQj D Q vib C 6B, Q vib C 10B, Q vib C 14B etc., and for
the O branch lines jQj D Q vib 6B, Q vib 10B, Q vib 14B etc.
For the case of the non-rigid rotor we have for the S, O and Q branches
SJ; jQj D Q vib C 4B 6DJ C 3/2 8DJ C 3/23
6.6.10
where J D 0, 1, 2, . . .,
OJ; jQj D Q vib 4B 6DJ 1/2 C 8DJ 1/23
6.6.11
where J D 2, 3, 4, . . ., and
QJ; jQj D Q vib
6.6.12
It can be seen that, for the non-rigid rotor, the interline spacings are not constant; they
decrease slightly as J increases.
176
The Raman Effect
We now consider the more realistic case in which vibrationrotation interaction is taken
into account. We begin by introducing the rotational constant Bv , the mean value of B in
the vibrational state v which is given by

h
1
Bv D
6.6.13
82 c0 r 2 v
where is the reduced mass of the molecule, r is the internuclear distance and [1/r 2 ]v
the mean value of [1/r 2 ] for the vibrational state v. The use of a mean value is justifiable
since the period of a vibration is small compared with the period of rotation. A simplistic
analysis would suggest that Bv would be somewhat smaller than Be , the rotational constant
corresponding to the equilibrium separation re . From the standpoint of classical mechanics
with increasing vibration anharmonicity would make the mean nuclear separation greater,
(see Fig. 6.2) so that Iv > Ie and Bv < Be . This is usually, but not invariably, the case.
A quantum mechanical treatment shows that Bv is related to Be as follows:
Bv D Be e v C 1/2 C . . .
6.6.14
where e is small compared to Be . Higher terms are very rarely needed.

If the nature of the dependence of the vibrational potential energy on the internuclear
distance is known, the averaging of the rotational constant can be carried out and a
theoretical value of e obtained. For example, for a vibrational potential energy function
which has quadratic, cubic and quartic terms the following result is obtained:
e D 24
Be3 re3 g
Be2

6
Q e3
Q e
6.6.15
where Q e is the harmonic vibration wavenumber (unit: cm1 ) and g is the coefficient
of the cubic term in the potential energy function when this has units of cm1 . The
(b)
Energy
(a)
Equilibrium
separation
Internuclear separation
Figure 6.2 Comparison of (a) harmonic and (b) anharmonic potentials for a diatomic
molecule.
177
coefficient g determines the asymmetry of the potential energy curve. The first term in
eq. (6.6.15) is almost invariably the dominant one and so e is almost always positive
and hence Bv < Be .
It is interesting to note however, that if g D 0 and the potential energy curve becomes a
symmetric curve associated with an harmonic oscillator, e becomes negative and hence
Bv > Be . This is because although the mean value of r increases as v increases, it is the
mean value of 1/r 2 that is involved in eq. (6.6.13) and the behaviour of this cannot be
completely generalized as in the initial simplistic approach.
The rotational term for a diatomic molecule in the vibrational state is thus given by
Q v J D Bv JJ C 1
F
6.6.16
if centrifugal stretching is neglected. The effects arising from changes in Bv with vibrational state are much greater than those associated with centrifugal stretching and so
eq. (6.6.16) is normally an adequate approximation. However, if centrifugal stretching is
also taken into account we have
Q v J D Bv JJ C 1 Dv J2 J C 12 C Hv J3 J C 13 C . . .
F
6.6.17
where the centrifugal stretching constants Dv and Hv depend on the vibrational state. For
Dv we express this as follows
Dv D De C e v C 1/2 e v C 1/22 C . . .
6.6.18
In principle a similar expansion may be made for Hv but it is usually sufficient to write
Hv He . However in most cases Hv can be neglected and then we have
Q v J D Bv JJ C 1 Dv J2 J C 12
F
6.6.19
If we neglect centrifugal stretching, then using eq. (6.6.16) for Fv J and the selection
rule J D C2 we obtain for the S branch of a vibrationrotation band
SJ; jQj D Q vib C 6Bv0 C 5Bv0 Bv00 J C Bv0 Bv00 J2
6.6.20
with J D 0, 1, 2, . . .
Similarly for the O branch J D 2 we have
OJ; jQj D Q vib C 2Bv0 3Bv0 C Bv00 J C Bv0 Bv00 J2
6.6.21
with J D 2, 3, 4, . . .; and for the Q branch J D 0 we have

QJ; jQj D Q vib C Bv0 Bv00 J C Bv0 Bv00 J2
6.6.22
with J D 0, 1, 2, . . . . The structure of the Q branch is essentially determined by Bv0 Bv00 ,

0. Of course eqs. (6.6.20),
that is by B1 B0 D e for the vibrational transition 1
(6.6.21) and (6.6.22) reduce to eqs. (6.6.7), (6.6.8) and (6.6.9), respectively, if we set
Bv0 D Bv00 D B.
The general pattern of a vibrationrotation Raman band may now be deduced. We
shall assume that Bv0 < Bv00 and thus Bv0 Bv00 is negative. Then, from eq. (6.6.20) we see
178
The Raman Effect
that the S branch consists of a series of lines, with larger wavenumber shifts than the
pure vibration line, the first line having a shift of about 6B from the pure vibration line
and successive lines having spacings of about 4B, but with this spacing decreasing as
J increases. From eq. (6.6.21) we see that the O branch consists of a series of lines
with smaller wavenumber shifts than the pure vibration line, the first line having a shift
of about 6B from the pure vibration line and successive lines having spacings of about
4B, but with this spacing increasing as J increases. From eq. (6.6.22) we see that the Q
branch will consist of a series of closely-spaced lines with smaller wavenumber shifts than
the pure vibration line. Figure 6.3 shows typical energy levels and the resultant Stokes
J=7
J=6
G( ) + F(J)
J=5
J=4
J=3
=1
J=2
J=1
J=0
J = +2
J = 0
J = 2
J=7
J=6
J=5
J=4
J=3
(a)
u=0
J=2
J=1
J=0
Stokes
S
Q
O
J = +2 J = 0 J = 2
v~ increasing
~
v1
(b)
Figure 6.3 (a) Non-rigid and anharmonic AB molecule vibration and rotation energy levels
and (b) associated Stokes vibrationrotation Raman spectra.
179
Raman spectrum. Often the difference between Bv0 and Bv00 is so small that the lines of
the Q branch repeat overlap when the resolution is insufficient. The Q branch will then
appear much more intense than the O and S branches; and if Bv0 D Bv00 all the Q branch
lines are superimposed.
We now consider the case when centrifugal stretching is also taken into account. Then
Q v J and the selection rules J D C2 (S branch), J D 2 (O
using eq. (6.6.19) for F
branch) and J D 0 (Q branch) we obtain:
SJ; jQj D Q vib C 6Bv0 C 5Bv0 Bv00 J C Bv0 Bv00 J2
9
3
11
DJ0 DJ00 DJ0 C DJ00 2J C 3 DJ0 DJ00 2J C 32
16
2
8
1 0
1 0
00
3
00
DJ C DJ 2J C 3 DJ DJ 2J C 34
6.6.23
2
16
with J D 0, 1, 2, . . .
OJ; jQj D Q vib C 2Bv0 3Bv0 C Bv00 J C Bv0 Bv00 J2
9
3
11
DJ0 DJ00 C DJ0 C DJ00 2J 1 DJ0 DJ00 2J 12
16
2
2
1 0
1
C DJ C DJ00 2J 13 DJ0 DJ00 2J 14
6.6.24
2
16
with J D 2, 3, 4, . . . and
QJ; jQj D Q vib C Bv0 Bv00 JJ C 1
DJ0 DJ00 J2 J C 12
6.6.25
with J D 0, 1, 2, . . .
6.6.4 Homonuclear diatomic molecule: nuclear spin degeneracy

As already indicated, for homonuclear diatomic molecules the nuclear spin degeneracy
gN takes different values for odd and even J values; and the gN values depend on the
nuclear spins of the atoms involved. We shall not consider here the theory underlying
nuclear statistics, detailed accounts of which are available in a number of spectroscopic
texts, as for example, Herzberg (1950), King (1964) and Hollas (1998). For our purposes
it will suffice to know the nuclear spins, nuclear statistics and gN values and the relative
weightings of rotational line intensities which result. This information is presented in
Table 6.6 for a number of homonuclear diatomic molecules.
We see that in those cases where the nuclear spin is not zero a distinct alternation in
the intensities of successive rotational lines is produced. In the cases of 16 O2 and 18 O2
where the nuclear spin is zero, gN is zero for even J values. Thus there are no rotational
levels with even J and alternate lines are absent in their rotational Raman spectra. On
this basis, using eq. (6.6.4) for the rigid rotor case we can deduce that for 16 O2 and 18 O2
the ratio of the wavenumber separation between the first rotational line in the Stokes
180
The Raman Effect
Table 6.6 Nuclear spin statistics and relative intensities of rotational Raman lines in some
diatomic molecules.
16
Species
Nuclear spin, I
Nuclear spin degeneracy, T D 2I C 1
Statistics
Ground electronic state
Nuclear spin statistical weight gN for
J odd
J even
Relative intensities of rotational
Raman lines
J odd
J even
a
O2 ,
18
0
1
Bose
3
g
O2
H 2 , T2 ,
19
1
2
F2
D2 ,
14
N2
35
Cl2
3
2
2
Fermi
C
g
1
3
Bose
C
g
4
Fermi
C
g
1
a
3
1
3
6
5
3
1
0
3
1
1
2
5
3
Even J levels unpopulated.
spectrum and the first rotational line in the anti-Stokes spectrum to the wavenumber
separation of successive lines in either the Stokes or anti-Stokes rotational spectrum will
be 20B/8B D 5/2. This is found experimentally to be the case. If lines of odd J were
missing the corresponding ratio would be 12B/8B D 3/2.
These gross intensity alternations due to nuclear spin statistics are imposed on the
more gradual intensity change with J value produced by the population factor Ni and the
Placzek-Teller b factors which are discussed in the next section.
There is an additional factor to be considered in the rotational Raman spectrum of
the oxygen molecule because its electronic ground state is 3
g . Coupling between the
electronic spin angular momentum and the overall molecular rotation causes a splitting
of the rotational levels into triplets. This is considered later in Section 6.9 which deals
with contributions of the orbital and spin angular momenta of electrons.
6.6.5 Intensity distribution

We begin with the population factor Ni in eq. (6.3.39) and consider first pure rotational
Raman transitions. For heteronuclear diatomic molecules gN D 1, and thus it follows from
Q
eqs. (6.5.1) and (6.5.10) that in the rigid rotor approximation where FJ
D BJJ C 1
N2J C 1 expfBJJ C 1hc0 /kTg
6.6.26
Ni D
2J C 1 expfBJJ C 1hc0 /kTg
J
For sufficiently large T or small B the sum in the denominator of eq. (6.6.26) can be
treated as a continous function of J and replaced by an integral over dJ between the
limits J D 0 and J D 1. This has the value kT/hc0 B.
Figure 6.4 shows a plot of the function 2J C 1 expfBJJ C 1hc0 /kTg against J for
T D 300 K and B D 10.44 cm1 (the value for the ground electronic state of HCl). As
2J C 1 increases linearly with J this function first increases with increasing J despite the
steady decrease of the exponential term with J. The function reaches a maximum value
181
(2J + 1) exp {F (J )hc0 /kT }
4.0
3.0
2.0
1.0
10
12
1
O
O
Figure 6.4 A plot of 2J C 1 expfFJhc
0 /kTg against J; T D 300 K, and FJ/cm
D 10.44JJ C 1.
at J D 3 and subsequently decreases as J increases, becoming very small at J D 12.

Assuming that the function is continuously differentiable its maximum is given by

1
kT
T 1

6.6.27
D 0.5896
Jmax D
2Bhc0 2
B 2
Thus Jmax increases with decreasing B, and increasing T. For example, for B D 1.92 cm1
(ground electronic state of CO), Jmax D 7 at 300 K and the population remains significant
beyond J D 20. It should be noted that the number of molecules in the lowest rotational
level Ji D 0, is not zero.
The Placzek-Teller b factors which are also involved in the intensity formulae are also
functions of J but as we shall see when these factors are discussed later they are relatively
insensitive to the J value.
Thus, overall, for pure rotational scattering by heteronuclear diatomic molecules the
intensity distribution in the branches for which J 6D 0 is determined mainly by Ni
and only to a limited extent by the Placzek-Teller b factors. For homonuclear diatomic
molecules, nuclear spin imposes additional gross intensity variations as already discussed
in Section 6.6.4.
We now consider vibrationrotation Raman spectra. For both homonuclear and
heteronuclear diatomic molecules the absolute population distribution which relates to the
intensities of O and S rotational lines in the Stokes Raman vibrationrotation spectrum
associated with the vibrational transition from vi D 0 to vf D 1, is the same as for the
corresponding pure rotational bands associated with the vibrational level vi D 0. This
is because it is the populations of the initial rotational levels in the initial vibrational
level vi which determine the intensities. For the anti-Stokes Raman vibrationrotation
spectrum, which is associated with the vibrational transition from vi D 1 to vf D 0,
although the relative distribution amongst the rotational levels remains the same, in the
vi D 1 vibrational level the absolute population of all the rotational levels is reduced by
the factor exp f3hc0 Q /2kTg. This factor controls the population of the initial vibrational
182
The Raman Effect

Table 6.7 Ratio of the number of molecules in the first to that
in the zeroth vibrational level for 300 K and 1000 K for some
diatomic molecules.
Molecular
formula
H2
HCl
N2
CO
O2
S2
Cl2
I2
a
Vibrational
wavenumbera
Q vib /cm1
expfhc0 Q vib /kTg

For 300 K
For 1000 K
4160.2
2885.9
2330.7
2143.2
1556.4
721.6
556.9
213.2
2.16 109
9.77 107
1.40 105
3.43 105
5.74 104
3.14 102
6.92 102
3.60 101
2.51 103
1.57 102
3.50 102
4.58 102
1.07 101
3.54 101
4.49 101
7.36 101
In this Table Q vib D Q e 2Qe xe D Q 0 Q 0 x0 . See Chapter 5, p. 125.
level vi D 1. In most molecules this factor is so small that the intensities of anti-Stokes
vibrationrotation bands are very weak. This is illustrated by Table 6.7 which gives the
ratio of the number of molecules in the v D 1 state to that in the v D 0 state for a number
of diatomic molecules at 300 K and 1000 K.
We now proceed to obtain expressions for I/2; ?s , ?i and I/2; jjs , ?i for pure
rotation and vibrationrotation Raman spectra of a diatomic molecule using eq. (6.3.39)
together with eq. (6.3.40) or (6.3.41). We consider first I/2; ?s , ?i . It can be seen
from Chapter 5, Section 5.5.2 that
I/2; ?s , ?i D kQ Q s4 Ni hyy 2fi iI
6.6.28
It follows from the discussion on pages 164 and 165 and from eq. (6.3.40) which relates
isotropic averages of the type hxx 2fi i to G 0 fi and G 2 fi , that when J D 0 there
can be both isotropic and anisotropic contributions to I/2; ?s , ?i as both G 0 fi and
G 2 fi can then be non-zero. However, when J D C2 there can only be an anisotropic
contribution to this intensity since then only G 2 fi can be non-zero.
It is now necessary to obtain formulae for G 0 fi and G 2 fi for the case of a
diatomic molecule. As already indicated such a molecule can be regarded as a special
case of a symmetric top with K D 0 and K D 0. When these conditions are introduced
into the general definitions of G 0 fi and G 2 fi given by eqs. (6.3.24) and (6.3.25)
we obtain
6.6.29
G 0 vf Jf :vi Ji D 32Ji C 1jhvf jajvi ij2
with J D 0, so that Jf D Ji and
G
2
f 2 i 2
vf Jf :vi Ji D 2Ji C 1bJ2
f 0:Ji 0 jhv j0 jv ij
6.6.30
with J D 0, 2 so that Jf D Ji , Ji 2. In these and subsequent equations which relate

to K D 0, the subscript K is omitted from the Placzek invariant G 0 and G 2 .

183
p p
Using Table 6.4 we can replace j2
0 j by 2/ 3 for a diatomic molecule so that
G
2
f
i 2
2
vi Ji :vi Ji D 2Ji C 1bJ2
f 0:Ji 0 3 jhv j jv ij
6.6.31
with J D 0, 2.
The values of bJ2
f 0:Ji 0 may be obtained from Table 6.2 by setting K D 0 in the expresf D Ki K D 0. Then the non-zero
sions for bJ2
f Kf :Ji Ki which are associated with K
values are as follows.
JJ C 1
J D 0 : bJ2
6.6.32
i 0:Ji 0 D
2J 12J C 3
3J C 1J C 2
J D C2 : bJ2
6.6.33
i C2,0:Ji 0 D
22J C 12J C 3
3JJ 1
6.6.34
J D 2 : bJ2
i 2,0:Ji 0 D
22J 12J C 1
The case with J D 2 does not arise for pure rotational Raman spectra, as already
explained in Section 6.4 of this chapter. Table 6.8 illustrates how these three factors vary
with Ji . It can be seen that they are relatively insensitive to the Ji value particularly for
Ji > 2; and for Ji > 10 they are coming close to their limiting values of 1/4, 3/8 and 3/8,
respectively. Thus, as indicated earlier in this section, the Ji dependence of the intensity is
determined mainly by the interplay of the degeneracy factor 2Ji C 1 and the exponential
factor expfBJi Ji C 1hc0 /kTg. The last column of Table 6.8 also illustrates one sum
property of these factors, which follows from the properties of the Wigner 3-j functions.
We now consider intensities for pure rotational transitions in the ground vibrational state
vf D vi D 0. Using eq. (6.6.29) for G 0 fi eq. (6.6.31) for G 2 fi and appropriate
a
Table 6.8 Values of bJ2
f :Ji and their sum properties.
Ji
bJ2
i :Ji
bJ2
i C2:Ji
bJ2
i 2:Ji
bJ2
f :Ji
Jf
0
1
2
3
4
5
6
..
10
..
20
a
0
0.40
0.29
0.27
0.26
0.26
0.26
..
0.25
..
0.25
1.0
0.60
0.51
0.48
0.45
0.44
0.43
0
0
0.20
0.25
0.29
0.30
0.31
0.41
..
0.375
0.34
..
0.375
1.0
1.0
1.0
1.0
1.0
1.0
1.0
..
1.0
..
1.0
Since Kf D Ki D 0 here, a contracted form of the Placzek Teller b2

factor is used in which K is omitted. (See also Reference Tables 6.1
and 6.2).
184
The Raman Effect
values of the PlaczekTeller b2 factors given by eqs. (6.6.32) and (6.6.33) we can obtain
from eq. (6.3.40) expressions for hyy 2fi i for J D 0 and J D 2 which, when inserted
in eq. (6.6.28), give the following results:
J D 0, vf D vi D 0

4JJ C 1
s
i
4
2
2
6.6.35
I/2; ? , ? D kQ Q s Ni ja0 j C
j 0 j I
452J 12J C 3
J D C2,
vf D vi D 0
s
I/2; ? , ? D
kQ Q s4 Ni

2J C 1J C 2
2
j 0 j I
152J C 12J C 3
6.6.36
In the above equations we have made the following substitutions:

ja0 j2 D jhvf D 0jajvi D 0ij2
2
j 0 j D jv D 0j jv D 0ij
6.6.37
6.6.38
We now turn to I/2; jjs , ?i . It can be seen from Chapter 5, Section 5.5.2 that
I/2; jjs , ?i D kQ Q s4 Ni hxy 2fi iI
6.6.39
Using eq. (6.6.39) with eqs. (6.3.41), (6.6.31) and (6.6.33) we obtain the following
results:
J D 0, vf D vi D 0

JJ C 1
s
i
4
2
j 0 j I
I/2; jj , ? D kQ Q s Ni
6.6.40
152J 12J C 3
J D C2, vf D vi D 0
s
I/2; jj , ? D
kQ Q s4 Ni

J C 1J C 2
2
j 0 j I
102J C 12J C 3
6.6.41
The results for intensities of rotational lines associated with the fundamental vibrational
vi D 0 can be obtained from the above formulae by introducing
transition vf D 1
vibrational transition polarizability tensor invariants in place of equilibrium tensor invariants and remembering that J D 2 is now allowed. The required formulae are:
J D 0Q branch

4JJ C 1
I/2; ?s , ?i D kQ Q s4 Ni ja10 j2 C
6.6.42
j 10 j2 I
452J 12J C 3
J D 2 (S branch)

s
kQ Q s4 Ni
kQ Q s4 Ni
I/2; ? , ? D
J D 2 (O branch)
s
I/2; ? , ? D

2J C 1J C 2
2
j 10 j I
152J C 12J C 3
6.6.43

2JJ 1
2
j 10 j I
152J 12J C 1
6.6.44

J D 0 (Q branch)

s
I/2; jj , ? D
kQ Q s4 Ni

JJ C 1
2
j 10 j I
152J 12J C 3
J D 2 (S branch)

s
I/2; jj , ? D
kQ Q s4 Ni
J D 2 (O branch)

J C 1J C 2
2

I
102J C 12J C 3 10
185
6.6.45
6.6.46

JJ 1
2
I/2; jj , ? D

6.6.47
I
102J 12J C 1 10
Here we have introduced a shorthand notation for the invariants of the vibrational transition polarizability tensor and written
s
ja10 j2
kQ Q s4 Ni
for jhvf D 1jajvi D 0ij2 , and j 10 j2 for jhvf D 1j jvi D 0ij2
It should be noted that in a diatomic molecule both ja10 j2 and j 10 j2 are non-zero
since a and both belong to the same symmetry class as the vibration of a diatomic
C
molecule, namely C
g for homonuclear diatomic molecules (D1h point group) or for
heteronuclear diatomic molecules (C1v point group).
For I/2; jjs , ?i which involves anisotropic scattering only, the ratio of the total
intensity of the Q branch lines to the total intensity of the O and S branch lines is
given by
1
JJ C 1
15 J 2J 12J 3
6.6.48
1 J C 1J C 2
1
JJ 1
C
10 J 2J 12J C 1 10 J 2J C 12J C 3
If the temperature is high enough and/or the value of B small enough so that the major
contribution to the intensity comes from molecules having large values of J, the above
ratio tends to 1:3. In this case only a quarter of the total anisotropic scattering is in the
Q branch.
If we introduce the Placzek polarizability theory and write
a D a0 C a0 Q0
0
D 0 C Q0
6.6.49
6.6.50
where a0 D a/Q0 , 0 D /Q0 and Q0 is the amplitude of the normal coordinate

of vibration we can make substitutions in the intensity formulae given above based on
the following equations :

vf D vi
avf vi D a0
6.6.51
vf D vi C 1 avf vi D bv vi C 11/2 a0

vf D vi
vf vi D 0
6.6.52
vf D vi C 1 vf vi D bv vi C 11/2 a0
Chapter 5, Section 5.7 discusses these relationships.
186
The Raman Effect
Here bv is defined by eq. (5.7.8) and vi is the vibrational number of the initial vibrational
state. For the fundamental vibrational transition we set vi D 0.
Degrees of depolarization /2 :?i for the various branches may be readily obtained
from the above intensity formulae; and it is straightforward to extend the procedures
detailed above to obtain intensity formulae for other cases of pi and ps . The isotropic
averages of squares of transition polarizability tensor components for diatomic molecules
for rotation and vibrationrotation transitions are given in Reference Table 6.1. Intensity
and polarization formulae for diatomic molecules are given in Reference Table 6.2 for
various cases of pi , ps and
.
6.7 SYMMETRIC TOP MOLECULE

6.7.1 Introduction
In general, non-linear molecules have three principal moments of inertia, Ia , Ib and Ic
where a, b and c are the principal inertial axes. The corresponding rotational constants
A, B and C are defined as follows:
h
h
h
; BD
; CD
6.7.1
AD
2
2
2
8 c0 IA
8 c0 IB
8 c0 IC
Here A, B and C have the dimensions of reciprocal length, and we use the conventional
unit cm1 for these quantities.
By convention the labels are assigned so that Ia Ib Ic , that is A B C. If Ia <
Ib D Ic the molecule is said to be a prolate symmetric top; if Ia D Ib < Ic then the
molecule is said to be an oblate symmetric top. If Ia D Ib D Ic the molecule is said to be
a spherical top. If Ia 6D Ib 6D Ic the molecule is said to be an asymmetric top. Diatomic
and linear polyatomic molecules normally constitute special cases of a prolate symmetric
top with 0 D Ia < Ib D Ic and with the quantum number K D 0 in a electronic state,
K D 1 in a electronic state, K D 2 in a electronic state and so on.
In a planar molecule the moment of inertia about the axis perpendicular to the plane
is equal to the sum of the two in-plane principal moments of inertia. Consequently the
moment of inertia about the axis perpendicular to the plane must be the largest. Thus
this axis is the c axis and we have Ic D Ia C Ib . If, in addition, the planar molecule
is an oblate symmetric top, then Ia D Ib and so Ic D 2Ib . These relationships are only
true for molecules in their equilibrium configurations. Some molecules which are strictly
asymmetric tops may have two moments of inertia which are accidentally nearly equal.
If Ia < Ib Ic the molecule is a prolate near-symmetric top, and if Ia Ib < Ic the
molecule is an oblate near-symmetric top. Accidental near-spherical tops are also possible
but they are rare.
In this section we consider the vibrationrotation Raman spectra of a prolate symmetric
top. The corresponding results for an oblate symmetric top can be obtained by replacing
A by C.
Q see Section 6.4.
This is in accord with our definition of rotational terms F,
However, if the molecule does not have a 1 ground electronic state, electronic orbital and spin angular
momenta have to be taken into account.
187
6.7.2 Symmetric Top: Pure Rotation

If vibrationrotation interaction is ignored the rotational term value for a rigid prolate
symmetric top is given by
Q
FJ,
K D BJJ C 1 C A BK2
6.7.2
Here J, which can take the values 0, 1, 2, . . ., is the quantum number relating to the total
angular momentum P of the nuclear framework which is given by P D [JJ C 1]1/2h .
The projection of P along the figure axis, which we define as Pa for a prolate symmetric
top, is also quantised. Its magnitude is given by Pa D Kh where K can take the values
0, 1, 2, 3, . . . , J. Since Pa is a component of P, K cannot be greater than J, so that
for a given K value we have the condition J D K, K C 1, K C 2, . . . . The value of the
component Pa may be positive or negative and is given by Pa D kh where for a given
J value, k D 0, 1, 2, . . . , J. Clearly K D jkj. It follows from eq. (6.7.2) that states
which differ only in the sign of k will have the same energies. Thus a pair of states
with wave functions characterised by Ck and k exhibit double degeneracy. Classically
this may be thought of as arising from clockwise or anticlockwise rotation about the top
axis resulting in the same magnitude of the angular momentum. For k D 0 (and K D 0)
there is no angular momentum about the top axis and so no degeneracy. We see that the
levels designated by the unsigned quantum number K are doubly degenerate except when
K D 0. This degeneracy cannot be removed by either external or internal fields.
The projection of P along a space-fixed axis defined by an applied field is given by
Mh where M can take the values 0, 1, 2, . . . , J, that is 2J C 1 values in all. In
the absence of an applied field all the 2J C 1 states which differ only in M values have
the same rotational energy because the orientation of the angular momentum is space is
immaterial. Then, as the rotational term is independent of M, an energy level specified
by the quantum numbers J and K is 2J C 1-fold degenerate when K D 0 and 22J C 1
degenerate when K > 0.
The rotational quantum numbers associated with a symmetric top are summarized
below:
J D 0, 1, 2, 3, . . .
k D 0, 1, 2, 3, . . . , J
K D 0, 1, 2, 3, . . . , J
M D 0, 1, 2, 3, . . . , J
with the additional restriction that for a given K D jkj
J D K, K C 1, K C 2, . . .
6.7.3
For the greater part of the treatment of rotational transitions in symmetric top molecules
that follows we can use the unsigned quantum number K. However in some special cases
It is however removed by Coriolis interaction and by l-doubling, see Section 6.7.3.
188
The Raman Effect
the sign of the quantum number becomes relevant and it is then necessary to introduce
the signed quantum number k.
It should be noted that in the treatment of selection rules based on irreducible transition
polarizability components given in section 6.2 of this chapter, we have used the set of
rotational functions of the form jJKMi. If we had used the complete set of rotational
functions jJkMi the selection rules for J and M would be unchanged and the selection
rules for k would be drawn from Ck D 0, 1, 2 and k D 0, 1, 2, depending
j
on the value of j in m .
Q
We see from eq. (6.7.2) that for a prolate symmetric top the dependence of FJ,
K on K
Q
is determined by the relative magnitudes of A and B. Also when K D 0, FJ,
K depends
only on Ib , the moment of inertia about an axis perpendicular to the figure axis. When
K J the energy is determined principally by Ia , the moment of inertia about the figure
Q
axis since, when J D K, FJ,
K D AJ2 C BJ which is dominated by AJ2 when J is large.
The rotational energy levels and degeneracies of a symmetric top, both prolate and
oblate cases, are illustrated in Fig. 6.5 for K D 0, 1, 2, 3. It should be noted that for
K D 1, the lowest value of J is 1, for K D 2, the lowest value of J is 2 and so on. Also
for each value of K the rotational levels have exactly the same spacing assuming as in
eq. (6.7.2) that there is no centrifugal distortion. It will be seen later that some centrifugal
distortion terms are K dependent.
We now consider the rotational Raman spectra which result from the allowed transitions
between the rotational energy levels of a rigid symmetric top. For the S branch for which
J D C2 and K D 0 we have
Q C 2, K FJ,
Q
SJ; jQj D FJ
K
6.7.4
Then, using eq. (6.7.2) for the rotational term we obtain
SJ; jQj D 4BJ C 3/2
6.7.5
where J D 0, 1, 2, . . . . It follows that S0; jj D 6B, S1; jj D 10B, S2; jj D
14B and so on. Thus in the S branch the first rotational line has a wavenumber shift of 6B
from the exciting line and successive rotational lines have equal interline spacings of 4B.
Similarly for the R branch, for which J D C1 and K D 0, we have
RJ; jQj D 2BJ C 1
6.7.6
where J D 1, 2, 3, . . . . The J D 0 case is excluded since J D C1 does not operate

for K D 0 because the PlaczekTeller factor bJ2
f Kf :Ji Ki for J D C1, K D 0 involves
K2 and so is zero when K D 0; and for K D 1 the lowest J value is 1. It follows that
R1; jQj D 4B, R2; jQj D 6B and so on. Thus in the R branch the first rotational line
has a wavenumber shift of 4B from the exciting line and successive rotational lines have
equal interline spacings of 2B.
It is conceptually useful to consider the energy levels of a rigid prolate symmetric top,
defined by eq. (6.7.2) as a set of energy ladders each ladder associated with a particular
value of K. This is illustrated in Fig. 6.6(a), and for K D 0, 1 and 2. The selection rule
K D 0 means that allowed J transitions occur only between J levels in the same ladder
of energy levels. The energy spacings in each ladder are independent of K although their
189
~
F(J, K)/cm1
Prolate top
Oblate top
400
J
J
(17)+
300
(15)
(13)+
200
(11)
100
(9)+
(34)
8 +
(30)+
7
(26)
6 +
(22)+
5
(18)
+
(7)
(14)+
3
(5)+
(10)
2
+
(6)+
1
(3)
(1)+
K=0
K=1
(34)+
(30)
+
7
(26)+
(22)
+
(18)+
(34)
+
8
(14)
+
(10)+
(30)+
(26)
+
(22)+
(15)+
(13)+
K=3
(30)
(26)
J
J
8 (34)+
8 (34) J
+
8
7 (30)+
7 (30)
+
7
6 (26)+
+ 6 (26)
5 (22)+
+ 5 (22)
4 (18)+
+ 4 (18)
5
(11)+
(9)+
(22)
(18)
3
(7)+
(3)+
(1)+
K=2
(34)
(5)+
(18)
4
+
(14)+
J
(17)+
K=0
(14)
3
2
1
0
(10)
(6)
K=1
3 (14)+
3 (14)
+
2 (10)+
2
+
1
K=2
K=3
Figure 6.5 The rotational energy levels of a prolate symmetric top and an oblate symmetric
top for K D 0,1,2,3. The degeneracies of the levels, 2J C 1 for K D 0 and 22J C 1 for
K > 0, are given in brackets. The parities of the levels are indicated by the signs C or . For
K > 0 the parities of the components that are doubly degenerate with respect to K are given.
For non-planar molecules each component in the oblate top is also accompanied by one of
opposite parity because of the degeneracy with respect to inversion doubling. The rotational
energies have been calculated using B D 5 cm1 , A B D 2.5 cm1 for the prolate top; and
B D 5 cm1 , C B D 2.5 cm1 for the oblate top.
absolute values increase by A B 2K C 1 between ladder K and ladder K C 1. The

wavenumber shifts associated with the allowed J transitions in a given ladder are said
to constitute a sub-band. As these wavenumber shifts are the same in each energy ladder
the resultant spectrum is a superposition of a number of identical sub-bands. Thus for
example the number of coincident components in each S line is in principle J C 1, the
190
The Raman Effect

~
F (J,K )/cm1
J = +1
J = +2 J
J = +1
J = +1
J = +2 J
J = +2 J
2
1
0
2
1
K=0
K=1
K=2
Figure 6.6(a) Raman-active transitions between the rotational levels of a prolate symmetric
top for K D 0, K D 1 and K D 2. J D C1 (R branch) transitions are indicated by " and
J D C2 (S branch) transitions by . Note the absence of J D C1 transitions when K D 0.
Stokes
anti-Stokes
R branch
S branch
R + S branches
v~1
v~ increasing
Figure 6.6(b) Rigid prolate symmetric top molecule: R branch, S branch, R and S branches
combined, in the Stokes and anti-Stokes rotation Raman spectra.
number of energy ladders available for a given initial J value. However, the number of
accessible K values will depend on the temperature of the sample.
The composition of the resultant spectrum is illustrated in Fig. 6.6(b) which shows the
R branch, the S branch and also the R and S branches combined. It can be seen from these
diagrams or from comparison of eqs. (6.7.4) and (6.7.5) that the lines of the R branches for
191
which J is even in RJ coincide with the lines of the S branches. As a result there is an
apparent intensity alternation. This is quite distinct from any intensity alternations arising
from nuclear spin effects. Also, as the S branch extends to larger wavenumber shifts than
does the R branch, this apparent intensity alternation does not extend over the whole
rotational spectrum. This is not illustrated in Fig. 6.6(b) which relates only to small J
values. The R and S branches are observed as both Stokes and anti-Stokes Raman spectra.
When centrifugal stretching is taken into account, the rotational term for a symmetric
top is given by
Q
FJ,
K D BJJ C 1 C A BK2 DJ J2 J C 12 DJK JJ C 1K2 DK K4
6.7.7
It then follows that
SK J; jQj D 4B 6DJ J C 3/2 4DJK K2 J C 3/2 8DJ J C 3/23
6.7.8
where J D 0, 1, 2, . . . and
RK J; jQj D 2BJ C 1 2DJK K2 J C 1 4DJ J C 13
6.7.9
where J D 1, 2, 3, . . . .
These formulae contain two centrifugal stretching terms and as a result, although the
general pattern of the R and S branches is largely unchanged, the interline spacings are
no longer exactly 4B and 2B, respectively. Also since the DJK term is K dependent, the
level spacings are not quite identical in each energy ladder and the components of a line
with different K values no longer coincide exactly (see Fig. 6.7).
6.7.3 Symmetric Top: VibrationRotation

For a symmetric top the selection rules for vibrationrotation bands depend on the
symmetry of the vibration involved.
We consider first totally symmetric vibrations. We assume initially no vibrationrotation
interaction and that the rigid rotor approximation applies. Then using the rigid rotor term
K=3
100
K=2
50
K=1 K=0
105/cm1
n increasing
Figure 6.7 Diagrammatic representation of the non-coincidence of components associated with the Stokes S branch transition J D 3
J D 1 of a symmetric top arising
from the K-dependent centrifugal stretching term DJK . The shifts are calculated using
DJK D C105 cm1 .
192
The Raman Effect
Q (J, K) given by eq. (6.7.2) and the selection rules J D C2, K D 0 we have for the
F
Stokes S branch
6.7.10
SJ; jQj D Q vib C 4BJ C 3/2
where J D 0, 1, 2, . . . . Similarly for the R branch (J D C1, K D 0) we have
RJ; jQj D Q vib C 2BJ C 1
6.7.11
where J D 1, 2, 3, . . . . For the Q branch (J D 0, K D 0) we have

QJ; jQj D Q vib
6.7.12
For the P branch (J D 1, K D 0) we have

PJ; jQj D Q vib 2BJ
6.7.13
where J D 1, 2, 3, . . . ; and for the O branch (J D 2, K D 0)

OJ; jQj D Q vib 4BJ 1/2
6.7.14
where J D 2, 3, 4, . . . .
It is helpful to extend to vibrationrotation transitions the energy ladder concept introduced earlier. We now have to consider two sets of energy ladders, one associated with
the lower vibrational level 00 and the other with the upper vibrational level 0 . This situation
is illustrated in Fig. 6.8. When there is no vibration-rotation interaction so that Bv0 D Bv00
and A0v D A00v and centrifugal stretching is ignored, the energy ladders are the same in each
vibrational level. The selection rule K D 0 restricts J transitions to energy ladders in
the upper and lower vibrational levels with the same K value and hence the resultant
spectrum is a superposition of the allowed J transitions associated with each energy
ladder. In other words the sub-bands are identical and coincide exactly. The situation
parallels that for the pure rotational spectrum except that now five branches are allowed.
These five rotational branches are illustrated separately and combined in Fig. 6.9, for a
Stokes vibrationrotation spectrum for a limited range of J values.
We see that on one side of the single-line Q branch, R branch lines of even J coincide
with S branch lines; and on the other side P branch lines of even J coincide with O
branch lines. Thus an apparent intensity alternation is produced. However, since the S
and O branches extend over a larger wavenumber range than the R and P branches, this
apparent intensity alternation does not extend over the whole range. This is not illustrated
in Fig. 6.9 which relates only to small J values.
When rotationvibration interaction is taken into account, then neglecting centrifugal
distortion the rotational term for a symmetric top in the vibrational state v is given by
Q v J, K D Bv JJ C 1 C Av Bv K2
F
6.7.15
Using the selection rules J D C2, K D 0 we have for the S branch

SK J; jQj D Q vib C 6Bv0 C 5Bv0 Bv00 J C Bv0 Bv00 J2
C [A0v Bv0 A00v Bv00 ]K2
6.7.16
193
J
7
11
Energy
Upper
vibrational
level
F (J, K )
10
4
9
8
7
6
5
4
3
2
1
0
7
10
6
5
Lower
vibrational
level
F (J, K)
7
6
5
4
3
2
1
0
K=0
K=1
K=2
K=3
K=4
Figure 6.8 Rotational levels of a prolate symmetric top for K D 0, 1, 2, 3 and 4 for the
vibrational levels v0 and v00 . Transitions between J levels are restricted to J D 2, 1 and
0 with K D 0.
where J D 0, 1, 2, . . . . Similarly for the R branch (J D C1, K D 0) we have

RK J; jQj D Q vib C 2Bv0 C 3Bv0 Bv00 J C Bv0 Bv00 J2
6.7.17
where J D 1, 2, 3, . . . . For the Q branch (J D 0, K D 0) we have

QK J; jQj D Q vib C Bv0 Bv00 JJ C 1
6.7.18
194
The Raman Effect

S branch
DJ=+2
DK=0
n~1
R branch
DJ=+1
DK=0
n~1
Q branch
DJ =0
DK=0
n~1
P branch
DJ =1
DK=0
O branch
n~1
DJ =2
DK=0
n~1
Resultant spectrum
n increasing
n~1
Figure 6.9 Stokes Raman vibrationrotation spectrum associated with a totally symmetric
vibration for a symmetric top with Bv0 D Bv00 , A0v D A00v and centrifugal stretching ignored. The
five branches are shown separately and also superimposed to give the resultant Stokes Raman
spectrum. Each branch is itself the superposition of the allowed J transitions in each K
sub-band.
where J D 0, 1, 2, . . . . For the P branch (J D 1, K D 0) we have

PK J; jQj D Q vib Bv0 C Bv00 J C Bv0 Bv00 J2
6.7.19
where J D 1, 2, 3, . . . , and for the O branch (J D 2, K D 0) we have

OK J; jQj D Q vib C 2Bv0 3Bv0 C Bv00 J C Bv0 Bv00 J2
6.7.20
where J D 2, 3, 4, . . . . In the symbols SK J, RK J etc., in eqs. (6.7.16) to (6.7.20), J

and K are the quantum numbers of the lower state for the transition involved.
The general pattern of a vibrationrotation band may be deduced from these equations.
We first disregard the term [A0v Bv0 A00v Bv00 ]K2 and recall that Bv0 < Bv00 and hence
(Bv0 Bv00 ), which is small, is negative. Then the Q branch will consist of a series of
closely-spaced lines of lower wavenumber than Q vib . The R branch lines of even J will
almost coincide with S branch lines; and P branch lines of even J will almost coincide
195
with O branch lines. The spacings of the R and S branch lines will decrease slightly as J
increases. The spacings of the P and O branches will increase slightly as J increases.
When the factor [A0v Bv0 A00v Bv00 ]K2 is taken into account the energy spacings
are not quite the same in each energy ladder (defined by K) and so the contributions from
the different energy ladders no longer coincide. The allowed transitions in the Kth energy
ladder give rise to what is termed a sub-band consisting of O, P, Q, R and S branches;
and the resultant overall spectrum is made up of (J C 1) such sub-bands which differ
slightly from each other. The origin of the Kth sub-band is the Q branch associated with
that K value and the set of origins is given by the sub-band formula with J D 0. Thus
setting J D 0 in eq. (6.7.18), we obtain
QK ; jQj D Q vib C [A0v Bv0 A00v Bv00 ]K2
6.7.21
which gives the origins of the sub-bands.

For a non-rigid symmetric top the situation becomes more complicated. The rotational
term for a symmetric top in the vibrational state v, including centrifugal stretching, is
given by an appropriately labelled form of eq. (6.7.7):
Q v J, K D Bv JJ C 1 C Av Bv K2 DvJ J2 J C 12
F
DvJK JJ C 1K2 DvK K4
6.7.22
It is useful at this point to make a further embellishment of the symbols used so far and
introduce the presuperscript symbols O, P, Q, R and S to indicate the K values. Thus Q Q
and Q S label branches for which J D 0, K D 0 and J D C2, K D 0, respectively,
and Q QK J) and Q SK J) label particular lines within those branches.
We now use the selection rules with eq. (6.7.22) to obtain formulae for the wavenumbers of the various branches. For the Q branch, (J D 0, K D 0) we obtain
Q
Q D Q vib C B0 B00 JJ C 1 C [A0 B0 A00 B00 ]K2 DJ0 DJ00
QK J; jj
0
00
0
00
J2 J C 12 DJK
DJK
JJ C 1K2 DK
DK
K4
6.7.23
where the upper and lower vibrational state constants are denoted by single and double
primes, respectively, and the subscript v is dropped. The quantum numbers J and K refer to
the lower state from which the transitions originate. For the O branch (J D 2, K D 0)
and the S branch (J D C2, K D 0) we obtain

3 0
Q
Q
OK J and SK J; jQj D Q vib C
B B00 C [A0 B0 A00 B00 ]K2
4

9 0
3 0
00
00
2
0
00
4
DJ DJ DJK DJK K DK DK K
16
4

3
0
00
0
00
0
00
2
C B C B DJ C DJ DJK C DJK K m
2

1
11 0
0
00
00
0
00
2
C
B B DJ DJ DJK DJK K m2
4
2
1 0
1
DJ C DJ00 m3 DJ0 DJ00 m4
6.7.24
2
16
196
The Raman Effect
where m D 2J C 1) for the O branch and (2J C 3) for the S branch. For the P branch
(J D 1, K D 0) and the R branch (J D C1, K D 0) we have
Q
0
00
PK J and Q RK J; jQj D Q vib C f[A0 B0 A00 B00 ]K2 DK
DK
K4 g
0
00
C [B0 C B00 DJK
C DJK
K2 ]m
0
00
C [B0 B00 DJ0 DJ00 DJK
DJK
K2 ]m2
2DJ0 C DJ00 m3 DJ0 DJ00 m4
6.7.25
where m D J for the P branch and J C 1 for the R branch.

Since the difference between corresponding upper and lower state constants is quite
small we can get a basic picture of the general appearance of these bands by setting
A0 D A00 D A and B0 D B00 D B and ignoring all the centrifugal distortion terms. Equations (6.7.23) to (6.7.25) then reduce to eqs. (6.7.10) to (6.7.14) and the pattern of the
bands is essentially as already described for these equations.
If we set A0 D A00 D A and B0 D B00 D B and retain the centrifugal distortion terms but
0
00
with DJ0 D DJ00 D DJ and DJK
D DJK
D DJK , then each of the lines splits into (J C 1)
lines, one for each value of K. These lines are separated by 2DJK 2K C 1) for the P and
R branches and by 4DJK 2K C 1) for the O and S branches, as is readily verified from
the above equations.
We now consider non-totally symmetric, non-degenerate, vibrational modes. Raman
bands associated with such vibrations can occur only in those symmetric top molecules
which have a fourfold axis of symmetry (point groups C4 , C4v , C4h , D2d , D4 , D4h and
S4 ). The selection rules are J D 0, 1, 2 and K D 2 only. Using these with the
rotational term given by eq. (6.7.22) we obtain the results given below. In these formulae
the upper sign goes with K D C2 and the lower sign with K D 2.
For the Q branch, (J D 0, K D 2) we have
O,S
QK J; jQj D sub O,S QK C fJ, K 2
6.7.26
where the sub-band origins, which are defined to be those wavenumbers associated with
J D 0, are given by
0
sub O,S QK D Q vib C 4[A0 B0 4DK
] 4[A0 B0 8D0 ]K
0
0
0
00
C [A0 B0 A00 B00 24DK
]K2 8DK
K3 DK
DK
K4
6.7.27
and where
0
00
fJ, K 2 D [B0 B00 DJK
K 22 C DJK
K2 ]JJ C 1
DJ0 DJ00 J2 J C 12
6.7.28
As noted previously, the upper and lower vibrational state constants are denoted by single
and double primes, respectively, and the subscript v is dropped.
For the P branch (J D 1, K D 2) and the R branch (J D C1, K D 2)
we have

O,S
197
P, RK J; jQj D Q vib C [A0 B0 K 22 A00 B00 K2 ]

0
00 4
0
00
K 22 DK
K ] C [B0 C B00 DJK
K 22 DJK
K2 ]m
[DK
0
00
C [B0 B00 DJ0 DJ00 DJK
K 22 C DJK
K2 ]m2
6.7.29
where m D J for the P branch and m D J C 1) for the R branch.

For the O branch (J D 2, K D 2) and the S branch (J D C2, K D 2)
we have
O,S
3
O, SK J; jQj D Q vib C B0 B00 C [A0 B0 K 22 A00 B00 K2 ]
4
3 0
9 0
00
K 22 DJK
K2 ]
DJ DJ00 [DJK
16
4
0
00 4
[DK
K 24 DK
K ]
3
0
00
C [B0 C B00 DJ0 C DJ00 DJK
K 22 DJK
K2 ]m
2
11
1 0
00
0
00
K 22 C DJK
K2 ]m2
C [B B DJ0 DJ00 DJK
4
2
1
1
DJ0 C DJ00 m3 DJ0 DJ00 m4
6.7.30
2
16
where m D 2J C 1) for the O branch and m D 2J C 3) for the S branch. The energy
ladder concept will again prove helpful in understanding the implications of these rather
complicated formulae. The selection rule K D 2 for non-totally symmetric modes means
that J transitions now occur only between energy ladders whose K value differs by 2.
As a result, even when the vibrational dependence of the rotational constants is neglected,
(that is A0 D A00 D A and B0 D B00 D B) and centrifugal stretching is ignored (all D terms
zero), the sub-bands will not coincide. Also there will now be two sets of sub-branches,
one with K D C2 and one with K D 2. It follows from eq. (6.7.22) that the spacings
of each energy ladder K differ by (A BK2 in the approximation under consideration. A
transition from J D 0 in the energy ladder K to J D 0 in energy ladder K C 2K D C2)
involves an energy difference of (A B[K C 22 K2 ] D 4A BK C 1); and for K
to K 2K D 2) an energy difference of 4A BK 1). Thus unlike the case for
totally symmetric modes, the Q sub-branch origins do not coincide. They are separated by
4A B) in this approximation and follow a simplified form of eq. (6.7.27) namely
sub D Q vib C 4A B 4[A B]K,
6.7.31
where the C sign applies to K C 2 and the sign to K D 2. It should be noted that
the first line (K D 0) in the series of Q branches with K D C2 occurs at Q vib C 4A B),
whereas the first line (K D 1) in the series of Q branches with K D 2 occurs at Q vib .
The spacing of the sub-band origins of the O,S P, R and O,S O, S branches (defined by
m D 0, K D 0, 1, 2, . . .) in eqs. (6.7.29) and (6.7.30) is also 4A B). These transitions of
198
The Raman Effect
spacing 4A B) are superimposed on a finer line structure in O,S P, R branches of spacing
2B, (obtained by letting m D J D 1); and of spacing 4B in the O,S O, S branches,
(obtained by letting m D 2 for J D 1). The general appearance of such bands is quite
complex.
In the treatment of transitions between non-degenerate states we could neglect Coriolis
interactions because they are weak; they contribute only to the vibrational dependence of
the rotational constant. However, the situation is quite different in the case of degenerate
vibrational states. Significant Coriolis interaction between vibration and rotation can then
arise, producing a component of vibrational angular momentum along the figure axis
which adds to or subtracts from the pure rotational angular momentum associated with
this axis. The magnitude of this component is jt lth j where t , the dimensionless Coriolis coupling coefficient of the t-th mode, has magnitude 0 jt j 1 and the quantum
number lt takes the values vt , vt 2, vt 4 . . . , vt where vt is the vibrational quantum
number of the t-th vibrational mode. When the vibrational motions associated with the
degenerate vibration are perpendicular to the figure axis, as for example in a planar X3
molecule which has one degenerate vibration with jt j D 1, the magnitude of the vibrational angular momentum associated with the figure axis is jlth j as for linear molecules.
In symmetric tops generally where there is more than one degenerate vibration the vibrational angular momentum associated with the figure axis can be less than jlth j since in
general 0 jt j 1. For example the molecule CF3 H (point group C3 ) has three doubly
degenerate vibrations, 4 , 5 and 6 with Coriolis coefficients of 0.95, 0.79 and 0.81,
respectively.
When Coriolis interactions are taken into account the definition of Fv J, K) given in
eq. (6.7.22) must be replaced by
Q v J, K, l D Bv JJ C 1 C Av Bv K2 2Ak klk
F
DvJ J2 J C 12 DvJK JJ C 1K2 Dv K4
6.7.32
This formula neglects certain higher order interactions which are not important here. In the
third term on the righthand side of this equation we have introduced the signed quantum
number k because the relative signs of the rotational and vibrational angular momenta
are important. The term 2Ak klt is positive when the vibrational angular momentum
has the same direction as the rotational angular momentum and negative when these two
momenta have opposite directions. The quantum number lt takes the values specified in
the previous paragraph.
When vt D 0, lt D 0 and eq. (6.7.32) reduces to eq. (6.7.22). The case for vt D 1, corresponding to single excitation of a degenerate vibrational mode t, is the one most frequently
encountered and we shall confine ourselves to this case henceforth. The rotational energy
levels associated with a singly excited E species vibration of a prolate symmetric top are
shown in Fig. 6.10. As vt D 1 then lt D 1 and all the levels are split into two components except when k D 0. The splitting is proportional to k (and therefore K) and also
In this section we use the t-th vibrational mode instead of the k-th vibrational mode used elsewhere, in
order to avoid confusion with the signed quantum number k.
See Section 6.8.1 in this Chapter.
199
J
J
J
J
7
5
~
F(J,K,l)
4
2
1
0
E
E
E
A2
A1
A1
A2
A2
A1
A1
A2
A2
A1
A1
A2
A2
A1
+lt
+lt
lt
E
E
6
E
5
E
4
2
1
E
3
E
+lt
lt
A2
A1
A1
A1
A2
A2
A1
A1
E
6
E
5
A2
A2
A1
lt
E
4
E
E
E
A2
E
E
E
k = +1, l = 1
k = +1, lt = +1
k = 1, l = +1
k = 1, lt = 1
K=0
K=1
K=2
K=3
Figure 6.10 Rotational term values associated with an E vibrational state of a prolate
symmetric top.
to t . For K > 0 there is a double degeneracy of each of the levels produced by this
splitting. For example when K D 1 and jlk j D 1 there are two doubly degenerate levels;
Q v (J, K, l) of 2At arising from k D C1, lt D C1 and
one pair with a contribution to F
k D 1, lt D 1 and a second pair with a contribution of C2At , arising from k D 1,
lt D C1 and k D C1, lt D 1. The labels Cl and l are attached to the components
according as to whether k and l have the same or different signs, respectively. The two
pairs of doubly degenerate rotationvibration levels can be shown to have the symmetry
species E and A1 C A2 . The degeneracy of the E levels cannot be removed but that of
the A1 , A2 pairs can. The splitting of these pairs is called l-type doubling and the effect is
small except when K D 1. It can be observed in high resolution spectra but we shall disregard it in the treatment that follows. The rotational energy levels associated with a singly
excited E species vibration of a prolate symmetric top are shown, appropriately labelled,
in Fig. 6.10. Note that one degenerate pair of levels of E class symmetry is shown unsplit
whereas the other degenerate pair of levels of A1 C A2 symmetry is shown split.
We now apply eq. (6.7.32) to cases of rotationvibration transitions from the totally
symmetric ground state (A species, all vt D 0) to a degenerate excited vibrational state
See for example Hougen (1962).
200
The Raman Effect
Table 6.9 Raman selection rules for (Cl) and (l) components in transitions from a totally
symmetric ground state to a degenerate excited state.a
Point groups C3 C3 , C3v , C3h , D3 , D3d , D3h , S6

K D C1 :
ECl
A
K D 1 :
El
A

K D C2 :
El
A
K D 2 :
ECl
A
Point groups C4 C4 , C4v , C4h , D4 , D4h
K D C1 :
K D 1 :
ECl
El
K D 2 :
A
A
Qsub D 2[A1 B]
Qsub D 4[A1 C 12 B]
Qsub D 2[A1 B]
Qsub D 4[A B]
Point groups Cn with n 5Cn , Cnv , Cnh , Dn , Dnh , Dnd with n odd, and Sm with 12 m odd

K D C1 :
E1 Cl
A
Qsub D 2[A1 B]
K D 1 :
E1 l
A

A
K D C2 :
E2 Cl
Qsub D 4[A1 12 B]
A
E2 l
K D 2 :
Point groups Dnd with n even, and Sm with 12 m even
(a) Point groups D2d and S4

K D C1 :
El
A
K D 1 :
ECl
A
K D 2 :
Qsub D 2[A1 C B]
Qsub D 4[A B]
(b) Point groups D4d and S8

K D C1 :
K D 1 :
E3 l
E3 Cl
A
A
K D C2 :
K D 2 :
E2 Cl
E2 l
A
A
Qsub D 2[A1 C B]
Qsub D 4[A1 12 B]
The states are characterized by their symmetry species under the operation Cn (or Sm ) only. Q sub is the
coefficient of K in the expressions for the sub-band origins. K D K0 K00 in the sense of the arrow.
(E species, vt D 1, all other v D 0) of a symmetric top molecule. The selection rules

depend on the point group to which a molecule belongs and are presented in Table 6.9.
Using these selection rules with eq. (6.7.32) we obtain the results given below for Q
branch transitions (J D 0), associated with various combinations of K and l.
The K D 1, l transitions arise for molecules belonging to the following point
groups: C3 (C3 , C3v , C3h , D3 , D3d , D3h , S6 ); Cn with n 4Cn , Cnv , Cnh , Dn , D2h );
Dnd with n odd; and Sm with 12 m odd. For these transitions
P,R
C,
QK
J; jQj D Q sub P,R QK
C f1 J, K
where the sub-band origins are
6.7.33
201
C,
0
0
Q sub P,R QK
D [Qvib C A0 1 2 B0 DK
] 2[A0 1 B0 2DK
]K
0
0
00
00
0
2
0
3
C [A B A B 6DK ]K 4DK K
0
00
DK
DK
K4
6.7.34
and where
0
00
K 12 C DJK
K2 ]JJ C 1
f1 J, K D [B0 B00 DJK
DJ0 DJ00 J2 J C 12
6.7.35
The K D 1, l transitions arise for molecules belonging to the point groups Dnd with
n even, and Sm with 12 m even. For these transitions
P,R
,C
,C
QK
J; jQj D Q sub P,R QK
C f1 J, K
6.7.36
where the sub-band origins are

,C
0
0
D [Qvib C A0 1 C 2 B0 DK
] 2[A0 1 C B0 2DK
]K
Q sub P,R QK
0
0
00
00
0
2
0
3
C [A B A B 6DK ]K 4DK K
0
00
DK
DK
K4
6.7.37
and where f1 J, K) is given by eq. (6.7.35).

The K D 2, l transitions arise for molecules belonging to the point groups Cn
with n 5 (Cn , Cnv , Cnh , Dn , Dnh , Dnd with n odd, and Sm with 12 m odd) and D4d and
S8 . For these transitions
O,S
C,
C,
QK
J; jQj D Q sub O,S QK
C f2 J, K
6.7.38
where the sub-band origins are given by

C,
0
0
Q sub O,S QK
D Q vib C 4[A0 1 B0 4DK
] 4[A0 1 12 B0 8DK
]K
0
C [A0 B0 A00 B00 24DK
]K2
0
0
00
8DK
K3 DK
DK
K4
6.7.39
and where
0
00
f2 J, K D [B0 B00 DJK
K 22 C DJK
K2 ]JJ C 1 DJ0 DJ00 J2 J C 12
6.7.40
The K D 2, l transitions arise for molecules belonging to the point groups C3 (C3 ,
C3v , C3h , D3 , D3d , D3h , S6 ). For these transitions
O,S
,C
,C
QK
J; jQj D Q sub O,S QK
C f2 J, K
6.7.41
where the sub-band origins are given by

,C
0
0
D Q vib C 4[A0 1 C B0 4DK
] 4[A0 1 C 12 B0 8DK
]K
Q sub O,S QK
0
0
C [A0 B0 A00 B00 24DK
]K2 8DK
K3
0
00
DK
DK
K4
and where f2 J, K is given by eq. (6.7.40).
6.7.42
202
The Raman Effect
For K D 2 transitions in molecules belonging to the point groups C4 (C4 , C4v ,

C4h , D4 , D4h ) and D2d and S4 the corresponding modes of vibration are antisymmetric
rather than degenerate with respect to the four-fold axis of symmetry. This case of a
non-totally symmetric, non-degenerate transition has been discussed above, and the band
wavenumbers are given by eqs. (6.7.26) to (6.7.28).
The wavenumber expressions for the remaining transitions with J D 1 and 2 are
not given here. They are easily derived from eq. (6.7.32) by use of the proper selection
rules. These transitions provide a fine structure in the degenerate band, with J D 1
transitions giving the P and R branches in which the line spacing is 2B, and the J D 2
transitions giving the O and S branches in which the line spacing is 4B, when the difference
between the upper- and lower-state B values and centrifugal stretching are ignored.
The formulae of eqs. (6.7.33) to (6.7.42) are relatively complicated and discussion of
the general forms of the Raman spectra is facilitated by introducing some simplifications.
If we ignore all centrifugal stretching terms, eqs. (6.7.35) and (6.7.40) no longer involve
K and we have
f1 J, K D f2 J, K D B0 B00 JJ C 1
6.7.43
If in addition we set A0 D A00 D A and B0 D B00 D B then eq. (6.7.34), which is associated
with K D 1, l transitions reduces to
C,
D Q vib C [A1 2 B] 2[A1 B]K
Q sub P,R QK
6.7.44
Also eq. (6.7.42), which is associated with K D 2, l transitions, simplifies to

C,
Q sub O,S QK
D Q vib C 4[A1 C B] 4[A1 C 12 B]K
6.7.45
For completeness we draw attention to the relation of eqs. (6.7.43) and (6.7.45) to
eqs. (6.7.28) and (6.7.31), which latter relate to K D 2 transitions in certain point
groups where the corresponding non-degenerate modes of vibration are antisymmetric
with respect to the four-fold axes of symmetry (see page 196). When D 0, eq. (6.7.45)
reduces to eq. (6.7.31); and when all centrifugal stretching terms are set equal to zero,
eq. (6.7.28) reduces to eq. (6.7.43). All the simplifications are summarized in Table 6.9.
It is evident from these simplified formulae for a prolate top that the spectra will consist
of a series of Q branches whose line spacing depends on K; and each branch will have
a rotational structure which depends on (B0 B00 ) and J. As (B0 B00 ) is usually small
the rotational structure may not be resolved. The form and position relative to Q vib of
these Q branches depends on the sign and magnitude of and the relative magnitudes of
A and B (for a prolate top A > B). Thus the separation of the P,R QK branches changes
by 2[A1 B] or by 2[A B] if D 0; and the separation of the O,S QK branches
changes by 4[A1 C 1/2 B] or 4[A B] if D 0. Also the R QK and S QK branches are
at higher wavenumbers and the P QK and O QK branches at lower wavenumbers relative to
Q vib if D 0. If is positive the branch separation of the P,R QK branches decreases relative
to the case of D 0 and can be very small. In an extreme case where A1 < B the
positions of these bands are reversed relative to Q vib . For the O,S QK branches the branch
separation decreases if is negative. The situation for an oblate top may be analyzed
203
similarly. It is necessary to replace A by C in eqs. (6.7.44) and (6.7.45) and to remember

that C < B.
The operation of different selection rules for infrared and Raman transitions enables
values of both A and to be obtained if both types of spectra are studied. Consider for
example a molecule belonging to the D3h point group. For the transition from A01 to the
degenerate state E0 the Raman selection rule is K D 2 and the spacing of the O,S QK
branches is 4[A1 C /2 B]. However, in the infrared the selection rule is K D 1
and the P,R QK branches have spacings of 2[A1 B]. Thus if these two spacings
can be measured both A0 and can be obtained provided that B is known.
6.7.4 Intensities
A general treatment of intensities of rotation and vibrationrotation Raman scattering has
been given in Section 6.3.3; the statistical distribution of the molecular population has
been developed in Section 6.5; and the intensity of rotation and vibrationrotation Raman
bands of diatomic molecules has been treated in some detail in Section 6.6.5.
It follows from these sections that the intensities of rotation and vibrationrotation
Raman transitions of symmetric top molecules are determined by the product of three
terms: Ni the number of molecules in the initial level, the appropriate Placzek-Teller b
factors, and the square of the appropriate transition polarizability component. We now
consider each of these in turn.
The population factor Ni is given in general by
Q v C FJ,
Q
NgN gJK exp fG
Khc0 /kTg
Ni D
6.7.46
Q
Q
gN gJK exp fGv C FJ,
Khc0 /kTg
v,J,K
Q v is the vibrational term value and FJ,

Q
Here G
K the rotational term value which
is given by eq. (6.7.22) or eq (6.7.32) if Coriolis interaction is significant. gJK is the
statistical weight of the initial state which is given by
gJK D 2J C 1 if K D 0
6.7.47
gJK D 22J C 1 if K 6D 0
6.7.48
gN is the nuclear spin factor which depends on the quantum number K, the value of
the nuclear spin, the spin statistics of the equivalent nuclei, the equilibrium molecular
symmetry and the symmetry of the vibrational state. Consider for example the case of a
non-planar XY3 molecule (or indeed any C3v molecule with only three identical atoms of
nuclear spin 12 outside the symmetry axis). Then, assuming the K doubling is not resolved,
for a non-degenerate vibrational state the rotational levels with K D 3q6D 0) have double
the statistical weight of the levels K D 3q 1, producing an alternation in weights of 2,
1, 1, 2, 1, 1, 2, . . . . For a degenerate vibrational state a similar alternation results for both
the Cl and l sublevels, but for the Cl sublevels the rotational levels with K D 3q C 1
and for the l sublevels the rotational levels with K D 3q 1 have the higher statistical
weight. If the identical atoms have zero nuclear spin, only certain states can exist; for nondegenerate vibrational states only those with K D 3q, and for degenerate vibrational states
204
The Raman Effect
only half of those with K D 3q 1. If the identical atoms have a nuclear spin of 1 there is
still an alternation of statistical weights but the ratio is reduced and approaches unity for
larger nuclear spins. These conclusions are specific to C3v molecules as defined above.
Nuclear spin statistics for symmetric top molecules in general have been treated in detail
in a series of papers by Weber (1980, 1981, 1982, 1988).
The Placzek-Teller factors bJf Kf ;Ji Ki involve both J and K. From Tables 6.2 and 6.8 we
see that for S branch lines the values of these factors decrease as K increases. For a prolate
symmetric top A > B and hence the exponential factor in eq. (6.7.46) also decreases as K
increases. Thus the maximum intensity will occur at low K values. However, for an oblate
symmetric top, C < B and the exponential factor in eq. (6.7.46) increases with increasing
K and the maximum intensity occurs at high K values. In either case the intensity maxima
for the various J transitions occur at different K values.
The transition polarizability components which can be involved depend on J and K.
For J D 0 and K D 0 there can be both isotropic and anisotropic contributions, namely
i
f 2 i
hvf j0
0 jv i and hv j0 jv i, respectively. For J D 1, 2 and K D 1, 2 there can
i
f 2 i
be only anisotropic contributions, namely hvf j2
1 jv i and hv j2 jv i The equilibrium
symmetry of the molecule and the symmetries of its vibrations determine whether a
transition polarizability component is zero or non-zero. These symmetry requirements
have already been discussed in Section 6.3.2. Their implementation requires reference to
Table 6.4 and Reference Table 5.3.
Formulae for intensities and polarization properties for rotation and vibrationrotation
transitions in symmetric top molecules are given in Reference Table 6.3 and some general
intensity and polarization formulae are given in Reference Table 6.4.
6.8 LINEAR MOLECULES

6.8.1 Rotation and Vibration-Rotation Raman Spectra
The results already obtained for diatomic molecules also apply to some, but not all, aspects
of linear molecules.
Thus the rotational term for a linear molecule in the ground vibrational state is given
by eq. (6.6.5). As the pure rotation selection rules for a linear molecule are also J D 0
and J D C2 it follows that its pure rotation Raman spectrum will consist of an S branch
for which the wavenumbers are given by eq. (6.6.6); and the Stokes and anti-Stokes pure
rotation Raman spectrum will have the pattern shown in Fig. 6.1(d).
In centrosymmetric molecules nuclear spin statistics will also play a role. For example,
as 16 O and 18 O have zero nuclear spins there are no rotational levels with odd J in the
centrosymmetric molecules C16 O2 or C18 O2 . Thus their pure rotational Raman spectra
will have lines with J even only and a line spacing of 8B.
Further, the rotational term associated with non-degenerate vibrations of a linear
molecule is given by eq. (6.6.19) and the selection rules are J D 0 and 2. Thus the
vibrationrotation Raman spectrum consists of S, O and Q branches the wavenumbers for
which are given by eqs. (6.6.20), (6.6.21) and (6.6.22), respectively, where all centrifugal
stretching terms are neglected but the dependence of B on the vibrational quantum number
205
has been retained. If the centrifugal stretching terms are retained then the wavenumbers
of the S, O and Q branches are given by eqs. (6.6.23), (6.6.24) and (6.6.25), respectively.
When degenerate vibrations are considered there is an additional angular momentum
about the internuclear axis associated with such vibrations in which atoms move perpendicular to the molecular axis. This angular momentum is given by lh, where for the kth
vibrational mode, l is a signed quantum number with values given by
6.8.1
l D vk , vk 2, vk 4, . . . vk
and vk is the vibrational quantum number of that mode. Thus for example when
vk D 1 as in fundamental vibrational bands, l D 1; when vk D 2, l D 0, 2; and when
vk D 3, l D 1, 3 and so on. Taking this additional factor into account the rotational
Q v J, l) is given by
term F
Q v J, l D Bv [JJ C 1 l2 ] Dv [JJ C 1 l2 ]2
F
6.8.2
Here J D jlj, jlj C 1, jlj C 2, . . . , so that the levels J D 0, 1. . . . jlj 1, do not occur. For
example, when vk D jlj D 1, the J D 0 level is missing and when vk D jlj D 2 the J D 0
and J D 1 levels are missing and so on.
Since the angular momentum thus generated is about the figure axis, a linear molecule
may be treated as a prolate symmetric top with K D jlj. In the absence of the quantum
number l, eq. (6.8.2) reduces to eq. (6.6.19) as expected. The general selection rules
applicable to eq. (6.8.2) are J D 0, 2 if l D 0, J D 0, 1, 2 if l 6D 0, and C 6! ,
s 6! a.
Q v J, l) is independent of the sign of l. However, this is only
According to eq. (6.8.2), F
true if Coriolis interaction is neglected. When this type of vibrationrotation interaction
is taken into account eq. (6.8.2) must be replaced by
q
Q v J, l D Bv [JJ C 1 l2 ] k vk C 1JJ C 1 Dv [JJ C 1 l2 ]2 6.8.3
F
4
Here qk is a parameter which depends on the coupling between the rotational and vibrational motions for the kth vibration. The effect of the term involving qk is to split the
rotational term values which differ only in the sign of l by an amount Q given by
qk
Q D vk C 1JJ C 1
6.8.4
2
When vk D 1, eq. (6.8.4) simplifies to Q D qk JJ C 1. This splitting is known as l-type
doubling. It decreases with increasing jlj and is appreciable only for jlj D 1. The parameter
qk is small and has values of the order of 104 to 105 cm1 .
The components of the doubled level are distinguished by their parity and are labelled C
or (or s or a); and allowed transitions are subject to the parity selection rules. Examples
of rotational levels with l-type doubling with their parity labels occur in Fig. 6.11.
Equation (6.8.2) may be used to obtain the following general expression for vibrationrotation Raman shifts for linear molecules:
0
0
jQj D Q vib C B0 [J0 J0 C 1 l 2 ] DJ0 [J0 J0 C 1 l 2 ]2
B00 [J00 J00 C 1 l002 ] C DJ00 [J00 J00 C 1 l002 ]2
See Page 198 of this Chapter.

See Page 166 in this Chapter.
6.8.5
206
The Raman Effect
For non-degenerate fundamental bands, representing stretching vibrations, l0 D l00 D 0.

For degenerate fundamental bands, representing bending vibrations jl0 j D 1, l00 D 0 and
each upper state rotational level is split into a doublet (l-type doubling) because of the
vibration-rotation interaction. We now illustrate the application of this general formula to
specific cases, treating first pure rotation and then vibration-rotation Raman spectra.
For pure rotational spectra in the ground vibration state the Raman shifts are obtained
from eq. (6.8.5) by setting Q vib D 0, l0 D l00 D 0, B0 D B00 D B, DJ0 D DJ00 D DJ where B
and DJ refer to the ground vibrational state and using the selection rule J D C2 so that
J0 D J00 C 2. This procedure yields eq. (6.6.6) which we have already indicated applies
to this case.
For pure rotational spectra within an excited non-degenerate vibrational state we set
Q vib D 0, l0 D l00 D 0, B0 D B00 D Bv , and DJ0 D DJ0 D DvJ where Bv and DvJ refer to the
excited vibrational state, and use the selection rule J D C2. The Raman shifts are given
by a modified form of eq. (6.6.6) in which Bv and DvJ replace B and DJ .
For pure rotational spectra within a degenerate vibrational state Q vib D 0, jl0 j D jl00 j 6D 1,
0
B D B00 D Bv , and DJ0 D DJ00 D DvJ , and the selection rule is now J D C1 and C2. Using
these conditions with eq. (6.8.5) yields the following formulae for the Raman shifts of
the R and S branches when the effects of l-type doubling are excluded:
6.8.6
RJ, l; jQj D 2Bv J C 1 4DvJ [J C 13 l2 J C 1]
and
SJ, l; jQj D 4Bv 6DJ J C 3/2 8DvJ [J C 3/23 l2 J C 3/2
6.8.7
For pure rotational spectra within a singly excited degenerate vibrational state we set
jl0 j D jl00 j D 1 in eqs. (6.8.6) and (6.8.7) and then we have
RJ; jQj D 2Bv J C 1 4DvJ JJ C 1J C 2
6.8.8
SJ; jQj D 4Bv J C 3/2 8DvJ J C 1J C 2J C 3/2
6.8.9
The intensity of the R branch varies as 1/J relative to the S branch so that the S branch
dominates the Raman spectrum.
We now consider vibration-rotation Raman spectra. For vibration-rotation spectra associated with a fundamental transition of a non-degenerate vibration, Q vib 6D 0, l0 D l00 D 0,
B0 6D B00 , and DJ0 6D DJ00 and the selection rule is J D 0, 2. Using these conditions
with eq. (6.8.5), the Raman shifts of the O, Q and S branches are found to be given by
eqs. (6.6.23) to (6.6.25) which we have already indicated apply to this case.
For vibration-rotation Raman spectra associated with a fundamental transition of a
degenerate vibration, Q vib 6D 0, jl0 j D 1, l00 D 0, B0 6D B00 , DJ0 6D DJ00 and the selection rule
is J D 0, 1, 2. Applying these conditions to eq. (6.8.5) yields the following formulae
for the Raman shifts of the O, P, Q, R and S branches:
For the Q branch
QJ; jQj D Q vib B0 DJ0 C B0 B0 C 2DJ0 JJ C 1 DJ0 DJ00 J2 J C 12
6.8.10
For the P and R branches
P, RJ; jQj D Q vib B0 DJ0 C B0 C B00 C 2DJ0 m C B0 B00 C DJ0 C DJ00 m2
6.8.11
207
where m D J for the P branch and J C 1 for the R branch.

For the O and S branches

1 0
1 0
1 0
00
00
0
00
00
O, SJ; jQj D Q vib B C 3B DJ 9DJ C B C B C DJ 3DJ m
4
16
2

1
1
B0 B00 7DJ0 11DJ00 m2
C
4
2
1 0
1
DJ C DJ00 m3 DJ0 DJ00 m4
6.8.12
2
16
where m D 2J C 1 for the O branch and 2J C 3 for the S branch.
The pure rotational Raman spectra of C16 O2 in the ground vibrational state and in the
vibrational state in which the degenerate bending vibration 2 has v D 1 serve to illustrate
some of the types of transitions considered above. The rotational levels of the ground
vibrational state of C16 O2 (symmetry C
g ) are shown at (b) in Fig. 6.11. As already
explained odd J levels are absent. The associated pure rotational Stokes and anti-Stokes
Raman spectrum (J D C2) is shown at (e) in Fig. 6.11 and this has lines with even J
only. The rotational levels of the vibrational state associated with the degenerate bending
vibration 2 with v2 D 1 (symmetry u ) are shown at (a) in Fig. 6.11. In this vibrational
state the l degeneracy is removed and each J level is split into a doublet, one member
of which has parity C and the other parity . As a consequence, although the rotational
levels with parity are absent in the ground vibrational state, transitions with both even
and odd J values can occur in this vibrational state. The Stokes and anti-Stokes rotational
Raman spectrum for this vibrational state arising from J D 1 is shown at (c) and that
arising from J D 2 at (d) in Fig. 6.11. In C16 O2 , Q 2 D 667 cm1 and because of this
relatively low value the population of the excited vibrational state in which v2 D 1 is
about 8% of the ground state population. Thus the observed Stokes and anti-Stokes pure
rotational Raman spectrum will consist of a superposition of the stronger ground state
spectrum (e) and the weaker excited vibrational state spectra (c) and (d). The difference
in the values of B0 and B1 is rather small so that transitions with J D 2 and J even in
both vibrational states will overlap. However, transitions with J D 1 and J odd only
arise in the excited vibrational state and so stand alone in the spectrum. The resulting
wavenumber pattern of the overall rotational Raman spectrum is shown at (f) in Fig. 6.11.
6.8.2 Intensities
The treatment of intensities of Raman bands of linear molecules is generally similar to
that given in Section 6.7.4 for symmetric top molecules. The population factor Ni is given
Q
by a modified form of eq. (6.7.46) in which FJ,
K) is replaced by the rotational term
Q
Q
appropriate to linear molecules, namely FJ)
for non-degenerate vibrations and FJ,
l)
Q
or FJ, l ) for degenerate vibrations. The appropriate Placzek-Teller b factors can be
obtained from Table 6.2 by replacing K by l. Then for non-degenerate vibrations l D 0
and for degenerate vibrations jlj 6D 0.
208
The Raman Effect

J =8
s
+a
+a
s
s
+a
+a u
s
s
+a
+a
s
s
+a
+a
s
5
4
3
2
1
(a)
J =8
+s
+s
+s
3
2
1
0
a
+s
a
+s
DG
(b)
(c)
(d)
(e)
J = 1
J=
7 6 5 4 3 2 1
1 2 3 4 5 6 7
J = 2
J= 6
J = 2
J= 6
4
4
2
2
1
0
4
4
6
6
+g
Total
(f)
n~ 1
Anti-Stokes
Stokes
Figure 6.11 Rotational energy levels and pure rotational Raman spectrum for C16 O2 .
(a) Rotational energy levels of the 011 0 vibrational state (u ) (b) Rotational energy levels
of the 000 0 vibrational state (C
g ) (c) Stokes and anti-Stokes rotational lines arising from
J D 1 in the 011 0 state. (d) Stokes and anti-Stokes rotational lines arising from J D 2 in
the 011 0 state. (e) Stokes and anti-Stokes rotational lines arising from J D 2 in the 000 0
state. (f) Overall Stokes and anti-Stokes rotational spectrum c C d C e.
6.9 CONTRIBUTIONS FROM ELECTRONIC ORBITAL AND SPIN

ANGULAR MOMENTA
In the treatment of a diatomic molecule given in Section 6.6 of this chapter we ignored
any contribution from the angular momenta of the electrons, either orbital or spin or both.
We now consider the contribution from the orbital angular momentum of the electron. A
diatomic molecule then becomes a prolate symmetric top with Ia < Ib D Ic and Ia 6D 0.
Of course, Ib D Ic is very much greater than Ia . Nevertheless, as the electrons rotate
209
much faster than the much heavier nuclei the angular momenta about the different axes
are similar in magnitude.
The rotational term for a diatomic molecule considered as a rigid symmetric top in
which the electronic angular momentum is coupled to the linear axis with component
along the axis of h, is given by
Q
6.9.1
FJ,
D BJJ C 1 C A B2
The rotational constants B and A are defined in eq. (6.7.1). Here they involve respectively
Ib , the moment of inertia of the molecule about an axis through its centre of gravity and
perpendicular to the internuclear axis, and Ia the moment of inertia of the electrons about
the internuclear axis. The quantum number relates to the angular momentum of the
electrons about the internuclear axis and
J D , C 1, C 2 . . .
6.9.2
Equation (6.9.1) is analogous to eq. (6.7.2) with replacing K and the condition on J
given by eq. (6.9.2) is analogous to that on K given by eq. (6.7.3). It is evident that
rotational levels with J < do not occur. is generally restricted to small integral
values.
It follows from eq. (6.9.1) that for each value of the energy levels are shifted
in magnitude by an amount (A B2 which is constant for a given electronic state.
However, the spacings between the energy levels remain the same as those for the simpler
cases based on eq. (6.6.1).
A more general form of eq. (6.9.1), which takes into account centrifugal stretching and
the dependence of B on the vibrational state is given by
Q v J, D Bv JJ C 1 C Av Bv 2 Dv2 J2 J C 12
6.9.3
F
which has some similarities to eq. (6.7.22). The selection rules are as follows:
when D 0 :
J D 0, 2
when 6D 0 :
J D 0, 1, 2
6.9.4
The permitted branches and the forms of the resulting rotation and vibrationrotation
Raman spectra follow those already deduced for the symmetric top in Section 6.7.
We consider next examples of molecules for which contributions from electron spin
angular momentum have to be taken into account. One such example is the oxygen
molecule whose ground electronic state is 3
g so that D 0 and the electron spin
quantum number S D 1. Coupling between the electronic spin angular momentum and
the overall molecular rotation causes a splitting of the rotational levels into triplets. As a
result its pure rotational Raman spectrum has an additional fine structure, the most significant feature of which is that each normal rotational line is accompanied by a pair of
so-called spin satellites, one at a slightly higher wavenumber and one at a slightly lower
wavenumber. The intensity of these spin satellites falls off rapidly with increasing rotational quantum number. We shall not develop a quantitative theory of such spectra here.
Another example is provided by the nitric oxide molecule which has a 2 ground
electronic state so that D 1 and S D 12 . Coupling between the electronic orbital angular
210
The Raman Effect
momentum and the angular momentum due to the spin of the uncoupled electron causes
a splitting of the ground electronic state into two states 2 1/2 and 2 3/2 which are
separated by about 120 cm1 . Each electronic state has its own set of vibrationrotation
levels and pure rotation spectra with both J D C1 (R branch) and J D C2 (S branch)
are allowed. However as the J values are half-integral, alternate lines of the R branches
do not coincide with the lines of the S branches. In addition, transitions between the two
electronic states 2 1/2 and 2 3/2 are allowed. A quantitative treatment of this situation
is not given here.
The examples considered above illustrate only some of the various types of coupling
known generally as Hunds coupling cases.
6.10 SPHERICAL TOP MOLECULES

A spherical top molecule has Ia D Ib D Ic and can also be regarded as a limiting case of
the symmetric top. The rotational term for a spherical top is then given by
Q
FJ
D BJJ C 1 DJ J2 J C 12
6.10.1
provided Coriolis coupling is ignored. In practice an additional centrifugal distortion term

which cannot be written in closed form should be included.
Such molecules have a ground state polarizability tensor which is spherically symmetric,
that is the diagonal tensor components xx , yy and zz are all equal and non-zero, and all
0
off-diagonal components are zero. Thus only 0
0 is non-zero and G 0 fi is non-zero if
J D 0, K D 0. Therefore no pure rotational Raman scattering will be exhibited by
spherically symmetric molecules, and Rayleigh scattering will be completely polarized.
However, vibrationrotation Raman spectra can arise for those vibrations of spherically symmetric molecules which have appropriate symmetry. For fundamental vibrational
transitions this requires a normal mode of vibration to belong to the same symmetry class
2
as one or more of the polarizability components 2
0 or K , K D 0, 1, 2. For
example the triply degenerate vibration of a tetrahedral molecule belongs to the symmetry
2
class F2 of the point group Td and the polarizability components 2
1 , 2 also belong
to this symmetry class. Thus the selection rules will be J D 0, 1, 2, K D 1, 2.
There is a 2J C 12 -fold degeneracy associated with each J level of a spherical rotor.
One 2J C 1 factor arises from the 2J C 1 values that M can take. The second 2J C 1
factor arises because, as a symmetric rotor becomes a spherical rotor, all levels with the
same J but different values of K become degenerate. Since K can take J C 1 values
and all levels with K D 0 are doubly degenerate, the second 2J C 1 factor results.
However, the triply degenerate modes of tetrahedral molecules are split by Coriolis
interaction and this results in three vibrational sublevels each of which has its own set of
rotational levels whose term values are given by
Q C J D Bv JJ C 1 C 2Bv k J C 1 DJ J2 J C 12
F
Q 0 J D Bv JJ C 1 DJ J2 J C 12
F
Q
J D Bv JJ C 1 2Bv k J DJ J J C 1
6.10.2
6.10.3
6.10.4
211
where Bv refers to the vibrational state v and k to the kth vibrational mode. There is
no selection rule for the three Coriolis sub-levels and thus there are 15 branches for an
F2 A1 transition, and five for each sub-band. Although these formulae are expected to
give a reasonably accurate representation of the rotational terms for heavy molecules they
have been shown to be insufficient to describe the spectrum of CH4 , for example.
Much more complete theoretical treatments have been developed both for rotational
terms and for intensities but they are not considered here.
6.11 ASYMMETRIC TOP MOLECULES

In asymmetric top molecules Ia 6D Ib 6D Ic and their rotational Raman spectra can exhibit
considerable complexity. Unlike the cases we have so far considered there are no
closed formulae for rotational terms which are generally applicable to asymmetric top
molecules, and straightforward equations characterizing transition wavenumbers are no
longer possible. However, rotational term values can be developed using asymmetry
parameters which are measures of the degree to which a molecule deviates from being
a symmetric top. Formulae based on such approximations can be useful especially when
the asymmetry is small and J values are low. However, computational methods have
largely replaced the approximation approach. The general procedure is to expand the
wave functions of the molecule in terms of an orthogonal set of functions, as for example
the symmetric top functions, and set up the secular equations for the unknown coefficients
and energies. The resulting secular determinant can be broken down into a number of subdeterminants the order of which increases with J. The solution of these sub-determinants
yields the energy levels and expansion coefficients. These procedures are quite standard
operations with modern computing facilities, even for large J. We shall not consider the
case of the asymmetric top any further here.
6.12 EPILOGUE
Rotation and vibration-rotation Raman spectroscopy has been widely applied to the fruitful
study of the structure of molecules. The resolution available has improved greatly over
Hg 2537
Figure 6.12 Pure rotation Raman spectrum of 14 N2 photographically recorded by Rasetti

(1930). The exciting radiation was from a mercury arc with wavelength 253.7 nm
(Q1 D 39,420 cm1 ). The relatively strong line to the right of the exciting line is a mercury
arc line of wavelength 253.48 nm superimposed on a rotational Raman line. Spectral range
200 cm1 .
212
The Raman Effect

Q branch
O branch
Intensity
S branch
8 14 20 2
2370
2330
2290
Wavenumber shift / cm1
Figure 6.13 Vibration-rotation Raman spectrum of 14 N2 obtained by Bendtsen (1974).

This shows the Q branch and parts of the O and S branches associated with the
vibrational transition v D 1
v D 0. The exciting radiation was from an argon ion laser,
Q1 D 20,487 cm1 (wavelength 488.0 nm).
80
Intensity/arbitrary units
70
60
50
40
30
20
10
2330.5
2330.0
2329.5
2329.0
2328.5
2328.0
2327.5
2327.0
2326.5
Wavenumber shift / cm1
Figure 6.14 Part of the fully resolved Q branch of 14 N2 obtained by Bendtsen and
Rasmussen (2000). It was recorded with a Fourier transform spectrometer with a resolution
of 0.015 cm1 .
213
the years and high resolution Raman spectroscopy of gases has now reached the stage
of Doppler-limited resolution. The spectra shown in Figures 6.12, 6.13 and 6.14, all of
which relate to 14 N2 , give an indication of the progress from 1930 to 2000. The intensity
alternation in the bands arises from the different nuclear spin statistical weights gN for J
odd and J even in 14 N2 (see Table 6.6).
Rotation and vibration-rotation Raman spectroscopy is also finding increasing
application in many diverse fields. Some examples are: the study of temperatures and
species concentration in combustion processes and in the earths upper atmosphere; the
investigation of gaseous species in gas bubbles taken from the Vostok ice core at a depth
of 3316 m which may provide information on the paleoclimatic composition of the earths
atmosphere over the last 100,000 to 250,000 years.
The reading list for this chapter in Further Reading provides many references to detailed
accounts of the applications of rotation and vibration-rotation Raman spectroscopy.
REFERENCES
Bendtsen, J. (1974). J. Raman Spectrosc. 2, 133.
Bendsten, J. and Rasmussen, F. (2000). J. Raman Spectrosc. 31, 433.
Herzberg, G. (1950). Molecular Spectra and Molecular Structure, I. Spectra of Diatomic Molecules,
D. van Nostrand: New York.
Hollas, J. M. (1998). High Resolution Spectroscopy, John Wiley: Chichester.
Hougen, J. T. (1962). J. Chem. Phys. 37, 1433.
King, G. W. (1964). Spectroscopy and Molecular Structure, Holt, Rinehart and Winston: New
York.
Mills, I. M. (1964) Molecular Physics 8, 363.
Placzek, G. and Teller, E. (1933). Zeitschrift fur Physik 81, 209.
Rasetti, F. (1930). Zeitschrift fur Physik 61, 598.
Weber, A. (1980). J. Chem. Phys. 73, 3952.
214
The Raman Effect

INTRODUCTION
Reference Tables 6.1 to 6.4, relate to the following cases only: illumination with
linearly polarized radiation such that Ey0 6D 0, Ex0 D 0, I D 12 c0 0 Ey20 and observation
of I/2; ?s , ?i , I/2; jjs , ?i and /2; ?i ; illumination with circularly polarized
R s,
R i D I0;
L s,
L i ,
radiation I D 12 c0 0 [Ex20 C Ey20 ] and observation of I0;
L s,
R i D I0;
R s,
L i ) and P 0 D P 1 . The scattering system is an
I0;
assembly of N randomly oriented molecules, which are non-absorbing and non-chiral.
For a general Stokes transition from a lower state defined by the set of vibration and
rotation quantum numbers v00 R00 to an upper state defined by the set of vibration and
rotation quantum numbers v0 R0 , we may write
I
; ps , pi D kQ Q1 Q v0 ,R0 :v00 ,R00 4 gN00 a2 , 2 ,
I
and for 0 the corresponding differential scattering cross-section per molecule
I
; ps , pi
NI
For randomly oriented molecules, transitions between M levels (degenerate in the absence
of fields) are taken care of by isotropic averaging; thus the quantum numbers M can be
excluded from the set of rotation quantum numbers R. In these formulae Q v0 R0 :v00 R00 is the
wavenumber associated with the transition v0 R0
v00 R00 , N00 is the number of molecules in
the initial (lower) state. (Here, following accepted convention, the upper level is indicated
by 0 and the lower level by 00 and hence Q v0 ,R0 :v00 ,R00 is always positive.) For pure rotational
transitions g, the statistical weight of the initial (lower) state, is the product of the nuclear
spin degeneracy gN and the rotational degeneracy gR ; for vibrationrotation transitions
vibrational degeneracy must also be taken into account.
For anti-Stokes transitions it is necessary to replace (Q1 Q v0 ,R0 :v00 ,R00 4 by (Q1 C
Q v0 ,R0 :v00 ,R00 4 , and N00 by N0 since the initial state is now the upper state. Also g, gN
and gR now refer to the upper (initial) state.
For a diatomic molecule in the ground electronic state, assuming zero angular
momentum of the electrons about the bond axis ( D 0) and zero resultant electron spin
(S D 0) only one vibrational quantum number v and one rotational quantum number J are
needed to define the energy for a system of randomly oriented molecules. The isotropic
averages of the relevant squares of the transition polarizability tensor components are
given in Reference Table 6.1 for permitted rotation and fundamental vibrationrotation
transitions of a diatomic molecule in the approximation of the polarizability theory. These
isotropic averages are functions of a20 and 20 or a0 2 and 0 2 , and J-dependent
Placzek-Teller b2 factors which determine the anisotropic contribution.
Values of the function a2 , 2 ,
for calculating scattered intensities, depolarization
ratios, and reversal coefficients for permitted rotation and vibrationrotation transitions
0
; ps , pi D
215
in a diatomic molecule are given in Reference Table 6.2. The functions a2 , 2 ,

were obtained by using the appropriate isotropic averages of squares of the transition
polarizability tensor components from Reference Table 6.1. For example I/2; jjs , ?i
for the pure rotation transition v, J C 2
v, J was obtained using hxy 2v;JC2:vJ i from
Reference Table 6.1. The functions in Reference Table 6.2 also apply to the pure
rotation transitions of linear molecules and to the vibrationrotation transitions of linear
molecules in which the vibration is totally symmetric. For the vibrationrotation transitions
the factors v C 1bv2 v D C1 and vbv2 v D 1 which have been omitted from this
Table must be included.
For other rotors and vibrators, the functions a2 , 2 ,
) and the formulae for
h/2; ?i i and P 0 D P 1 , follow the same general pattern, differing essentially
only in the Placzek-Teller b2 factors. These determine the anisotropic contributions
and depend on the rotation quantum numbers and the permitted transitions. This is
illustrated by Reference Table 6.3 which gives the function a2 , 2 ,
for intensities and
formulae for depolarization ratios, and reversal coefficients for pure rotational scattering
from a symmetric top molecule. The rotational energy of a symmetric top molecule is
determined by two quantum numbers J and K, and the anisotropic contribution to the
scattering is determined by the Placzek-Teller b2 factors bJ0 ,K00 :J00 ,K00 since the selection
rule K D 0 operates for pure rotation transitions. Reference Table 6.3 also applies to
vibrationrotation transitions of symmetric top molecules in which the vibration is totally
symmetric (K D 0 also applies), if a20 is replaced by a0 2 and 20 by 0 2 , and
v C 1bv2 or vbv2 included.
Generalization to cover most cases of rotation and vibrationrotation transitions is
possible. Thus Reference Table 6.4 gives the general forms of the function a2 , 2 ,

and the formulae for /2; ?i ) and P 0 D P 1 ; and Table 6.2 gives the general
forms of the Placzek-Teller b2 factors. These determine the anisotropic contributions.
Using the selection rules summarized in Table 6.5 formulae for intensities and polarization
properties may be obtained for permitted rotation and vibrationrotation transitions in
linear, symmetric top and spherical top molecules.
Finally, the nuclear spin and rotational degeneracy factors: for heteronuclear diatomic
molecules gN D 1; but for homonuclear diatomic molecules gN depends on the nuclear
spin of the equivalent nuclei and can be different according as J is odd or even. Some
specific cases are considered in Table 6.6. For the nuclear spin degeneracy of polyatomic molecules see pages 203204. For linear molecules gR is (2J C 1); for symmetric
top molecules gR is 2J C 1 for K D 0 and 22J C 1) for K 6D 0; and for spherical top
molecules gR is 2J C 12 . Both gN and gR refer to the initial state.
v, J
v, J
v C 1, Jb
v C 1, J 2b
JJ C 1
2J 12J C 3
4 2 2
b
45 J,J 0
4 2 0 2
b
45 J,J
2
D
bJ2,J
1 2
0 2
b
15 J2,J
1 2 0 2
b
15 J,J
1 2
2
b
15 JC2,J 0
1 2 2
b
15 J,J 0
hxy 2v0 J0 :v00 J00 i
3J C 1J C 2
22J C 12J C 3
4 2 0 2
b
45 J,J
a0 2 C
4 2
2
b
45 JC2,J 0
a20 C
hxx 2v0 ,J0 :v00 J00 i
2
D
bJC2,J
v D C1, J D 2
v D C1, J D 0
v D 0, J D C2
v D 0, J D 0
Selection rulesa
Assumes zero angular momentum of electrons about bond axis D 0 and zero resultant electron spin S D 0.
Assumes applicability of Placzek polarizability theory; factor v C 1b2v has been omitted.
c
A contracted form of the Placzek Teller b factor is used here and in Reference Table 6.2 since K D 0.
v, J
v, J C 2
c 2
bJ,J
v, J
Lower state v00 J00
v, J
Upper state v0 J0
2 2 2
b
45 J,J 0
2 2 0 2
b
45 J,J
3JJ 1
22J 12J C 1
2 2
0 2
b
45 J2,J
a0 2
2 2
2
b
45 JC2,J 0
a20
hxx v0 ,J0 :v00 J00 yy v0 J0 :v00 J00 i
Reference Table 6.1 Isotropic averages of squares of transition polarizability tensor components for diatomic molecules with D 0,
S D 0 for rotation and vibration-rotation transitions.
216
The Raman Effect
S
C2
2
C2
4 2 0 2
b
45 J,J
4 2
b
0 2
45 JC2,J
a0 2 C
O
y
1 2
b
0 2
15 JC2,J
1 2 0 2
b
15 J,J
1 2
b
0 2
15 J2,J
4 2
b
0 2
45 J2,J
1 2 2
b
15 J,J 0
1 2
b
2
15 JC2,J 0
4 2 2
b
45 J,J 0
I/2; jjs , ?i
4 2
b
2
45 JC2,J 0
a20 C
I/2; ?s , ?i
a2 , 2 ,
2
3
4
2
45a0 2 C 4bJ,J
0 2
2
3bJ,J 0 2
3
4
3
4
45a20 C 4bJ,J 20
2
3bJ,J 20
/2; ?i
45
0 2
2
20
bJ,J 0 2
45
2
bJC2,J
a0 2 C
45
bJ2,J
2
45
20
2
bJ,J 2
0
45
2
bJC2,J
a20 C
L s,
L i
I0;
R s,
R i
I0;
15
2
2bJC2,J
2
20
20
0 2
2bJ,J 0 2
15
15
2
2bJ2,J
15
2
2bJC2,J
2
2bJ,J 20
15
R s,
L i
D I0;
L s,
R i
I0;
a2 , 2 ,
2
45a0 2 C bJ,J
0 2
2
6bJ,J 0 2
2
45a20 C bJ,J 20
2
6bJ,J 20
P 0 D P 1
2
2
The factors v C 1b2v (for v D C1 and vb2v (for v D 1 have been omitted. For definitions of bJ,J
and bJ2,J
see Reference Table 6.1.
Ramana
(vibration
rotation)
Raman
(rotation)
v J
Branch Selection
rules
Rayleigh
Type of
scattering
s R i
s L i
s R i
sL i
R ,
L ,
L ,
R
Reference Table 6.2 a2 , 2 ,
for I/2; ?s , ?i , I/2; jjs , ?i , I0;
D I0;
, I0;
D I0;
,
i
1
/2; ? and P 0 D P for diatomic molecules.

217
2
2
bJC1,K:J,K D
3
4
3
4
45a20 C 4 20
2
3bJ,K:J,K 20
/2; ?i
3K2 [J C 12 K2 ]

JJ C 1J C 22J C 1
1 2
b
2
15 JC2,K:J,K 0
4 2
b
2
45 JC2,K:J,K 0
1 2
b
2
15 J,K:J,K 0
1 2
b
2
15 JC1,K:J,K 0
4 2
b
2
45 J,K:J,K 0
4 2
b
2
45 JC1,K:J,K 0
a20 C
I/2; jjs , ?i
a2 , 2 ,

I/2; ?s , ?i
[JJ C 1 3K2 ]2
JJ C 12J 12J C 3
C2
bJ,K:J,K D
C1
Raman
(rotation)
v J K
Selection
rules
Branch
Rayleigh
Type of
scattering
bJC2,K:J,K D
2
2
45a20 C bJ,k:J,k 20
2
6bJ,K:J,K 20
P 0 D P 1
3[J C 12 K2 ][J C 22 K2 ]

2J C 1J C 22J C 12J C 3
15
45
2
2bJC2,K:J,K 20
bJC2,K:J,K 20
2
15
45
2
2
2bJ,K:J,K 20
15
2bJC1,K:J,K 20
2
bJ,K:J,K 20
45
R s,
L i
D I0;
L s,
R i
I0;
bJC1,K:J,K 20
2
a20 C
L s,
L i
D I0;
R s,
R i
I0;
a2 , 2 ,
s R i
s L i
s R i
s L i
R ,
L ,
L ,
R ,
Reference Table 6.3 a2 , 2 ,
for I/2; ?s , ?i , I/2; jjs , ?i , I0;
D I0;
, I0;
D I0;
;
i
1
/2; ? and P 0 D P for rotation transitions in symmetric top molecules.
218
The Raman Effect
R 6D 0
R D 0
R 6D 0
4 2
b i i 20
45 R R
4 2
b i i 0 2
45 R R
4 2
b
0 2
45 Rf Ri
a0 2 C
4 2
b
2
45 Rf Ri 0
a20 C
I/2; ?s , ?i
1 2
b
0 2
15 Rf Ri
1 2
b i i 0 2
15 R R
1 2
b
2
15 Rf Ri 0
1 2
b i i 20
15 R R
I/2; jjs , ?i
a2 , 2 ,
3
4
45a0 2 C 4bRi Ri 0 2
2
3bR2i Ri 0 2
3
4
45a20 C 4bRi Ri 20
2
3bR2i Ri 20
/2; ?i
15
45
2bRf Ri 0 2
15
45
2
2bR2i Ri 0 2
20
bR2i Ri 0 2
15
2bR2f Ri
15
2bR2i Ri 20
L s,
R i
I0;
i
R ,
L s
D I0;
bRf Ri 0 2
2
a0 2 C
45
20
20
45
bR2i Ri
bR2f Ri
a0 2 C
R s,
R i
I0;
s
L ,
L i
D I0;
a2 , 2 ,
The factors v C 1b2v (for v D C1 and vb2v (for v D 1 have been omitted. For definitions of bR2f Ri see Table 6.2.
v D 1
v D 0
R D 0
Selection rules
2
45a0 2 C bRi Ri 0 2
6bR2i Ri 0 2
2
45a20 C bRi Ri 20
6bR2i Ri 0 2
P 0 D P 1
s R i
s L i
R ,
L ,
Reference
Table 6.4 General
forms
of
a2 , 2 ,

for
I/2; ?s , ?i , I/2; jjs , ?i , I0;
D I0;
,
s R i
s L i
L ,
R ,
I0;
D I0;
, /2; ?i , P 0 D P 1 for rotation and vibrationrotation transitions.

219
221
7
Vibrational Resonance Raman
Scattering
Time comes from the future, which does not
yet exist, into the present which has no duration,
and goes into the past which has ceased to exist.
St Augustine
7.1 INTRODUCTION
Vibrational resonance Raman scattering is a very extensive subject. The theoretical treatment necessarily involves consideration of the vibrational levels not only of the initial
electronic state, which throughout this chapter is taken to be the ground state, but also
of one or more excited electronic states. Consequently the formulae are, in general, both
complicated and cumbersome; and the key principles involved are easily obscured. We
shall therefore only consider in detail the more simple cases so that the essential features
can be clearly established. More complicated cases will be treated only qualitatively or
semi-quantitatively.
Vibrational resonance Raman scattering can be divided into two categories: discrete
resonance Raman scattering, in which the frequency of the incident radiation is in the
region of discrete vibronic transitions; and continuum resonance Raman scattering, where
the incident frequency is above the excited state dissociation limit in a continuum or
coincides with the continuum of a purely repulsive state. These two categories of Raman
scattering have considerably different properties and are treated separately. They are
compared with normal Raman scattering in Fig. 7.1.
We begin with discrete resonance Raman scattering which is treated in Sections 7.2
to 7.8. Two different theoretical approaches are used. The first, which we consider
222
The Raman Effect
(a)
(b)
(c)
(d)
Figure 7.1 Comparison of (a) normal Raman scattering, (b) discrete resonance Raman
scattering, (c) continuum resonance Raman scattering arising from the continuum above the
dissociation limit of a bound state, (d) continuum resonance Raman scattering arising from
the continuum of a purely repulsive state.
in Sections 7.2 to 7.7, is based on formulae derived using perturbation theory. These
formulae describe steady state scattering processes and contain no explicit reference to
time. The second, which we consider in Section 7.8, involves a time-dependent formulation and provides an alternative description of resonance Raman scattering. It will emerge
from these treatments that vibrational resonance Raman scattering is a powerful method
for the exploration of excited electronic states of molecules and can yield information not
accessible by other means.
7.2 VIBRATIONAL TRANSITION POLARIZABILITY TENSOR

COMPONENTS IN THE RESONANCE CASE, BASED ON
PERTURBATION THEORY
The key formulae for vibrational resonance Raman scattering were developed in Chapter 4,
Section 4.12, using perturbation theory. They are reproduced here for convenience. The
component of the vibrational transition polarizability tensor in the resonance case is
given by
eg vf :eg vi D AVI C BVI C CVI C DVI
7.2.1
where
rr
hv
1
jvrr
k ihvk jvk i
k
AVI D p 0eg er p 0er eg
r r
h
r er vr :eg vi 1 ie vk
fg
vk
ig
7.2.2
The condensed nomenclature involving A, B, C and D terms is introduced in Chapter 4, Section 4.8.1
and developed in subsequent sections of that chapter.
223
rr
hv
heks er
jQk jvrr
1
0
0
k ihvk jvk i
k
D 2 p eg es
p er eg
er es
h
vr
k
k
fg
VI
ig
rr
hv
1
hekr es
jvrr
0
0
k ihvk jQk jvk i
k
p
p

g
r
s
g

e e
ee
er es
h 2
vr
k
k
fg
ig
7.2.3
rr
hv
1 hekg et
jQk jvrr
0
0
k ihvk jvk i
k
p

p

r
g
t
r
ee
ee
h 2 eg et
vr
k
k
fg
CVI D
ig
fg
ig
rr
hekt eg hvk jvrr
1
k ihvk jQk jvk i
0
0
p

p

eg er
er et
eg et r er vr :eg vg 1 ier vrk
h 2
v
k
k
7.2.4
VI
ig
hvfg jQk jvrr ihvrr
1
0
0
k
k
k 0 jQk 0 jvk 0 i
D 3 p eg es
p es0 eg
er es er es0
h
vr ,vr
k
k
heks er hekr es0
k0
7.2.5
0
In these equations the exclusions er 6D eg , es , es 6D er and et 6D eg apply, and an additional superscript has been introduced on the vibrational quantum numbers which appear
in the numerators to emphasize to which electronic state they belong. Thus, the symbols
ig
fg
vk and vk indicate that the initial and final state vibrational quantum numbers vik and
f
rr
vk respectively relate to the ground electronic state jeg i. Similarly, symbols such as vk
rs
and vk indicate that the intermediate state vibrational quantum numbers vrk relate to the
intermediate electronic states jer i and jes i respectively with the exclusion that r, s 6D g.
Electronic states of vibrational quantum numbers appearing as subscripts on frequency
terms in denominators are normally evident and further labelling is not usually needed.
The exclusions given here apply throughout this chapter.
In obtaining eqs. (7.2.2) to (7.2.5) a number of conditions appertaining to the resonance
situation were introduced and we now review these briefly. As 1 approaches a particular
vibronic absorption frequency er vr :eg vg , the excited electronic state jer i will dominate the
k
k
sum over states, and thus it is normally adequate to consider one, or at the most two,
resonant electronic manifolds. Also, because of their relative smallness, terms with nonresonant denominators can be neglected. Further, we can restrict the sums over vibrational
states to one sum over the states jvrr
k i; and we need consider coupling of the resonant
electronic state jer i to only one excited electronic state, jes i in the BVI terms and jet i in
0
the CVI terms, and only two excited states jes i and jes i (where s may be equal to s0 ) in
the DVI term. Similarly, we can restrict the summation to a single normal coordinate Qk
in the BVI and CVI terms and a pair of normal coordinates Qk and Qk 0 (where k and k 0
may be the same) in the DVI term.
In the next section we examine the nature of the four terms AVI , BVI , CVI and DVI
and assess their relative importance. In Sections 7.3 to 7.7 we consider in more detail a
number of special cases of discrete vibrational resonance Raman scattering.
224
The Raman Effect
Continuum resonance Raman scattering is treated in Section 7.9 using both perturbation
theory and the time-dependent formulation.
7.3 COMPARISON OF THE AVI , B VI , C VI AND D VI TERMS

7.3.1 The AVI term
VI
The AVI term, for which the full notation is A eg vt :eg vi , is defined in eq. (7.2.2). For
this term to be non-zero, two conditions must be met. Both the transition dipole moments
p 0eg er and p 0er eg which appear as a product must be non-zero; and the product of
fg
rr ig
the vibrational overlap integrals, hvk jvrr
k ihvk jvk i, must also be non-zero for at least
value.
one vrr
k
The first condition is straightforward. It simply requires the resonant electronic transition jer i
jeg i to be electric-dipole allowed. For strong scattering the transition dipole
moments involved should have appreciable magnitudes. Thus excitation within the contour
of an intense absorption band, as for example a band arising from a charge transfer mechanism or a transition, would be favourable. Excitation within the contour of a weak
band, like one resulting from a ligand-field or spin-forbidden transition, would not be
likely to produce a significant A-term.
We now consider the second condition, restricting the discussion to the case of harmonic
potential functions. If the vibrational wave functions are all orthogonal, as will be the case
if, for each vibrational mode, its vibrational manifolds in the two electronic states jeg i and
fg
rr ig
jer i are identical, the vibration overlap integrals hvk jvrr
k i and hvk jvk i are zero unless
fg
ig
rr
rr
vk D vk and vk D vk and there is therefore no contribution from the AVI term to
Raman scattering. However, if non-orthogonal vibrational wave functions are involved
the AVI term can be non-zero. For any one vibrational mode of a molecule, say the kth
vibrational mode, non-orthogonality can arise in two ways. The first of these is if for the
kth mode its classical vibration frequency is different in the electronic states jeg i and jer i,
g
that is kr 6D k . Such an inequality will arise if the shape of the potential energy surface
is different in the two electronic states. The second source of non-orthogonality is if there
is a displacement Qk of the potential energy minimum along the normal coordinate
Qk as between the electronic states jeg i and jer i. Symmetry considerations require that
such a displacement can occur only for totally symmetric modes unless the molecular
symmetry is changed in the excited state. If the molecular symmetry is the same in the
ground and excited states, then geometric changes are confined to combinations of bond
length and bond angle changes that, overall, do not alter the molecular symmetry. This
means that the geometric changes are confined to those along totally symmetric symmetry
coordinates, which necessarily relate only to displacements along totally symmetric normal
coordinates.
In this and subsequent sections we shall use the full notation for the AVI , BVI , CVI and DVI terms when it
is expedient to show the components and transitions involved.
Chapter 5, Section 5.9 discusses vibrational frequencies.
Chapter 5, Section 5.8 discusses symmetry and spectroscopic activity.
225
|er
|eg
Q
(a)
=
Qk = 0
wrk
wgk
Q
(b)
wgk
Qk = 0
wrk
|er
Qk
Qk
|eg
Q
(c)
Qk 0
wrk = wgk
Q
(d)
Qk 0
wrk wgk
Figure 7.2 Diagrams of the potential energy V associated with the kth vibrational mode
(normal coordinate Qk ) for the ground electronic state jeg i and the excited electronic state
jer i for four cases (a) (d), defined in outline above and in more detail in the text.
The two sources of non-orthogonality can operate separately or together. Thus we

can distinguish four different situations relating to the potential energy curves for the
ground state jeg i and an excited electronic state jer i. These are illustrated in Fig. 7.2. In
case (a) the orthogonality of the vibrational wave functions would not be removed and
there would be no AVI term for any vibrational mode. In case (b) where Qk D 0 nonzero vibrational overlap integrals would result for both totally and non-totally symmetric
modes, whereas in cases (c) and (d) where Qk 6D 0 non-zero vibrational overlap integrals
would only arise for totally symmetric modes. In practice a significant change of classical
vibration frequency usually only occurs when there is also a displacement of the potential
energy minimum, hence case (d) is the most important.
The roles of the vibrational overlap integrals and the dipole electronic transitions in
determining whether the AVI term is zero or non-zero is illustrated in Fig. 7.3 for two
g
of the cases, (a) and (c). If kr D k and Qk D 0 then, unless vfg D vig , there is
always one vibrational overlap integral which is zero in the products of the two such
integrals in the numerators of the terms in AVI . Hence in this instance (case (a) in
g
Fig. 7.2) the AVI term only contributes to Rayleigh scattering. If however kr D k and
fg
ig
D v C 1 are allowed, and
Qk 6D 0 (case (c) in Fig. 7.2) transitions for which v
226
The Raman Effect

g
wkr = wk ; Q k = 0 (Case (a) in Fig. 7.2)

Vibrational part
r (r )
k
=1
r (r )
k
=0
Electronic part
r
(i)
(p)0ereg
(p)0eger
f (g)
k
=1
i(g)
k
=0
(ii)
(iii)
(iv)
(i)
r (r) = 0| i (g) = 0 0
k
k
(ii) kf (g) = 1| kr(r) = 0 = 0
(iii) kr (r) = 1| ki (g) = 0 = 0
(iv) kf (g) =1| kr (r) = 1 0
eg
wkr = wkg; Q k 0 (Case (c) in Fig. 7.2)

Vibrational part
r (r )
k
=1
r (r )
k
=0
Electronic part
er
(i)
(ii)
(iii)
(iv)
Qk
(p)0ereg
eg
(p)0eger
f (g)
k
=1
i (g)
k
=0
(i)
r (r) = 0| i (g) = 0 0
k
k
(ii) kf(g) = 1| kr(r) = 0 0
(iii) kr (r) = 1| ki (g) = 0 0
(iv) kf(g) = 1| kr(r) = 1 0
Figure 7.3 The dipole electronic transitions and vibrational overlap integrals involved in
AVI for cases (a) and (c) defined in Fig. 7.2.
227
Vibrational part
Electronic part
r (r )
k
=2
r (r )
k
r (r )
k
=1
=0
er
Qk
(ps)0ereg
(i)
(ii)
(pr)0eger
eg
i(g)
k
=2
i(g)
k
=1
i(g)
k
=0
fg
Figure 7.4 How an overtone transition jvk

term when Qk is large.
D 2i
(i) kr(r) = 2| ki(g) = 0 0

(ii) kf(g) = 2| kr(r) = 2 0
fg
jvk
D 0i can arise from the AVI
then the AVI term contributes to Raman scattering arising from fundamental transitions.
However, if the magnitude of Qk is sufficient, the product of the vibrational overlap
fg
ig
integrals hvk jvrr
ihvrr
not only for transitions
k fg
k jvk i may have significant magnitudes
ig
ig
fg
D 1 but also for transitions from vk D 0 to vk
D 2, 3, 4 . . . .
from vk D 0 to vk
As a result, the AVI term can give rise to overtones in the resonance Raman spectrum with
intensities comparable to that of the fundamental. This situation is illustrated in Fig. 7.4,
fg
ig
for the first overtone jvk D 2i
jvk D 0i.
In cases where a change of molecular symmetry accompanies the electronic transition,
the requirement that only displacements along totally symmetric normal coordinates are
allowed no longer applies. The totally symmetric condition now applies not to the groundstate point group but to the sub-group formed by the operations that are common to both
the ground- and excited-state point groups. An important case of change of molecular
symmetry accompanying an electronic transition arises through the JahnTeller effect.
For example, a non-linear molecule in an electronically degenerate state always distorts
in the excited electronic state into a less symmetric but more stable configuration in which
the electronic degeneracy has been removed. Such distortions are manifested in resonance
Raman spectra, and the nature of the vibrational resonance Raman spectrum depends on
the magnitude of the JahnTeller distortion.
7.3.2 The B VI term

The BVI term as defined in eq. (7.2.3) involves the vibronic coupling of the resonant
excited state jer i to one other excited state jes i. In principle, coupling to more than one
228
The Raman Effect
excited state could be involved but this is rarely necessary and is not considered here.
The magnitude of the BVI term is determined by a number of factors which we now
examine.
We consider first the vibronic coupling integrals which for convenience we now write
as heks er /er es and hekr es /er es where er es D er es . These relate to the vibronic
coupling of two excited electronic states jer i and jes i, and we recall from Chapter 4,
Section 4.8, that the vibronic coupling integral heks er is given by
O e /Qk 0 jer i
heks er D hes jH
7.3.1
and that for real wave functions hekr es D heks er .

The term heks er /er es is a measure of the strength of vibronic coupling of the states jer i
and jes i via the normal coordinate Qk . It should be appreciated that the HerzbergTeller
expansion is only valid for weak vibronic coupling, i.e. within the framework of the
adiabatic BornOppenheimer approximation. It is not applicable to cases in which there
is strong vibronic coupling, as when es er is of the order of vibrational frequencies (in
which case non-adiabatic coupling is important) or zero (such as for a degenerate state,
in which JahnTeller coupling occurs). Of course, when the HerzbergTeller expansion
is not valid, A- and B-term mechanisms cannot be considered separately.
For the integral heks er to be non-vanishing, the irreducible representation of the vibrational fundamental with normal coordinate Qk must be contained in the direct product
of the irreducible representations of the states jer i and jes i. Thus the BVI term can be
non-zero for both totally symmetric and non-totally symmetric fundamentals. However,
for a totally symmetric mode the states jer i and jes i must have the same symmetry. Now
electronic states of the same symmetry are rarely close together in energy and so the BVI
term will usually be unimportant as far as totally symmetric modes are concerned because
a large value of the frequency denominator er es is likely to be involved. Even when
this frequency difference is small the heks er factor will not have an appreciable magnitude.
We consider next the products of electronic transition dipole moments p 0eg es p 0er eg
and p 0eg er p 0es eg . For these products to be non-zero both the transitions jer i
jeg i
s
g
and je i
je i must be electric dipole allowed.
Finally we consider the products of vibrational transition integrals and overlap integrals,
fg
ig
fg rr
rr ig
jvk ihvrr
namely hvk jQk jvrr
k ihvk jvk i and hvk
k jQk jvk i. In the harmonic approxfg
ig
fg
rr
D vrr
imation for Qk , hvk jQk jvrr
k i is non-zero only if vk
k 1 and hvk jQk jvk i is
non-zero only if vrr
D vig 1. If we restrict ourselves to diagonal vibrational overlap
k
ig
fg
D vk and vk D vrr
integrals, that is those for which vrr
k
k , products of the vibrational
fg
ig
transition integrals and overlap integrals are non-zero only if vk D vk 1. If upper
vibrational levels of the kth mode are not populated, a situation often described as the
ig
ig
fg
fg
low temperature limit, then vk D 0k and vk D 1k so that only two such products
contribute. If we introduce the values of the vibrational quantum numbers explicitly and
also give them electronic state superscript labels, the two contributing products take the
fg
fg rr
ig
rr ig
VI
forms h1k jQk j0rr
j1k ih1rr
k ih0k j0k i and h1k
k jQk j0k i. The expression for B
229
now assumes the following special form which contains only two terms:
fg

rr ig
1
heks er
h1k jQk j0rr
BVI
0
0
k ih0k j0k i
p er eg
eg 1f :eg 0i D 2 p eg es
k
k
er es
er 0r :eg 0i 1 ier 0rk
h
fgk rrk rr

ig
1
hekr es
h1k j1k ih1k jQk j0k i
0
0
C 2 p eg er
p es eg
7.3.2
er es
er 1r :eg 0i 1 ier 1rk
h
k
where we have reverted to the full nomenclature. It is clear from these considerations that
BVI term scattering occurs only for fundamentals.
VI
The nature of the first term in B eg 1f :eg 0i defined in eq. (7.3.2) is illustrated in
k
k
Fig. 7.5(a). The electronic part of this diagram shows the transition electric dipole
p 0er eg , the vibronic coupling heks er /er es of the excited electronic state jer i to the
excited electronic state jes i and the transition electric dipole p 0eg es . The HerzbergTeller
coupling is active in the emission process, allowing participation of the transition electric
dipole p 0eg es . The vibrational part of the diagram shows the vibrational overlap integral
ig
fg
h0rr
jQk j0rr
k j0k i and the vibrational transition integral h1k
k i. It can be seen that the
ig
rr
r
jeg ij0k i which leads to
vibronic transition involved in the resonance is je ij0k i
a frequency denominator of er 0r :eg 0i 1 ier 0rk . This resonance is often referred to
k
k
simply as a 00 resonance.
VI
The nature of the second term in B eg 1f :eg 0i defined in eq. (7.3.2) is illustrated simik
k
larly in Fig. 7.5(b). The vibronic transition involved in the resonance is now
ig
jeg ij0k i which leads to a frequency denominator of er 1r :eg 0i 1 ier 1rk .
jer ij1rr
k i
k
k
This is often referred to simply as a 10 resonance. The HerzbergTeller coupling is now
active in the absorption process allowing participation of the transition electric dipole
p 0es eg .
7.3.3 The C VI term

The CVI term which is defined in eq. (7.2.4) relates to vibronic coupling of the ground
electronic state jeg i to an excited electronic state jet i. This is to be contrasted with the
BVI term which involves the vibronic coupling of two excited electronic states jer i and
jes i. As the energy separation of the ground electronic state and the excited electronic
states is normally large the CVI term is likely to be negligible and we shall not consider
it further.
7.3.4 The D VI term

The DVI term which is defined in eq. (7.2.5) involves the vibronic coupling of the excited
0
electronic state jer i to two other excited electronic states jes i and jes i. It may give rise to
first overtones (when k D k 0 ) and binary combination tones k 6D k 0 . This term is likely
to be very small and will also not be considered further.
230
The Raman Effect

Electronic part
Vibrational part
|es
heser
|er
1
0
|er
(i) 0kr(r)|0ki(g) 0
(ii) 1kf(g)|Qk|0kr(r) 0
(ps)ereg
(pr)eges
(i)
(distinguished by )
(ii)
|eg
2
1
0
|eg
(a) Resonant vibronic transition |er |0kr(g) |eg |0ki(g)

(or 00 resonance). HT coupling active in emission process.
Electronic part
Vibrational part
es
heres
|er
1
0
er
(i) 1kr(r)|Qk|0ki(g) 0 (distinguished by )

(ii) 1kf(g)|1kr(r) 0
(ps)eseg
(pr)eger
(i)
(ii)
|eg
2
1
eg
(b) Resonant vibronic transition |er |1kr(r) |eg |0ki(g)

(or 10 resonance). HT coupling active in absorption process.
Figure 7.5 Diagrammatic illustrations of the nature of the two terms in eq. (7.3.2).
(a) Electronic part and vibrational part of the first term. (b) Electronic part and vibrational
part of the second term.
7.3.5 Subsequent developments

The further application of eqs. (7.2.2) to (7.2.5) necessarily involves consideration of the
symmetry of the scattering molecule and of its vibrations. To establish on a quantitative
basis the characteristic features of vibrational resonance Raman scattering we confine the
231
detailed treatment to the simplest case of a molecule which has only one totally symmetric
mode. Other cases involving molecules with more than one totally symmetric mode and
molecules with non-totally symmetric modes are dealt with in less detail.
7.4 AVI TERM RAMAN SCATTERING FROM MOLECULES WITH

TOTALLY SYMMETRIC MODES
7.4.1 AVI term Raman scattering from molecules with one totally
symmetric mode
Examples of species possessing only one totally symmetric mode are diatomic molecules,
tetrahedral or square planar MX4 molecules and octahedral MX6 molecules. For these
molecules only the diagonal components of the transition tensor are non-zero.
From eq. (7.2.2), the contribution of the AVI term to a typical diagonal component,
VI
Axx eg vf :eg vi , of the transition polarizability tensor is given by
rr
hv
1
jvrr
k ihvk j0k i
k
D [px 0er eg ]2
r r
h
r er vr :eg 0i 1 ie vk
fg
VI
Axx eg vf ;eg 0i
k
vk
ig
7.4.1
Here we have set p 0eg er D p 0er eg and have assumed that, initially, all the molecules are
ig
ig
in the ground vibrational level of the ground electronic state, that is vk is given by 0k .
As we are considering a system which has only one totally symmetric normal coordinate
Qk , the vibrational overlap integrals are one-dimensional. Assuming harmonic potentials
for the states jeg i and jer i the several overlap integrals arising in eq. (7.4.1) may be
calculated using the following general recurrence formulae (Manneback, 1951).

b a 1/2 1/2
2

2k
k k
exp

7.4.2
h0bk j0ak i D
2
kb C ka

1/2

b
a
b
v
k
k
k
hvak jvbk C 1i D
hvak jvrk 1i
vbk C 1
ka C kb

1/2
va
2ka kb 1/2 a
hvk 1jvbk i
C
b
a
b
vk C 1
k C k

1/2

1/2
1
2ka
k
hvak jvbk i
7.4.3

vbk C 1
ka C kb
Chapter A14 considers this situation in detail.

Chapter A2 discusses the distinction between xx and .
232
The Raman Effect

1/2 a
vak
k kb
D
hvak 1jvbk i
vak C 1
ka C kb

1/2
2ka kb 1/2 a b
vbk
C
hv jv 1i
vak C 1
ka C kb k k

1/2

1/2
1
2ka
C
k
hvak jvbk i
vak C 1
ka C kb
hvak
1jvbk i
7.4.4
Here, ka and kb are the classical vibrational frequencies, and vak and vbk the vibrational
quantum numbers for the kth mode in the ground and excited electronic states jea i and
jeb i respectively. k is the dimensionless shift parameter defined by
1/2 a b 1/2
1
k k
k D
Qk
7.4.5
h
ka C kb
where Qk is the displacement of the electronic state potential minimum.
These recurrence formulae take on simpler forms in certain special cases. If ka D kb
then we have
h0bk j0ak i
D exp
hvak jvbk
C 1i D
hvak
1jvbk i
and

2k
2
1/2
vak
vbk C 1
vbk
a
vk C 1
7.4.6
hvak 1jvbk i k vbk C 11/2 hvak jvbk i
7.4.7
hvak jvbk 1i C k vak C 11/2 hvak jvbk i
1/2 a 1/2
1
k
Qk
k D
h
2
7.4.8
7.4.9
If the shift parameter k is very small the formulae become especially simple. Such a
case is considered later in this section.
It is evident from these formulae that the magnitude of the displacement Qk is a key
factor in determining the magnitudes of the overlap integrals.
Qk is related to the displacement Sk along the associated symmetry coordinate Sk
of the totally symmetric vibration as follows:
1/2
Qk D
k Sk
7.4.10
where
k is the reduced mass associated with the kth vibration, and Qk and Sk are
readily related to the internal coordinates associated with the kth vibration. For example,
1
1
for a homonuclear diatomic molecule X2 , Q D 2 /2 mX /2 r; for a tetrahedral molecule
233
1/
MX4 , Q D 2mX 2 r; and for an octahedral molecule MX6 , Q D 6mX /2 r. In these

relationships mX is the mass of the atom X and r is the change in the MX bond length.
As already indicated the products of the overlap integrals which appear in eq. (7.4.1)
fg
fg
can have appreciable magnitudes for values of vk D 2, 3, 4 . . . as well as vk D 1
provided that Qk is large enough. In such cases a number of overtones of k appear
in the resonance Raman spectrum giving rise to an overtone progression whose intensity
distribution depends on the magnitude of Q (or r). Provided that the resonant electronic state is associated with a charge transfer mechanism or a transition, then the
overtone progression will be intense.
VI
Since Axx eg vf :eg 0i as defined in eq. (7.4.1) is complex the intensity of the resonance
VI
Raman scatteringk willk be determined by jAxx eg vi :eg 0i j2 After careful consideration of
k
k
the product of two sums of complex quantities, we find that
VI
jAxx eg vi :eg 0i j2

hvfg jvrr ihvrr j0ig i
1
0
4
k
k
k
k
D 2 jpx r g j
2
2
e
e
vr C er vr
h
vr
k
C2

vr <vr
0
0
fg
fg r 0 r
ig
rr ig
hvk jvrr
jvk ihvkr r j0k ivrk vrk
k ihvk j0k ihvk
2r C 2r r 2r 0 C 2r r 0
vk
e vk
vk
e vk
C er vr
0
er vr
7.4.11
where
vrk D er vr :eg 0i 1 and vr 0 D er vr 0 :eg 0i 1
k
7.4.12
r0
It should be noted that vrk and vk both relate to the same electronic state jer i. By
following the procedures discussed in detail in Chapter 5, eq. (7.4.11) may be used to
obtain formulae for the intensity of vibrational resonance Raman scattering for particular
illuminationobservation geometries and polarization states of the incident and scattered
radiation. At this stage it is not necessary to consider the full expression for the intensity.
It will suffice to display explicitly the fourth power frequency term and write for Stokes
Raman scattering
VI
7.4.13
I D KN0i s4 jAxx eg vf :eg 0 j2
k
g f
e vk :eg 0ik ,
where s D 1
N0i is the number of molecules in the initial vibrational state
k
vik D 0ik and K is a constant for a given irradiance of the incident radiation and a given
experimental situation.
Using eqs. (7.4.11) to (7.4.13) we can calculate the relative intensities of the fundamental and overtone bands for a particular vibrational mode over the frequency range
covered by the absorption band involved in the resonance. A plot of these relative intensities against the excitation frequency 1 is called an excitation profile.
In experimental situations the scattered intensity is often determined by counting the number of photons
with scattered frequency s , falling on the detector in a given period of time. In such cases the frequency
factor s4 has to be modified. See Chapter A17, Section A17.2.3.
234
The Raman Effect
Far from resonance, vrk and vr 0 are essentially independent of 1 and thus the intenk
sity of pure vibrational Raman scattering is proportional to 1 eg vf :eg 0i 4 or in a good
k
approximation to 14 . The intensity is thus a smoothly varying function of the excitation
frequency. In the case of pure vibrational resonance Raman scattering however, vrk and
vr 0 are no longer independent of 1 and so the intensity can show a marked depenk
dence on 1 over the frequency range of the resonant electronic absorption band. This is
reflected in the form of the excitation profile.
Basing our discussion on eq. (7.4.11) we consider first some general characteristics
of an excitation profile and then some specific cases. In contrast to the first term in
eq. (7.4.11) which is a single summation of terms, each of which involves one vibrational
level vrr
of the intermediate electronic state jer i, the second term consists of a double
k
summation of terms which have cross-products
of vibrational overlap integrals involving
r 0 r
and
v
of
the
intermediate electronic state jer i. An
pairs of vibrational levels vrr
k
k
important consequence of the nature of the second term is that it can be positive or
negative leading to an increase or a decrease, respectively, in the scattered intensity
relative to that arising from the first term in eq. (7.4.11). This is often described
as an
r 0 r
interference between the contributions from the different levels vrr
and
v
.
The
second
k
k
term in eq. (7.4.11) will become negative if one of the overlap integrals is negative or if
either vrk or vr 0 is negative, a situation which arises when 1 is greater than er vr :eg 0i
k
k
k
or er vr 0 :eg 0i , respectively. In such cases, provided er vrk and er vr 0 are sufficiently small
k
k
k
so that the product vrk vr 0 plays the dominant role in the last term in the numerator of
k
the second term of eq. (7.4.11), the interference effect will cause the excitation profile
f
for a given vibrational transition from eg 0ik to eg vk to show a series of maxima and
minima in the intensity distribution as 1 changes across the frequency range of the
absorption band. However when er vrk and er vr 0 are large these interferences are not
k
apparent and the excitation profile is smooth and devoid of fine structure. The maxima of
the excitation profiles are then shifted to higher frequencies for successive overtones, the
differences in peak positions being approximately equal to the vibrational frequencies of
the excited state.
We now illustrate more specifically these characteristics of excitation profiles and
compare them with those of absorption profiles by considering a simple system. We define
this system, which corresponds to case (c) in Fig. 7.2, to have the following properties:
kr D ki D k and er vr D er vr 0 D 0.2k . We then calculate the relative values of the
intensity for selected values of the dimensionless shift parameter k (k D 0.1, 0.5, 1.0,
1.5 and 2.0) over the range of frequency values covered by the resonant absorption band.
The results are presented as excitation profiles in Fig. 7.6(a) for the fundamental (k ,
first overtone (2k ) and second overtone (3k ). The absorption spectra for the same k
values and the same frequency range are given in Fig. 7.6(b). In these plots the frequency
abscissae are expressed in terms of vrr
k , the vibrational quantum number in the excited
electronic state jer i. Resonance involving transitions to successive states jer ijvrr
k i occurs
r
r
as 1 is changed and successively satisfies the condition 1 D e vk eg 0i . The interferk
ence effects discussed above are clearly shown. For the smallest k value, k D 0.1, the
235
k = 2.0
Intensity
k = 1.5
k = 1.0
k = 0.5
k = 0.1
0
r (r)
k
r (r )
k
Frequency
(i)
Frequency
(ii)
Figure 7.6 (i) Raman excitation profiles for the fundamental k (

), first overtone
2k (- - - - ) and second overtone 3k ( . . . . ); (ii) absorption spectra for a case (c) system
where kr D ki D k , er vr D er vr0 D 0.2 k and k D 0.1, 0.5, 1.0, 1.5 and 2.0. The
frequency scale is expressed in terms of vrr
k . The absolute frequency is given by
er vrk eg 0ik .
236
The Raman Effect
excitation profile is limited to the fundamental transition and exhibits just two maxima
at vrk D 0 and vrk D 1. As k increases, more maxima occur and the position of greatest
intensity of the excitation profile shifts to higher frequency. Overtones are also observed
and their intensities increase relative to the intensity of the fundamental, becoming the
more intense in some regions.
It is evident that the parameters k and er vrk could be obtained by finding those values
which give the best fit between calculated excitation profiles and experimentally determined profiles. Thus excitation profiles can provide a sensitive method for determining
excited state geometries (via ) and lifetimes (via ).
In the foregoing we have assumed that bandwidths are determined entirely by homogeneous broadening. This is valid for the idealized case of a gas at low pressure which
is the basis for the treatments in this book. In more realistic situations inhomogeneous
broadening, which arises from variations in the molecular environment, plays an important
role. In such cases we must consider the total bandwidth which is given by
D C
7.4.14
where is the inhomogeneous bandwidth. To take into account and it is necessary

to convolute eq. (7.4.1) with an assumed band shape function, usually a Lorentzian distribution and integrate over all possible values of the electronic transition frequency; and
then use the result to calculate the intensity.
The details of such a procedure will not be given explicitly here. We shall content
ourselves with the following brief qualitative account of how excitation profiles are
affected by such considerations. In most instances is relatively small, but may range
from zero to quite substantial values. If excitation profiles are broad and unresolved,
is the dominant contribution to the bandwidth. Homogeneous and inhomogeneous
broadening have quite different effects on the excitation profiles. The inhomogeneous
bandwidth does not play any part in the interference phenomena, but increasing
while maintaining constant decreases the intensity of the overtones relative to that
of the fundamental. It is interesting to note that measurement of excitation profiles in
different environments, as for example in solvents or host lattices, makes it possible to
determine because this is largely independent of the molecular environment. This is in
contrast to absorption spectra which are not sensitive to the relative contributions of
and , only to their sum .
The treatment so far has assumed that only one excited electronic state contributes
to the transition polarizability. This assumption is not justified if two or more excited
electronic states are in close proximity to each other. The expression for the transition
polarizability given in eq. (7.4.1), which assumes D 0, must then be replaced by the
following which involves a sum over all relevant excited electronic states jerj i.
VI
Axx eg vf :eg 0i D

k
hvfg jvrj r ihvrj r jOig i

1
k
k
k
[px 0erj eg ]2
rj rj g i 1 i rj rj
h rj g
rj
e v
e v :e 0
e 6De
vk
Section 7.4.4 considers in more detail the case of AVI term scattering when k is very small.
7.4.15
237
In excitation profiles calculated using eq. (7.4.15) interferences are possible between
vibronic contributions from different excited electronic states as well as from the same
excited electronic state. These arise from sign differences at certain positions in the vrj
values where
7.4.16
vrj D erj vrj :eg 0i 1
k
They will also occur if k is positive for one state and negative for another. These
interference effects are decreased when is decreased. If is included the interference
effects also decrease as decreases.
It should be appreciated that when both and are considered and when more than
one electronic state is involved the situation becomes generally rather more complicated.
Often, because of lack of sufficient experimental data, it is assumed that is the same
for all vibronic levels of a given state and for vibronic levels of different states. However,
intuitively it would be expected that excited state lifetimes decrease for higher vibronic
levels leading to progressively larger values. Variations in between different excited
states will play a significant role in interference effects. If the experimental data are
sufficiently detailed such effects can be modelled. It is probably reasonable to assume
that , which depends on the chemical environment, does not vary markedly for different
vibronic levels of either single or multiple excited states.
7.4.2 AVI term Raman scattering from molecules with more than one
totally symmetric mode: general considerations
The treatment of the case of molecules with more than one totally symmetric mode is
straightforward if we assume that the normal coordinates of the excited state are the
same as those of the ground state, as then the multidimensional overlap integrals can
be expressed as products of one-dimensional ones. These products can then be inserted
in eq. (7.4.1) or eq. (7.4.15) and the corresponding intensity profiles calculated for the
single or multi-state cases. If these are fitted to experimental data, values can be deduced
of Q1 , Q2 , . . . the displacement of the excited state potential minimum along each
of the totally symmetric coordinates. The relationship between these displacements and
the geometrical changes which are the consequence of electronic changes is a little more
involved than in the single mode case since a knowledge of the transformation from
normal coordinates to internal coordinates is needed. This involves solution of the secular
equation but the information for this is usually available.
The AVI term resonance Raman spectra of molecules with more than one symmetric
mode may display overtone progressions for each of the active vibrations, that is those
for which there are normal coordinate displacements associated with the electronic transition. In addition there may be combination band progressions involving quanta of more
than one mode. It can be said that, in general, the normal coordinate with the greatest
displacement will give rise to the greatest intensity enhancement and the longest overtone
progression. On this basis it is possible to make qualitative deductions about the excited
state geometry from vibrational resonance Raman spectra even when actual calculations
cannot be performed.
238
The Raman Effect
7.4.3 AVI term Raman scattering from totally symmetric modes when
1k is very small
A special type of AVI term scattering arises when each of the displacements k is very
small, but non-zero. In this case it follows from eqs. (7.4.6) to (7.4.9) that, on the assumption that ka D kb , the overlap integrals have the following values
h1bk j1ak i D h0bk j0ak i D 1
7.4.17
h1ak j0bk i
7.4.18
h0ak j1bk i
D k
fg
or vrr
has a value of 2 or more and
Those overlap integrals in which either vk
k
fg
rr
vk 6D vk may be taken to be zero because they involve higher powers of k . As k is
very small, 2k for example is very much smaller than k . Consequently the summation
is restricted to contributions from vrr
D 1 and vrr
D 0 and so the
over the levels vrr
k
k
k
fg
products of overlap integrals will be non-zero for just two cases: vk D 0, corresponding
fg
to Rayleigh scattering; and vk D 1 corresponding to Raman scattering associated with
fundamental vibrational transitions. Thus, when the displacements k are taken to be very
small, overtones are forbidden.
In the small displacement approximation each vibrational mode can be treated as an
independent harmonic oscillator and this enables tractable expressions to be obtained for
the excitation profile of each of the totally symmetric fundamentals. For example, on
introducing the small displacement approximation into eq. (7.4.1) the xx component of
the transition polarizability for the kth fundamental is given by

1
1
1
AVI
0
2
7.4.19
C
xx eg 1f :eg 0i D [px er eg ] k
k
k
h
0rk ier 0rk 1rk ier 0rk
where 0rk and 1rk , which are special cases of vrk , defined in the equations (7.4.12) are
given by
0rk D er 0r :eg 0i 1 and 1rk D er 1r :eg 0i 1
7.4.20
k
Thus, as we are considering harmonic oscillators,

1rk 0rk D 1r :0i D k
k
7.4.21
It follows that the intensity of the resonance Raman scattering is given by

2
0r C 21r 20rk 1rk

4 1
0
4 2
k
k
I D KN0i 1 k 2 jpx er eg j k
20r C r2 21r C r2
h
k
7.4.22
where we have set er 1rk D er 0rk D r and, as we are considering harmonic fundamental
transitions, eg vf :eg 0i D k . Although it is not possible to obtain absolute values of the
k
displacement parameters in this approximation, it is clear that eq. (7.4.22) enables their
239
Intensity
0
00
10
w1
Figure 7.7 Excitation profile for AVI term scattering in the small displacement approximation. The 00 and 10 contributions are shown by - - - - curves. The interference
contribution and the overall intensity profile are shown by
curves. Note the interference is constructive between the 00 and 10 resonances (where 0rk 1rk is negative) and
destructive elsewhere (where 0rk 1rk is positive).
relative magnitudes, but not signs, to be deduced from the relative intensities of resonance
Raman bands.
The negative sign in the numerator of eq. (7.4.22) results in constructive interference
between the 00 and 10 maxima where 0rk 1rk is negative, and destructive interference elsewhere where 0rk 1rk is positive. Figure 7.7 illustrates an excitation profile for
AVI term scattering in the small displacement approximation. The 00 and 10 contributions and the interference contribution to the intensity are shown, as well as the overall
intensity profile which has the form of the sum of two Lorentzians modified by a small
interference term.
7.5 AVI TERM RAMAN SCATTERING INVOLVING

NON-TOTALLY SYMMETRIC MODES
As already indicated there are two possible mechanisms for AVI term Raman scattering
involving non-totally symmetric modes: a change of symmetry in the resonant excited
state, or excited state JahnTeller coupling.
7.5.2 AVI term scattering involving a change of molecular symmetry of

the resonant excited state
When a molecule undergoes a change of symmetry upon electronic excitation then the
AVI term active modes are those that are totally symmetric in the subgroup (or common
240
The Raman Effect
group) formed by the symmetry operations common to the ground state and excited state
point groups. An example is the ethene molecule which has planar D2h geometry in
the ground state and twisted D2d geometry in the first 1 B1u excited electronic state. The
common group is D2 , and by correlation between D2 and D2h we find that the AVI term
active fundamentals are those with ag and au symmetry. Now the au fundamental 7 , the
twist about the CDC bond, is itself Raman inactive. However, even quanta of 7 , namely
2 7 , 4 7 etc. have Ag symmetry and so are Raman active. Excitation within the contour
of the transition of ethylene has been found to give resonance Raman spectra
displaying progressions in the CDC mode and even harmonics of 7 au . This result
provides evidence for a twisted geometry for the ethene molecule in the first 1 B1u excited
electronic state.
This is an example of a general empirical rule that the vibrational coordinates responsible for converting a molecule from its ground state to its excited state geometry will
give rise to resonance-enhanced Raman bands.
7.5.3 AVI term scattering involving excited state JahnTeller coupling

If the resonant excited electronic state jer i is degenerate, AVI term activity may arise for
some non-totally symmetric modes. The non-totally symmetric vibrations belonging to
the irreducible representations contained in the symmetric direct product er er may
be responsible for JahnTeller coupling in the excited state. Some common examples of
excited states which show this behaviour are T2 Td , T1u Oh and Eu D4h , for which
the direct products are as follows:
T2 T2 D fA1 C E C T2 g C [T1 ]
T1u T1u D fA1g C Eg C T2g g C [T1g ]
Eu Eu D fA1g C B1g C B2g g C [A2g ]
Vibrations which transform as the antisymmetric part of the direct product (given in square
brackets) are not JahnTeller active because they do not lift the degeneracy of the excited
state. An illustration of this is provided by considering the fundamental vibrations of
b1g , b2g and a2g symmetry of the square planar [Pt(CN)4 ]2 ion which has D4h symmetry.
It can be seen from Fig. 7.8 that b1g and b2g vibrations remove the C4 axis which is
responsible for the double degeneracy, whereas an a2g vibration does not. It follows that
the b1g and b2g modes are JahnTeller active but the a2g mode is not.
7.5.4 Summary of excited state JahnTeller effects in resonance

Raman scattering
If the JahnTeller effect is small the result is the appearance of combination band
progressions involving quanta of totally symmetric modes plus one quantum of a
JahnTeller active vibration. If the JahnTeller effect is large then there appear
combination band progressions involving quanta of totally symmetric modes plus multiple
241
a2g
b1g
b2g
Figure 7.8 The effect of a2g , b1g and b2g distortions on a D4h species as for example
[Pt (CN)4 ]2 .
quanta of JahnTeller active vibrations, as well as overtone progressions of the latter. We

would expect the excited state JahnTeller effect to be a dynamic one so that no permanent
distortion of the molecule occurs. The major geometric changes accompanying excitation
to a degenerate excited state will therefore be those involving totally symmetric modes.
Thus large JahnTeller effects will be relatively uncommon.
7.6 B VI TERM SCATTERING INVOLVING VIBRONIC COUPLING

OF THE RESONANT EXCITED STATE TO A SECOND
EXCITED STATE
7.6.1 Introduction
BVI term scattering can arise for both totally symmetric and non-totally symmetric vibrations. Although the BVI term is usually very much smaller than the AVI term, it plays
the determining role in resonance Raman scattering when the AVI term is zero. We
therefore consider in some detail BVI term scattering from molecules with non-totally
symmetric modes, and then more briefly BVI term scattering from molecules with totally
symmetric modes.
7.6.2 B VI term scattering from molecules with non-totally symmetric

modes
The several factors involved in the BVI term have been discussed in Section 7.3.2. An
important factor in determining the magnitude of the BVI term, and hence the resonant
Raman intensity, is the vibronic coupling integral heks er defined in eq. (7.3.1). If a nontotally symmetric normal coordinate Qk is to give rise to BVI term activity it must be
242
The Raman Effect
effective in coupling the resonant state er to a second excited state es . This requires that
the following symmetry condition is fulfilled, namely that Qk is contained in the product
e r e s .
We now consider further the implications of eq. (7.3.2) which we rewrite in the
following form:

1
heks er
h1jQk j0i
BVI
0
0
p er eg
eg 1f :eg 0i D 2 p eg es
k
k
er es
0rk ir
h

k
1
h
h1jQ
j0
r
s
k
C 2 p 0eg er e e p es eg
7.6.1
er es
1rk ir
h
where we have set
fg
h1k
ig
rr
jQk j0rr
k i D h1k jQk j0k i D h1jQk j0i
ig
fg
h0rr
k j0k i D h1k
rg
j1k i D 1,
7.6.2
7.6.3
0rk and 1rk are defined in the eqs. (7.4.20) and we have taken er 1rk D er 0rk D r . The
basic assumptions involved in eqs. (7.3.2) and (7.6.1) have been discussed earlier in
Section 7.3.2. Here we just reiterate that an important consequence of these assumptions
is that BVI term scattering occurs only for fundamentals.
It follows from the nature of the products p 0eg es p 0er eg and p 0eg er p 0es eg in
eqs. (7.3.2) and (7.6.1) that a BVI term transition polarizability component is not symmetric
in the indices and . We therefore proceed to construct the components of the symmetric
tensor as and the antisymmetric tensor aa into which a general asymmetric second-rank
tensor a may be decomposed. Now
D s C a
where
s D
C
2
7.6.5
a D

2
7.6.6
and
Thus from eq. (7.6.1) and setting hekr es D heks er we have

k
1
h
VI
s
r
ee
h1jQk j0ifp 0eg es p 0er eg
B g f g i D 2
e 1k :e 0k
er es
2h

1
1
0
0
p eg er p es eg g
0rk ir
1rk ir
7.6.4
Chapter A10, Section A10.4.2, deals with tensor decomposition.
7.6.7
243
where the plus sign is to be taken throughout for s and the minus sign throughout for
a . The symmetric and antisymmetric tensor contributions to the intensity profiles are
then given by
VI
7.6.8
I D KN0i 1 k 4 jB g f g i j2
e 1k :e 0k
with
VI
jB Cg f g i j2
e 1 :e 0
k
1
D 4
4h

heks er
er es
2
fh1jQk j0ig2 fp 0eg es p 0er eg C p 0eg er p 0es eg g2
21r C 20r C 4r2 C 21rk 0rk

k
20r C r2 21r C r2

k
7.6.9
and

2

2
k
BVI

1
h
s
r
ee
f

fh1jQk j0g2 fp 0eg es p 0er eg p 0eg er p 0es eg g2
eg 1k :eg 0ik D 4h4 r s
e e
2

1r C 20r 21rk 0rk
k
k
7.6.10
20r C r2 21r C r2
k
Equation (7.6.8) combined with eqs. (7.6.9) and (7.6.10) may be used to calculate the
symmetric and antisymmetric tensor contributions to excitation profiles.
Comparing eqs. (7.6.8) and (7.6.10) with eq. (7.4.22) shows that the frequency dependence of the antisymmetric tensor contribution to the intensity of the BVI term scattering
is the same as for the AVI term for the case when k is very small. The antisymmetric
tensor contribution to the intensity exhibits constructive interference between the 00
and 10 maxima where 0rk 1rk is negative, and destructive interference elsewhere where
0rk 1rk is positive (Fig. 7.7). Using eq. (7.4.21) the numerator may be shown to be equal
to k2 . Also away from resonance, where er eg 1 k and r can be ignored we
can set the frequency denominator of eq. (7.6.10) equal to 0rk 4 , since 0rk 1rk . Thus
VI
the frequency dependence of jB g f g i j2 away from resonance is
e 1k :e 0k
k 2
0rk 4
7.6.11
We see that the magnitudes of the antisymmetric tensor components tend rapidly to zero
outside the resonance region.
The behaviour of the symmetric tensor contribution to the intensity is quite different
because the term involving 0rk 1rk has a positive sign. Thus destructive interference
between the 00 and 10 resonances and constructive interference elsewhere may occur
as shown in in Fig. 7.9 which gives the 00 and 10 contributions, the interference
The Raman Effect
Intensity
244
0
00
10
w1
Figure 7.9 Excitation profile for BVI term scattering (symmetric tensor case). The 00 and
10 contributions are shown by - - - - curves. The interference contribution and the overall
curves. Note the interference is destructive between the
intensity profile are shown by
00 and 10 resonances, and constructive elsewhere.
contribution and the overall intensity profile. Away from resonance where 0rk 1rk and
VI
r can be ignored, the frequency dependence of jB Cg f g i j2 is given by

4
0rk 2
e 1k :e 0k
7.6.12
which is effectively constant well away from resonance. Thus outside the resonance
region the antisymmetric tensor components are zero and Raman scattering is determined
entirely by a symmetric tensor with the Raman intensity following the normal fourth
power frequency law.
7.6.3 B VI term scattering from molecules with totally symmetric

modes
We stated in Section 7.3.2 that BVI term scattering may occur for totally symmetric modes
if the excited states jer i and jes i have the same symmetry. For totally symmetric modes
only the diagonal components of the transition tensor are non-zero and so the tensor is
necessarily symmetric. Thus the BVI term contribution from the xx component of the
transition polarizability to the intensity of totally symmetric modes can be obtained from
eqs. (7.6.7) to (7.6.9) by setting D D x and taking the C sign throughout. Hence we
have

2

4 BVI C
I D KN0i 1 k xx g f g i
7.6.13
k
e 1k :e 0k
245
and introducing
VI
Bxx Cg f g f
e 1 :e 0
k

1
heks er
D 2
fh1jQk j0igfpx 0eg es px 0er eg g
er es
h

1
1
C
0rk ir
1rk ir
7.6.14
we obtain

2
1
heks er
fh1jQk j0ig2 fpx 0eg es px 0er eg g2
I D KN0i 1 k 4
k
r
s
h
e e
2

2
2
1r C 0r C 4r C 21rk 0rk
k
k
0rk C r2 21r C r2

4
7.6.15
Intensity
The intensity profile has the form shown in Fig. 7.9 for the symmetric tensor contribution to BVI term scattering from a non-totally symmetric mode.
For totally symmetric modes interference between AVI and BVI term contributions
can now arise and will be important if the two terms have similar magnitudes. This is
illustrated in Fig. 7.10 where the AVI term relates to the small displacement approximation
and the AVI and BVI term contributions are equal. The interference shown in Fig. 7.10(c)
is constructive for the 00 and destructive for the 10 resonances and produces the
intensity profile shown in Fig. 7.10(d). The case illustrated has assumed that the transition
moments px 0eg es and px 0er eg have the same signs. If they have opposite signs then
destructive interference occurs for the 00 resonance and constructive interference for
the 10 resonance so that the peak at the 10 resonance then has the larger intensity.
00
(a)
10
00
10
(b)
00
(c)
10
00
10
(d)
w1
Figure 7.10 Excitation profiles illustrating AVI BVI term interference: (a) AVI term; (b) BIV
term; (c) interference contributions to the total intensity when the AIV and BVI terms are
equal; (d) resultant excitation profile with AVI BVI term interference.
246
The Raman Effect
7.7 SYMMETRY, RAMAN ACTIVITY AND DEPOLARIZATION

RATIOS
7.7.1 General symmetry considerations
The symmetry conditions that must be fulfilled for vibrational Raman activity have been
considered in Chapter 5, Section 5.8. However, the treatment given there relates to the
special case where the initial and final electronic states are the ground electronic state
which is taken to be non-degenerate and the polarizability is a function of the nuclear
coordinates only.
For consideration of the symmetry conditions in relation to AVI and BVI term Raman
activity it is necessary to state the symmetry requirement in a more general form. Let the
initial vibronic state jii have the representation i , and the final vibronic state jfi the
representation f . Then a transition polarizability tensor component fi can span the
irreducible representation R which is obtained from the relationship
R D i f \ T
7.7.1
where T indicates a set of irreducible representations into which the second rank transition polarizability tensor is decomposed under the point group symmetry of the molecule
in question. In other words the direct product i f and T must have one or more
representations in common for Raman activity.
If the BornOppenheimer approximation is applicable, as we have assumed in our
treatment of AVI and BVI term scattering, jii D jei ijvi i and hence
i D ei vi
7.7.2
f D ef vf
7.7.3
Similarly jfi D jef ivi i, and hence
Here ei and ef are the representations of the initial and final electronic states and vi and
vf the representations of the initial and final vibrational states. Introducing eqs. (7.7.2)
and (7.7.3) into eq. (7.7.1) we have
R D ei vi ef vi \ T
7.7.4
For the fundamental vibrational transition vi D 0 to vf D 1, with both jei i and jef i
the ground electronic state, which is taken to be non-degenerate, ei D ef D 1 and
vi D 1 where 1 is the totally symmetric representation. Hence
R D vf \ T
7.7.5
This is an alternative statement of the symmetry selection rule given in Chapter 5,

Section 5.8, for a fundamental vibrational transition. In effect eq. (7.7.5) states that the
247
irreducible representation R is in this case equal to vf which for a fundamental transition is given by Q the irreducible representation of the normal mode Q; and for Raman
activity there must be one or more transition tensor components belonging to this same
irreducible representation Q . The simple relationship given by eq. (7.7.5) will no longer
apply if the transition involves some kind of degeneracy. It is then necessary to revert to
eq. (7.7.4) so that the effect of degeneracy can be taken into account.
With the general symmetry conditions established we can now proceed to examine
Raman activity and depolarization ratios in a number of cases of resonance involving AVI
and BVI term scattering. Where a detailed analysis is possible we shall find that resonance
Raman scattering can yield more information than non-resonant scattering. In particular
the symmetry of resonant electronic transitions and electronic band assignments can often
be determined.
7.7.2 The AVI term

As already explained, a Raman band arising from a totally symmetric fundamental may
acquire resonance enhancement of its intensity from either the AVI term or the BVI term
of the transition polarizability tensor. However it is usually only the AVI term that is
important, and accordingly we consider this in some detail first.
For convenience we repeat below eq. (7.4.1) which gives the form of the AVI term for a
typical diagonal component of the transition tensor which we shall use in our discussion:
rr
hv
1
jvrr
VI
k
k ihvk j0k i
Axx eg vf :eg 0i D [px 0er eg ]2
k
k
h
er vr :eg 0i 1 ir
r
fg
vk
ig
7.7.6
We recall from earlier in this chapter that r D er 0rk D er 1rk . Equation (7.7.6) is based
on the assumption that only one resonant electronic manifold is involved, and that
ef D ei D eg .
If the ground electronic state is non-degenerate then the symmetry requirement for
Raman activity of a fundamental vibrational transition is that given by eq. (7.7.5). Since
the AVI term involves only totally symmetric modes vf D 1 , only those transition tensor
components belonging to 1 can be involved. With the exception of a few special cases
which we discuss below, only the diagonal components of the transition tensor belong
to the totally symmetric representation. Thus in the case of totally symmetric vibrations,
when deducing properties like depolarization ratios, we normally have to consider only
the diagonal components of the transition tensor. The few exceptions to this are of two
kinds. In a few point groups of low symmetry, namely C1 , Ci , Cs , C2 and C2h one
or more off-diagonal components of the symmetric transition tensor do belong to the
totally symmetric representation, and in the point groups C1 , Ci , Cs , Cn , Cnh and Snh n D
2, 3, 4, 5, 6 one or more off-diagonal components of the antisymmetric transition tensor
belong to the totally symmetric representation. However, off-diagonal components of the
transition tensor, whether symmetric or antisymmetric, can only arise from the BVI term.
Thus we can say that for the totally symmetric modes which derive their intensity solely
248
The Raman Effect
from the AVI term, properties such as depolarization ratios can be deduced using only
the diagonal components of the transition tensor. In cases where BVI term scattering
contributes significantly to the intensity of a totally symmetric mode, such deductions
will need modification and this will be dealt with later.
At resonance the AVI term imposes further restrictions on the transition tensor components involved. When the ground electronic state jeg i is non-degenerate the resonance
involves one particular electronic transition jer i jeg i because we have assumed only
one electronic manifold to be operative. This leads to selective enhancement of one or
more diagonal tensor components depending on the symmetry properties of the transition
electric dipole p 0er eg which is involved. As jeg i is being taken as non-degenerate, the
requirement is that r and g must belong to the same symmetry class for p 0er eg to be
non-zero.
We now discuss some examples by way of clarification. For convenience we shall
write xx for xx fi and so on in the rest of this section and also in Section 7.7.3. We
consider first the case where the molecular point group is C2v . Then er can be A1
(with z polarization or 1 D z , B1 (with x polarization) or B2 (with y polarization).
If resonance enhancement occurs via an electronic transition with x polarization then
only xx is enhanced and at resonance we can set yy D zz D 0. As all off-diagonal
components are zero it follows from the definitions of the mean polarizability a and the
anisotropy given in Chapter A14 by equations (A14.7.26) and (A14.7.32), respectively,
that a2 D 19 2xx and 2 D 2xx . Hence using eq. (5.5.13) we have /2; ?i D 1/3.
Similar arguments apply if the resonant electronic state is doubly degenerate. For
example, a molecule of D4h symmetry can have an electronic state of symmetry Eu
with x, y polarization. Thus xx and yy can be equally enhanced and so, at resonance,
xx D yy and zz D 0. Hence a2 D 49 2xx , 2 D 2xx and /2; ?i D 18 .
A molecule of high symmetry can have a triply degenerate electronic state. For example,
a molecule of Td symmetry has a triply degenerate electronic state F1u (with x, y, z
polarization) so that xx , yy and zz are equally enhanced. Thus the resonance situation
is the same as the non-resonance case and /2; ?i D 0 in both cases.
We can see that measurements of depolarization ratios of scattering which arises
entirely from the AVI term can give information about the symmetry of electronic states.
However, the clear-cut cases just discussed do not apply in a number of instances. Two
examples are: if either two non-degenerate electronic transitions lie close together or
a non-degenerate transition is in close proximity to a doubly degenerate one. In the
former case the /2; ?i value will depend on the relative magnitudes of say xx and
yy as illustrated in Fig. 7.11. Likewise, in the second case, the /2; ?i value will
depend on the relative magnitudes of say xx D yy and zz ; this situation is illustrated
in Fig. 7.12. As xx and yy in the first case, and xx D yy and zz in the second case
will have different frequency denominators involving er vr :eg 0i and er 0 vr 0 :eg 0i , the value
k
k
k
k
of /2; ?i will vary with 1 and so polarization dispersion will be exhibited. This
is in contrast to depolarization ratios involving just one resonant electronic state which
show no dispersion.
249
0.8
0.7
i
0.6 r(p/2; )
0.5
0.4
0.3
0.2
0.1
10
10
axx
ayy
Figure 7.11 Plot of /2; ?i against xx /yy for the case where two non-degenerate
electronic transitions contribute to the transition polarizability.
0.8
0.7
r(p/2; i)
0.6
0.5
0.4
0.3
0.2
0.1
10
0
2
axx
ayy
=
azz
azz
10
Figure 7.12 Plot of /2; ?i against xx /zz D yy /zz for the case where one degenerate
transition and one non-degenerate electronic transition contribute to the transition polarizability.
We now consider cases where the electronic ground state is degenerate. The symmetry
requirement is now that er eg must contain D x, y, z. As an example we
consider the doubly degenerate, E00g ground state of an octahedral molecule of point group
Oh . Now E00g E00g D A1g C F1g and therefore the Raman bands associated with 1 a1g
(and its overtones) can involve both the symmetric tensor components xx C yy C zz
which belong to a1g and the anti-symmetric tensor components xy , yz and zx which
250
The Raman Effect
belong to F1g . Resonance enhancement occurs via the electronic state F1u (x, y, z polarization). Of course the anti-symmetric components can only arise from BVI term scattering.
In this example both a and 2 will be non-zero and so /2; ?i may take any value
between zero and infinity depending on the relative magnitudes of a and 2 . The depolarization ratio will also show dispersion throughout the resonance region because the two
types of tensor components have different dependences on the excitation frequency.
A special case can arise if the ground electronic state is not orbitally degenerate but
posesses spin degeneracy. Strong spinorbit coupling may then cause mixing of the
ground state with degenerate excited states of the same spin degeneracy. In this way
the ground electronic state can acquire some degenerate character and the depolarization
ratios of totally symmetric modes may become anomalously large.
7.7.3 The B VI term

For convenience we give below eq. (7.6.1) in an expanded form to facilitate the ensuing
discussion.

r
r
g
O
1 heg jp jes ihes jH/Q
1
k 0 je ihe jp je ih1jQk j0i
BVI
eg 1f :eg 0i D 2
k
k
er es
0r ir
h
k
s
s
g
O
1 heg jp jer iher jH/Q
1
C 2
er es
1rk ir
h
7.7.7
Background information for eq. (7.7.7) will be found in Section 7.6.2.
In contrast to the situation for AVI terms there can be both diagonal and off-diagonal
BVI terms. In the former case the BVI term contributes to scattering arising from totally
symmetric modes. In the latter case the BVI term contributes to scattering arising from
non-totally symmetric modes and we shall concentrate mainly on this case.
The essential symmetry requirement for BVI term scattering is that given by eq. (7.7.5),
namely that R D uf \ Tf2g , because, in the approximation we are considering, only
fundamentals are involved in BVI term scattering. This requirement is based on the
assumption that the initial and final electronic states are the ground electronic state which
is taken to be non-degenerate.
We now consider the relationships between /2; ?i and the transition tensor invariants 2 and 2 for non-totally symmetric modes in a number of special cases. For
non-totally symmetric modes a D 0 and hence it follows from the definition of /2; ?i
in Reference Table 5.2(a) that
/2; ?i D
3
52
C 2
4 4
which may be recast in the useful form

3
2
4
i
/2; ?
D 2
f D
5
4

7.7.8
7.7.9
251
The general definitions of 2 and 2 in terms of transition tensor components given in

Chapter A14 by eqs. (A14.7.31) and (A14.7.29) lead to the following definition of f:

2
2
,
7.7.10
f D 2 D

C 2

,
where the sums are over all pairs of tensor components. However, if only one pair of components need be considered by virtue of selective enhancement at resonance, then we have
f D
j j2

j C j2
7.7.11
For simplicity we base our discussion on eq. (7.7.11) rather than eq. (7.7.10). The
function f is a measure of the anti-symmetry of the transition tensor. Thus for a purely
symmetric transition tensor D , so that 2 D 0, f D 0 and /2; ?i D 34
(depolarized scattering). For a purely anti-symmetric transition tensor D , so
that 2 D 0, f D 1 and /2; ?i D 1 (inverse polarization). For the general case
of a transition tensor of mixed symmetry 6D , so that 0 < 2 orf < 1 and
0 < /2; ?i < 1 (anomalous polarization).
We now explore the frequency dependence of /2; ?i in the vicinity of resonance.
VI
Because of the rather cumbersome expressions for terms like B eg 1f :eg 0i , it will prove
k
k
convenient to introduce a shorthand notation at this point. We therefore write eq. (7.7.7)
in the form
D lL C mM
7.7.12
VI
where we have introduced as a convenient shorthand for B eg 1f :eg 0i and
k
r
r
g
O
1 heg jp jes ihes jH/Q
es er
h 2
g
r
r
s
s
g
O
1 he jp je ihe jH/Q
mD 2
es er
h
1
LD
0rk ir
1
MD
1rk ir
lD
7.7.13
7.7.14
7.7.15
7.7.16
The terms l and m involve the electronic states jer i and jes i which participate in the
HerzbergTeller coupling. The term L involves 0rk and can give rise to 00 resonance;
the term M involves 1rk and can give rise to 10 resonance.
252
The Raman Effect
In this shorthand notation

D mL C lM
7.7.17
because, on the exchange of the subscripts and , l ! m and m ! l. Also, the

symmetric and antisymmetric tensor components s and a , respectively, are
given by
s D 12 f C g D 12 l C mL C M
7.7.18
a D 12 f g D 12 l mL M
7.7.19
Introducing eqs. (7.7.18) and (7.7.19) into eq. (7.7.11) we have

f D
jl mL Mj2

jl C mL C Mj2
7.7.20
Then using eqs. (7.7.15) and (7.7.16) we find that

1rk 0rk 2
jl mj2
2
f D 2 D

jl C mj2 1rk C 0rk 2 C 4r2
7.7.21
and using eqs. (7.4.20) and (7.4.21), which relate to 1rk and 0rk , we obtain
jl mj2
f D

jl C mj2
2
k
er 0g
k
2
2
k
C
1 C r2
2
7.7.22
The frequency part of this function in the braces f g is shown in Fig. 7.13. It has
the form of a simple Lorentzian, centered midway between the 00 resonance which
occurs when 1 D er 0g and the 10 resonance which occurs when 1 D er 0g C k . Its
k
halfwidth is r and the maximum value of the frequency function which occurs when
k
er 0g C
D 1 is
k
2
2
k
2
2
r
At both the 00 and 10 resonances the frequency function in the braces f g has a
value of
2
k
2

2
k C 2
r
2
253
( 2k )2
Gr2
( 2k )2
wk 2
( 2 ) + Gr2
2Gr
Gr
Gr
10
00
0
Frequency
wer0 g = w1
k
wer0g +
w + wk = w1
wk
= w1
2
g
e r0 k
Figure 7.13 The Lorentzian form of the frequency function given in the braces f
eq. (7.7.22).
g in
This tends to unity if r2 is small and then

l m 2

f D
l C m
at these points. An interesting special case occurs if either l D 0 or m D 0. Then

l m 2

l C m D 1,
and when the frequency function tends to unity f D 1 and /2; ?i ! 2. The condition l D 0 means that there is no 00 resonance and similarly the condition m D 0 means
that there is no 10 resonance. These conditions can arise if particular transition dipole
components are zero, because of the symmetry of the electronic states involved.
We now examine how the symmetries of the various electronic states involved affect
and . We shall proceed by considering illustrative examples.
We first define the initial and final electronic states to be the ground state which we
take to be non-degenerate. The initial and final vibrational states are necessarily j0i i and
j1f i, respectively, as explained in the derivation of the formulae for BVI term scattering.
We now consider as the first example the case where the two excited electronic states
jer i and jes i which are involved in the HerzbergTeller coupling are both non-degenerate.
jeg i be allowed with x polarization and the
We let the electric dipole transition jer i
s
g
electric dipole transition je i
je i be allowed with y polarization. Thus
heg jpy jer i D 0 D her jpy jeg i
7.7.23
hes jpx jeg i D 0 D heg jpx jes i
7.7.24
and
254
The Raman Effect
Introducing these conditions into eq. (7.7.7) we have for resonance involving the elecjeg i
tronic transition jer i

s
s
g
O
1 heg jpx jer iher jH/Q
1
k 0 je ihe jpy je ih1jQk j0i
7.7.25
xy D 2
er es
1rk i
h

r
r
g
O
1 heg jpy jes ihes jH/Q
1
k 0 je ihe jpx je ih1jQk j0i
yx D 2
7.7.26
er es
0rk ir
h
We see from eq. (7.7.25) that xy involves only the 10 resonance term. Similarly, we
see from eq. (7.7.26) that yx involves only the 00 resonance term.
Comparison of eqs. (7.7.25) and (7.7.26) shows that xy 6D yx . However, this
inequality is due entirely to the different frequency denominators 1rk ir and 0rk ir .
On 00 resonance yx predominates, whereas on 10 resonance xy dominates. As
already indicated limiting values of f D 1 and /2 :?i D 2 result.
As our second example we consider HerzbergTeller coupling between two doubly
degenerate excited states jer i and jes i. We take the non-degenerate ground electronic state
to have A1g symmetry and both jer i and jes i to have Eu symmetry. This is compatible
with a molecular point group symmetry of D4h . Transitions to each Eu state from an A1g
ground state are dipole allowed with both x and y polarizations. We now proceed to show
that for such a system the transition tensor will be antisymmetric with xy D yx .
It follows from eqs. (7.7.12) and (7.7.17) that if we can establish that l D m for this
system then xy D yx
Now as jer i is doubly degenerate
her jpy jeg i D heg jpx jer i
7.7.27
and as jes i is also doubly degenerate

heg jpx jes i D heg jpy jes i
7.7.28
Introducing these equalities into the definitions of l and m given by eqs. (7.7.13) and
(7.7.14), respectively, with D x and D y, it follows that l D m provided that
r
s
r
O
O
hes xjH/Q
k 0 je i D he yjH/Q
k 0 je i
7.7.29
where the cartesian components of the states jer i, and jes i are given in brackets.
Symmetry arguments show that eq. (7.7.29) is satisfied if D y and D x when Qk
has A2g symmetry. If D x and D y then the terms on each side of eq. (7.7.29) are
both zero. The symmetry arguments involved are rather specialized and are not treated
here. We merely indicate the fact that, just as matrix elements between eigenfunctions of
the angular momentum operators can be determined using 3-j symbols, in an analogous
manner, matrix elements of operators operating on wave functions classified according to
the representations of point groups can be determined using V coefficients. Indeed all the
255
concepts used for treatment of the angular momentum problem, 3-j, 6-j and 9-j symbols,
the WignerEckart theorem, reduced matrix elements and so on, are mirrored in a similar
set of concepts for point groups.
Adopting the results of the symmetry arguments we see that for the transition
f
g
je ij1k i
jeg ij0ik i associated with a normal coordinate Qk which belongs to the
irreducible representation A2g , xy D yx .
These symmetry considerations can be summarized using a state tensor. The state tensor
for the A2g representation of the point group D4h is as follows
D4h
Ey
p
2
A2g :
x
0
0
p
2
0
0
0
This state tensor is to be interpreted as follows. The entry in the xy position indicates
that xy involves an intermediate state of symmetry Ey and the entry in the yx position
indicates that yx involves an intermediate state of symmetry Ex ; and the numerical
factors show that xy D yx . State tensors may be constructed using V coefficients,
and these are listed in Table 7.1 for all the common point groups.
Table 7.1 Raman state tensors for important point groups, based on
Mortensen and Hassing (1980). The symbolism used is explained at the
end of the Table on page 262.
C2 , C2h
B
A : B00
0
C2v C2 , yz
B1
A1 : 0
0
0
B1 : 0
B1
D2 , D2h C2z , C2y
B0
B000
0
0
0
A
0
0
B2 0
0 A1
0 A1
0 0
0 0
B3 0
0
A : 0 B2 0
0
0 B1
0 0 B1
B2 : 0 0 0
B3 0 0
0
B:0
B
0
0
B0
A
A0
0
0 B2 0
A2 : B1 0 0
0
0 0
0 0
0
B2 : 0 0 A1
0 B2 0
0 B2 0
B1 : B3 0 0
0
0 0
0 0
0
B3 : 0 0 B1
0 B2 0
256
The Raman Effect

D2d S4 , C2y
Ex
p
2
A1 :
0
0
Ex
p
B1 :
0
Ex : 0
C3 C3 , T, S6 C3 , T, i
C3h C3 , h , T
0
0
0 Ey
Ey
p
2
0
B2
0
Ey
p
2
0
0
B2
0
Ex
B2 :
p
2
Ey : 0
Ex
E01y
p
0
E1y
p
0
2
0
A2
E1y
p
0
0
0
0 A2
0 0
0 0
Ey
p
2
Ex
A2 :
p
E0y
x
p
p
0
2
2
A : Ex
Ey
p
p

0
2
2
0
0
A
Ey
p
A
0
E
x
Ex :
p
0
0
2
0
Ex
0
0
E
1x
p
0
A1 :
E
p1x
2
0
E1x :
1x
E
p2
0
0
E2x : 0
E1x
Ex
p
2
Ey :
0
E1y
E1x
p
2
:
0
0
0
Ey
p
0
0
0
0 B2
0 0
0 0
Ey
p
2
E0y
0
E1y
p
2
0
0 0
E2y : 0 0
0 E1y
0
A2
0
257

C3v C3 , yz
Ex
p
2
A1 :
0
0
0
E
Ex :
px
2
E0x
D3 C3 , C2y , D3d C3 , C2y , i
Ex
p
2
A1 :
0
0
0
E
Ex :
px
2
0
D3h C3 , C2y , h
A11
E1x
p
2
:
0
0
0
E1x :
1x
E
p2
Ey
p
2
0
Ey
p
2
0
Ex
A2 :
p
A1
0
Ex
p
2
Ey :
0
A1
Ey
p
0
2
0
A2
Ey
p
0
0
A2
E0y
0
E1y
p
2
0
E1y
p
2
0
0
Ex
p
2
Ey :
0
E0x
0
0
Ex
A2 :
p
A22
0
0
0
0
0
0
0
A22
: 0
0 E1y
0
A21
E1y
E2x
E2y
:
1x
E
p2
0
E1x
p
2
:
0
Ey
p
2
0
0
0
0
0
Ey
p
2
E0y
Ey
p
2
0
A1
0
0
A2
Ey
p
2
0
E1y
p
2
0
E1y
0
p
2
0
0
0
0 0 A22
: 0 0 0
E1x 0 0
258
The Raman Effect

C4 C4 , T, C4h C4 , h , T
E
x
p
0
A:
E
px
2
0
0 0
Ex : 0 0
Ex
D4 C4 , C2y , D4h C4 , C2y , i
B1 :
A2
0
Ey
p
2
0
0
0
Ey
Ex
p
2
A1 :
0
Ex
p
B1 :
0
Ey
p
2
0
Ex
p
2
Ex : 0
0
0
C4v C4 , yz
Ex
p
2
A1 :
0
E0y
p
0
Ey
p
0
2
0
A
A
0
Ex
A2 :
p
A1
0
Ey
p
2
0
0
0
0 0 A1
Ex : 0 0 0
Ex 0 0
0
A
0
Ey
p
2
0 A2
0 0
0 0
Ey
p
2
0
0
0
Ey
0
0
Ey
p
0
0
0
Ex
B2 :
p
2
Ex
A2 :
p
Ey : 0
Ey
p
2
Ey : 0
Ex
E0y
p
2
Ey
p
2
0
Ex
B2 :
p
2
Ey
p
2
0
Ex
p
0
B:
E
px
2
0
0 0
Ey : 0 0
0 Ey
0
A2
0
0
A1
0
259

C6 C6 , T, C5 , C6h
C6v C6 , yz , C5v
E0y
Ex
p
p
0
2
2
A : Ex
Ey
p
0
p
2
2
0
0
A
0 0 A
E1x : 0 0 0
E1x 0 0
E1y
p
0
0
2
E2x : E1x
0
0
p
2
0
0
0
Ex
p
2
A1 :
0
Ey
p
0
2
0
A1
0 A1
0 0
0 0
E1y
p
0
2
0
0
0
0
E1x : 0
E1x
E2x : E1x
p
2
0
0
D6 C6 , C2y , D6h C6 , C2y , i, D5
Ex
p
0
2
Ey
A1 :
p
0
2
0
0
0
0
0
E1x : 0
0 E1y
E1y
p
0
2
E2x : E1x
0
p
2
0
0
E1y
E2y
0 0
0
: 0 0 A
0 E1y 0
E1x
p
0
0
2
E1x
:
p
0
0
2
0
0
0
A2 : E x
p
2
0
0 0
E1y : 0 0
0 E1y
E1x
p
2
E2y :
0
A2
0
A2
0
0
0
A2 : E x
p
2
0
0 0
E1y : 0 0
Ex 0
E1x
p
2
E2y :
0
Ey
p
2
0
0
0
A1
0
0
E1y
p
2
0
Ey
p
2
0
0
0
A2
0
0
0
E1y
p
0
2
0
0
0
260
The Raman Effect

C5h C5 , h , T
E12x : 0
E21x
A2
E11y
p
2
:
Ep11x
2
A2
0
0
E11x
E011y
p
2
E11y
p
2
0
E11x
p2
A1 : E011x
p
2
0
E12y : 0
0 E11y
E21y
E11x
p
:
0
A2
0
0
E11y
p
2
0
D5h C5 , C2y , h
A11
E11x
p
2
:
0
E12x : 0
E21x
:
Ep11x
2
A22
:
Ep11x
0
0
E21y
E11x
p
:
0
0
0 A22
E1x
E12y : 0
E11y
p
2
E11y
p
2
A22
E1y
0
A22
0
E11y
p
2
0
0
0
0
E11y
p
2
0
261

Td , O
Tx
p
3
A1 :
0
Tz
p
3
Ty
p
3
0
Tx
p
2
Ex :
0
Ty
p
2
0
Tx
T1z :
p
T2x
0
:
Tx
T2z :
p
2
0
Ty
p
2
Ty
p
2
0
0
Tx
p
6
Ey :
0
T1y :
0
Tx
p
2
0
Ty
p
6
0
2Tz
p
6
Tz
0 p
2
0
0
0
0
0
Tz
p
2
Ty
p
2
0
Tz
p
2
0
0
T1x :
Ty
0 p
2
T2y
:
0
Tx
p
2
0
0
0
Tz
p
2
T means :
T2 in point group Td
T1 in point group O
0
262
The Raman Effect

C1v C , v , D1h
E1x
p
2
A1 :
0
E1x
E2x
0
E1y
p
2
0
A2 :
1x
E
p2
A1
0 0 A1
:
0 0 0
E1x 0 0
E1y
p
0
0
E
:
1x
p
0
0
2
0
0
0
P 1y
E
E2y
: 0
E1y
p
2
0
0
0
0
0 A1
0 E1y 0
E1x
0
p
E1y
:
0
p
2
0
0
The symbolism used in the tables should be understood in the following way. The
symmetry of the vibrational mode is indicated to the left of each tensor. The symmetries
written in the tensors means that the entry in that position is due to an intermediate
state of that particular symmetry, while the numerical factors show the quantitative relation between the tensor elements. Primes are used to distinguish quantities that are not
numerically related although they refer to intermediate states of the same symmetry. For
point groups where one conventionally uses primes and double primes to indicate the
behaviour of the representations under the symmetry operation h , we have used instead
an extra subscript. Thus in C5h , E12 corresponds to E001 in the conventional notation. For
point groups where the time inversion operator is added to the spatial operators, we have
indicated this with the symbol T.
7.8 TIME-DEPENDENT FORMULATION OF RESONANCE

RAMAN SCATTERING
7.8.1 Introduction
So far we have described vibrational resonance Raman scattering in terms of formulae
which were derived using perturbation theory. These formulae describe a steady state
scattering process and contain no explicit reference to time. They are also difficult to
use for the numerical calculation of resonance Raman excitation profiles except in the
case of small molecules. This is mainly because the formulae involve summations over
many states and such summations are almost impossible to perform for larger polyatomic
molecules.
A different approach involves a time-dependent formulation. This may be obtained
using straightforward mathematical procedures which transform a time-independent
263
formula into an integral over time. The time-dependent formulation affords an alternative
description of resonance Raman scattering and gives new insights into the phenomenon.
In the following sections we show how an appropriate mathematical transformation
applied to the time-independent AVI term gives the corresponding time-dependent formula
and then illustrate its application to particular cases.
7.8.2 Transformation of the AVI term to a time-dependent expression

VI
We begin by writing the expression for A eg vf :eg vi obtained from eq. (7.2.2) in the
following shorthand form
r
VI
eg vf :eg vi D K vf vi
7.8.1
Kr D p 0eg er p 0er eg
7.8.2
k k
where
and
rr
1 hvk jvrr
k ihvk jvk i
h vr er vr :eg vi 1 ier vrk
fg
vf vi D
k k
ig
7.8.3
We next replace the frequency terms in the denominator by energies and introduce a
simplified notation to facilitate the writing of the mathematical processes. We then have

hfjrihrjii
7.8.4
vf vi D
E

E

E

i
r
i
1
r

ig
fg
f
represents
v
,
and
represents the
where i represents vk , r represents vrr
k
k
r i; also,
r
in
the
electronic
state
je
summation over all vrr
k
Er D h er vr , Ei D h eg vi , E1 D h 1 and D h er vr
Equation (7.8.4) can be transformed into a time-dependent expression by using the
mathematical identity

i 1
iDt
1
D
exp
dt
7.8.5
D
h 0
h
where in this instance D is the energy denominator in eq. (7.8.4), that is
D D Er Ei E1 i
As a result of this transformation
1 we obtain
i
VI
hfjrihrjii expfiEr Ei E1 it/hgdt
A eg vf :eg vi D
k
k
h 0
r
7.8.6
7.8.7
Since
O hg
hrj expfiEr t/hg D hrj expfiHt/
O is the hamiltonian of the excited state, we can rewrite eq. (7.8.7) as
where H

i 1
AVI
O hjii expfiEi C E1 C it/hgdt
hfjrihrj exp iHt/
eg vf :eg vi D
h 0
r
7.8.8
7.8.9
264
The Raman Effect
O h as acting on the right, then

If we regard the propagator exp iHt/
O hgjii D jiti
expfiHt/
7.8.10
Finally we perform closure over the vibrational states r and obtain

i 1
VI
hfjiti expft/hg expfiEi C E1 t/hgdt
7.8.11
A eg vf :eg vi D
h 0
The quantity jhfjitij expft/hg can be described as the Raman overlap. It is the
product of the modulus of a time-dependent vibrational overlap integral jhfjitij which
involves the final state wave function jfi and the initial state wave function propagated on
the electronic surface, that is jiti and a damping function expft/hg which decreases
exponentially with time.
7.8.3 The time-dependent interpretation of resonance Raman

scattering
The time-dependent expression for AVI
eg vf :eg vi given by eq. (7.8.11) leads to an alternative interpretation of resonance Raman scattering which we now describe making use
of Fig. 7.14.
We begin by considering Fig. 7.14(a). The resonance Raman process begins from the
state jii which is a vibrational eigenstate of the ground electronic surface and is represented
by a harmonic potential function. At time zero, interaction with the incident radiation of
frequency 1 and photon energy E1 D h 1 (or equivalently wavenumber Q 1 and photon
energy E1 D hcQ 1 ) results in a vertical transition to the excited electronic surface through
the involvement of the electronic transition moment p 0er eg . The vibrational state jii is
O and it responds by beginning to evolve with
now under the influence of a hamiltonian H
time. If there is a significant displacement of the excited state potential energy surface
relative to the ground state potential energy surface, the wave packet jiti moves rapidly
from its initial position and reaches the right-hand side of the excited state potential. The
wave packet in this position is indicated by a dashed line in Fig. 7.14(a). The wave packet
is then reflected and moves back to positions of shorter internuclear separations. This
process of oscillation between the two classical turning points continues but is gradually
damped out by the damping function expft/hg. During the process of propagation with
time, the wave packet jiti crosses regions of large overlap with the wave function jfi
in the ground electronic state. In Fig. 7.14(a) maximum overlap occurs in a very short
time, of the order of a few picoseconds, after the start of the propagation. This initial
short period of propagation is represented in Fig. 7.14(a) by the arrow 0 ! 1 where the
point 0 corresponds to the equilibrium separation in the ground state and the point 1
corresponds to the internuclear separation at which a maximum overlap occurs with the
final state jfi D j1i. The Raman overlap jhfjitij expft/hg therefore has a maximum
at this point of time and this is shown in Fig. 7.14(b) which is a plot of the Raman
overlap against time. After this time the Raman overlap decreases as jiti moves well
away from the point 1 to the point 2 where it reaches a minimum. It is then reflected and
increases again, reaching a second, but smaller maximum, at point 3. At point 4, which
265
0
4
4
1
3
5
2
2
^ t/h}
|i (t) = {expiH
Energy
|i (0)
E1
Eg
|f > = |1>
|i > = |0>
(a)
Raman overlap |<f |i(t ) exp{Gt/h}
Raman overlap
|<f |i(t ) exp{Gt/h}
Vibrational periods
1
2
01
3 4 5 (time points in (a))

exp{Gt/h}
0
(b)
20
40
60
Time/femtoseconds
80
100
Intensity (half-Fourier
transform)2
Raman excitation profile

(Square of half-Fourier transform)
Excitation energy hc n1 (= hw1)
(c)
Figure 7.14 The time-dependent interpretation of Raman scattering. (a) Interaction of the
incident photon with the electronic transition moment sends the initial vibrational state jii
to the excited electronic surface where it is propagated by the excited state vibrational
hamiltonian. (b) The Raman overlap jhfjitij expft/hg as a function of time. (c) The
excitation profile as a function of the incident photon energy E1 D hc0 Q 1 D h 1 is given by
the square of the half-Fourier transform of the overlap between jiti and the final vibrational
level jfi.
266
The Raman Effect
corresponds to one complete cycle of vibration, the Raman overlap goes to zero because
here jiti is orthogonal to jfi D j1i. There then follows reflection at the left-hand side
of the excited state potential energy and as the wave packet retraverses the surface further
maxima, but of decreasing amplitude are produced until jiti is fully damped out. These
processes are shown for three complete vibrational periods in Figs. 7.14(a) and 7.14(b).
VI
The intensity of resonance Raman scattering will be proportional to jA eg vf :eg vi j2
VI
where in the case under consideration vi D 0 and vf D 1. Since A eg vf :eg vi as defined
in eq. (7.8.11) is obtained by integration over the Raman overlap function, the Raman
excitation profile can be obtained directly by a half-Fourier transformation of the Raman
overlap. The square of the half-Fourier transform of the overlap between jiti and the
final vibrational level jfi is shown in Fig. 7.14(c).
7.9 CONTINUUM RESONANCE RAMAN SCATTERING

We now consider the situation in which the resonant electronic state is a repulsive excited
electronic state or the repulsive part of a bound excited electronic state. Such a situation
is conveniently discussed by considering a specific example, which we have chosen to
be the bromine molecule Br2 . The energy as a function of internuclear distance for this
molecule is shown in Fig. 7.15.
30
Continuum
Wavenumber/103 cm1
25
20
Intermediate
state
Dissociation
limit
B(30+u)
15
n1
10
ns
1u
1
g+
0
150
300
200
250
Internuclear distance/pm
350
Figure 7.15 Illustration of continuum resonance Raman scattering for the bromine
molecule Br2 . The wavenumber of the incident laser radiation Q 1 is in resonance with the
continuous states of the pure repulsive 1 1u excited state and the repulsive part of the bound
B3 0C
state which is above the dissociation limit at around 20 000 cm1 .
267
If the resonance Raman scattering is excited with radiation of wavelength 457.9 nm

(or wavenumber 21 840 cm1 ) the intermediate state is above the dissociation limit of the
bound B3 0Cu state which is around 20 000 cm1 . Thus the continuum of the repulsive
part of this state as well as the continuum of the repulsive state 1 1u play a role in the
continuum resonance Raman scattering.
In this situation the intermediate states can be assumed to be infinitely narrow and so it
is not necessary to include the damping term in expressions for the transition polarizability
components.
We consider first the implications of this for formulae obtained on the basis of perturbation theory. If we put er vr D 0 in eq. (7.2.2) for the AVI term we have
rr
hv
1
jvrr
k vk jvk i
k
D p eg er p er eg
h
er vr :eg vi 1
r
fg
VI
A eg vf :eg vi
ig
7.9.1
vk
To accommodate the pole which occurs for the continuum states at exact resonance
eq. (7.9.1) has to be rewritten in the following form:
1 hvfg jvrr ihvrr jvig i

1
VI
k
k
k
k
P
A eg vf :eg vi D
ri dri
h
ri
1
0

fg rr
rr ig

ihvk jvk ihvk jvk iri
7.9.2
ri D1
where we have written ri for er vr :eg vi , ri is the density of continuum states and P
is the principal part of the integral.
In the case of the time-dependent formulation if we set D 0 eq. (7.8.11) becomes

1 1
AVI
hfjiti expfiEi C E1 t/hgdt
7.9.3
eg vf :eg vi D
h 0
Equation (7.9.2) or (7.9.3) can be used to calculate resonance Raman intensities of
simple molecules as a function of 1 , the frequency of the exciting radiation.
The time-dependent formulation can provide an instructive picture of continuum resonance Raman scattering as the following account will demonstrate. We will consider the
vibrational Raman band arising from the transition from the vibrational state jii D j1i to
jfi D j2i (the first hot band) in 79 Br2 excited by laser radiation of wavelength 457.9 nm
(or wavenumber 21 840 cm1 ). The evolution of the scattering process with time is illustrated in Fig. 7.16, the caption to which defines the electronic and vibrational states
involved. At t D 0 the transition from the initial state jii D j1i to the excited electronic
state occurs and this is indicated by the vertical arrow in Fig. 7.16(a). Then the wavepacket
jiti propagates mainly on the repulsive 1 1u state until delay time t. The case when
t D 20 fs is shown in Fig. 7.16(b), and the case when t D 40 fs in Fig. 7.16(c).
It can be seen that the wavepacket jiti only overlaps with the final state jfi D j2i
for a very short propagation time. Only during this time has the function jiti a nonvanishing overlap with the final vibrational state jfi D j2i in the ground electronic state
268
The Raman Effect
and so gives rise to continuum resonance Raman scattering. It can be inferred from the
sequential diagrams in Fig. 7.16 that the most efficient resonant Raman transitions occur
within ultrashort times with an average delay time of about 14 fs for this transition.
The behaviour we have just described is quite different from that occurring when
bound excited states are involved. In such cases, as already explained and illustrated in
0.2
0 Femtoseconds
|1(t )
0.1
|2
|1
0
(a)
0.2
20 Femtoseconds
Energy/atomic units
|1(t )
0.1
|2
|1
0
(b)
0.2
40 Femtoseconds
|1(t)
0.1
|2
|1
0
4
(c)
Internuclear distance/atomic units
Figure 7.16 The time-dependent interpretation of continuum resonance Raman scattering

for 79 Br2 . The wavelength of the excitation radiation is 457.9 nm and the vibrational Raman
transition jvi D 1i to jvf D 2i is considered. The figures (a), (b) and
(c) show the potential
energy curves of the relevant electronic states, ground state X D 1 C
g (continuous line)
and excited states B D 3 0Cu (dashed line) and D 1 1u (dotted line) together with the
absolute values of the coordinate representations of the initial vibrational state jii D j1i,
final vibrational state jfi D j2i and the time-dependent state jiti at times t D 0, 20 and
40 fs, respectively. The excitation and de-excitation processes and the related unimolecular
dissociations are represented schematically by vertical and horizontal arrows. (The energy
gap between the states jii and jfi has been expanded for clarity of representation.).
269
(a)
10 20 30 40
Delay time / fs
50
60
Raman intensity
|f = 6| i(t) = 0 (t )|
Fig. 7.14, the wavepacket oscillates, with damping, within the classical turning points of
the potential surface. However, in continuum resonance Raman scattering the molecule
either dissociates or returns to the ground electronic state as a result of Raman scattering.
It follows that only the accelerated propagation away from the repulsive functions needs
to be considered for the overlap function jhfjitij. Non-vanishing vibrational overlaps
only occur within times of the order of 1020 fs and thus damping plays no part and
expft/hg 1.
It is instructive to consider how the overlap integral depends on time for a resonance
Raman vibrational transition in a diatomic molecule. Figure 7.17(a) shows the absolute
value of the overlap function jhf D 6jit D 0tij for the vibrational transition for the
iodine molecule I2 . This gives rise to the fifth overtone in the resonance Raman spectrum
of I2 when excited above the dissociation limit of the B3 0Cu electronic state with
radiation of wavelength 4880 cm1 (or wavenumber 20 490 cm1 ). The highest value
of the time-dependent overlap integral is reached around 16 fs after the molecule has
been excited into the dissociation continuum of the B state. Although this vibrational
transition corresponds to an overtone, the function hf D 6jit D 0ti is nearly identical to the wavefunction of the initial state ji D 0i from which the resonance Raman
scattering process takes place. Figure 7.17(b) shows the excitation profile of the Raman
transition which is obtained simply by making the half-Fourier transform of the timedependent overlap integral hfjiti and plotting this as a function of energy. This function
profile has its maximum around 19 000 cm1 which corresponds to Q l 6Q k where
Q 1 D 20 490 cm1 and Q k for 127 I2 D 215 cm1 .
The time-dependent formulation provides a useful explanation of why many overtones
are observed in the continuum resonance Raman scattering of diatomic molecules. During
the short period of propagation of jiti the wavepacket moves across regions which
have large overlaps with higher vibrational levels jfi in the ground electronic state. The
vibrational wave functions for these higher vibrational levels have large values in the
neighbourhood of their classical turning points, particularly at the outer turning point of
an anharmonic oscillator.
15000
20000
25000
Absolute wavenumber / cm-1
(b)
Figure 7.17 Continuum resonance Raman scattering in 127 I2 for the vibrational transition
ji D 0i to jf D 6i, involving the B3 0Cu electronic state. (a) shows jhf D 6jit D 0tj as
a function of time. (b) shows the excitation profile of this transition as a function of absolute
wavenumber. The maximum is at 19 000 cm1 .
270
The Raman Effect
0 fs 15 fs
40 fs
Energy
|i(t)
|i =|0
| f |1
|f =|3
Figure 7.18 Generation of the overtone pattern in continuum resonance Raman scattering
of a diatomic molecule. The propagating wavepacket brought up from the initial vibrational
level in the electronic ground state into the excited electronic state evolves rapidly with
time. While passing over regions with high vibrational overlaps the molecule can de-excite
to the various vibrational levels of the electronic ground state. Such overtone transitions are
schematically shown for v D 3 (second overtone) and an overtone with a large vibrational
quantum number change (v 1). The energy gap between state j3i and state j0i has been
expanded for clarity of presentation.
The situation is illustrated in Fig. 7.18. The upward arrow shows the vertical transition from the ground electronic state to the excited electronic state resulting from the
interaction of the incident radiation with the molecule. This process is controlled by the
transition electric dipole moment associated with the two electronic states. The propagating wavepacket jiti evolves rapidly with time in the excited electronic state and as
it passes over regions with high vibrational overlaps the molecule can be de-excited by
transitions to various vibrational levels of the ground electronic state, giving rise to overtone transitions. Two such overtone transitions are illustrated in Fig. 7.18. At 15 fs there
is high overlap with the vibrational level jfi D j3i and at 40 fs there is overlap with a
vibrational level for which jfi is rather large, v 1.
REFERENCES
Manneback, C. (1951). Physica 17, 1001.
Mortensen, O.S. and Hassing, S. (1980). Advances in Infrared and Raman Spectroscopy, (eds.),
R.J.H. Clark and R.E. Hester, 6, 1. Heyden: London.
271
8
Rotational and
VibrationRotation Resonance
Raman Scattering
That night we went to Birmingham
by way of Beachy Head.
G. K. Chesterton
8.1 INTRODUCTION
This chapter considers resonance rotational and vibrationrotation Raman scattering for
symmetric top molecules. Our objective is to derive expressions for the isotropic, antisymmetric and anisotropic contributions to the intensities of the O, P, Q, R and S resonance
rotational Raman transitions.
The treatment that follows is necessarily more complicated than that for normal nonresonance rotational and vibrationrotation Raman scattering. This is because in the
resonance case we cannot put er vr Rr :ei vi Ri D er vr :ei vi and er vr Rr D er vr as we did in
the non-resonance case; and consequently separation of the rotational terms from the
vibrational terms in the transition polarizability is no longer possible.
The treatment in this chapter uses 3-j and 6-j symbols quite extensively. The general
theory underlying these symbols is dealt with in Chapter A19. Specific formulae are
introduced without proof in the treatment that follows.
272
The Raman Effect
8.2 GENERAL EXPRESSION FOR alm fi FOR A SYMMETRIC

TOP MOLECULE
As explained in Chapter 6, rotational transitions are best treated using irreducible tranj
sition polarizability tensor components m fi because of their simple properties under
j
rotational transformation. For the present purpose it is convenient to introduce the m fi
through their relationships to the spherical transition polarizability components fi
using the following equation:

1
1 j
j
m
1/2
m
D
1 2j C 1
fi
8.2.1
m
fi

where, as explained in Chapter A14, Section A14.3.1

fi D 1 fi
8.2.2
The use of the unsigned vector dyads avoids the phase factor 1 and simplifies
eq. (8.2.1). The relations of the signed and unsigned vector dyads to 3-j coefficients
and C-coefficients is discussed in Chapter A19 (see particularly pages 541 and 542).
It follows from eq. (5.2.9), that when the non-resonant term is neglected

1 hfjp jrihrjp jii
fi D
8.2.3
h r6Di,f ri 1 ir
We now invoke the BornOppenheimer approximation and put
jii D jeg ijvi ijJi Ki Mi i
8.2.4
jri D jer ijvr ijJr Kr Mr i

g
jfi D je ijv ijJ K M i
8.2.5
8.2.6
where we have taken the initial and final electronic states to be the ground electronic state
jeg i. We recall that jer i characterizes the intermediate electronic state; that vi , vr and vf
are vibrational quantum numbers for the initial, intermediate and final vibrational states
respectively, and that Ji Ki Mi , Jr , Kr , Mr and Jf Kf Mf are the three angular momentum
quantum numbers characterizing a symmetric top molecule for the initial, intermediate and
final states, respectively. In the absence of a magnetic field the M levels are degenerate
and so
and
ri D er vr :eg vi C Jr Kr :Ji Ki
8.2.7
r D
8.2.8
r r
er vr J K
See footnote on page 283 of this Chapter. This discusses the definition of .
273
Introducing eqs. (8.2.4) to (8.2.8) into eq. (8.2.3) we obtain

1
fi D
h er vr Jr Kr Mr

r r r
r r r
i i i
hJf Kf Mf jhvf jheg jp jer ijvr ijJ K M ihJ K M jhvr jher jp jeg ijvi ijJ K M i
er vr :eg vi C Jr Kr :Ji Ki 1 ier vr Jr Kr

8.2.9
with the restriction r 6D i, f implicit.
In eq. (8.2.9) the spherical vector components p and p are defined relative to a
space-fixed spherical basis system. The next step is to relate p and p to the spherical
vector components p0 and p0 which are defined relative to a molecule-fixed spherical
basis system. The necessary transformations are of the forms

D0 p0
8.2.10
p D
0
and
p D
D0 p0
8.2.11
0
where the D0 are direction cosines defined in Chapter A9, Section A9.4 and Table A9.1.
Introducing these transformations into eq. (8.2.9) we have
1
fi D
h 0 0 r r

e v
r r r
JK M
hJf Kf Mf jhvf hjeg jD0 p0 jer ijvr ijJ M K ihJ M K jhvr jher jD0 p0 jeg ijvi ijJ K M i
8.2.12
The wave function of a symmetric top can be expressed in the form

2J C 1 1/2 J
DMK
jJKMi D
82
8.2.13
J
where DMK
is an element of a rotation matrix. Introducing eq. (8.2.13) into eq. (8.2.12)
we obtain
1
r
82 2 2Ji C 11/2 2J C 12Jf C 11/2
fi D
h 0 0 r r
e v
r r r
JKM
J
J
J
J
0
f
g
r
r
r
r
g
i
DM
f Kf D0 DMr Kr DMr Kr D0 DMi Ki hv jhe jp0 je ijv ihv jhe jp0 je ijv i

8.2.14
274
The Raman Effect
8.3 GENERAL EXPRESSION FOR ajm fi

j
We can now obtain an expression for the m fi by using eq. (8.2.14) with eq. (8.2.1).
J
If we also introduce the properties of the DMK
transformations, namely that

f

i
f Mf
Jf 1 Ji
Jf 1 Ji
J
J
K
2
DKf Mf Dmm0 DKi Mi D 1
8
Kf m0 Ki
Mf m Mi
8.3.1
we obtain the somewhat formidable result that

1
f
f
r
r

D
1K M K M Cm
j
fi
m
h
0
0 r r r r r i f
,, , ,e v J K M M M
1/2
i
1/2
2j C 1 2J C 1 2Jr C 12Jf C 11/2

1
1 j
Jf
1
Jr
Jf 1 Jr
m
Kf 0 Kr
Mf Mr
r

J
1 Ji
Jr
1 Ji
p0 eg vf :er vr p0 er vr :eg vi

Kr 0 K
Mr Mi
er vr :eg vi C Jr Kr :Ji Ki 1 ier vr Jr Kr 1
8.3.2
In eq. (8.3.2) we have introduced a condensed notation for the components of the dipole
transition vectors in the spherical basis and written, for example,
p0 eg vf :er vr D hvf jheg jp0 jer ijvr i
8.3.3
We have also made use of the fact that the M levels are normally degenerate and introduced a sum over all values of Mi and Mf . Thus eq. (8.3.2) gives all contributions to
j
m fi at the frequency associated with the transition from vi Ji Ki to vfJfKf .
8.4 CONTRACTION OF GENERAL EXPRESSION FOR ajm fi

The next stage in the development of eq. (8.3.2) is to perform summations over , and
j
Mr and obtain a contracted form of m fi in which the three Wigner 3-j functions in
eq. (8.3.2), involving , and Mr , are replaced by the product of a Wigner 3-j function
and a Wigner 6-j function neither of which involves , or Mr . There are, of course,
consequential changes in the exponent of the factor 1. The contracted expression is
1
r
f
r
j
1J CM K 2j C 11/2 2Ji C 11/2 2Jf C 11/2 2Jr C 1
m fi D
h 0 0 er vr Jr
Kr Mi Mf
j
Jf
m Mf
Ji
Mi

j Jf
Jr
1
Ji
1

Jf
1
Jr
Jr
1 Ji
p0 eg vf :er vr p0 er vr :eg vi
Kf 0 Kr
Kr 0 Ki
er vr :eg vi C Jr Kr :Ji Ki 1 ier vr Jr Kr 1
275
8.4.1
where f g indicates a Wigner 6-j symbol.
8.5 THE QUADRATIC TERM

j
To obtain expressions for intensities we must form the quadratic term jm fi j2 . In the
absence of an external field the Mi and Mf levels are degenerate. Thus we can sum over
all Mi and all Mf using the following orthogonality property of the Wigner 3-j symbols
j
when forming jm fi j2 :
f

Jf
1
Ji
j
J
J
j0
jj0 mm0
D
8.5.1
f Mi m
f Mi m0
M
M
2j
C
1
i
f
M
To simplify somewhat the resulting expression we assume that only one vibronic band
er vr is dominant in the resonance. The sum over er vr can then be omitted. On this basis
we obtain the following result:

1
2
jj
2Jr1 C 12Jr2 C 12Ji C 12Jf C 1
m fi j D 2
r
r
r
r
h 0 0 J J K K
1 2 1 2

f

j Jf Ji
j Jf Ji
J
1
Jr1
Jr 1 1
Jr2 1 1
Kf 0 Kr1
1f

r

J
1
Jr2
Jr1
1 Ji
J2
1 Ji
Kf 0 Kr2

Kr1 0 Ki
Kr2 0 Ki
jp0 eg vf :er vr j2 jp0 er vr :eg vi j2
er vr :eg vi C Jr1 Kr1 :Ji Ki 1 ier vr Jr1 Kr2 1
er vr :eg vr C Jr2 Kr2 :Ji Ki 1 C ier vr Jr2 Kr2 1
8.5.2
In forming the quadratic term, given by eq. (8.5.2), account has been taken of the
possible contribution of cross-terms resulting from the summation of intermediate resonant
rotational levels characterized by Jr1 , Kr1 and Jr2 , Kr2 . Such cross-terms imply that there
is more than one rotational pathway at the amplitude level for a given overall transition
giving rise to scattering. We shall discuss this in more detail later. In eq. (8.5.2) it has
been assumed for simplicity that the polarization indices p0 and p0 are the same for
both pathways, but the treatment can be adapted to the general case where the polarization
indices are different for each pathway.
Before examining the implications of eq. (8.5.2) we make one further simplification.
We shall consider the resonance to be associated with an electronic transition polarized
along the top axis, which by convention is the z axis. Thus in eq. (8.5.2) p0 D p0 D p00 ,
and 0 and 0 may be removed from the summation.
276
The Raman Effect
8.6 SELECTION RULES

The triangular condition on the upper row of the 6-j symbols allows isotropic scattering
(j D 0) to contribute to the Q branch (J D 0) intensity only; antisymmetric scattering
(j D 1) to contribute Q, P and R branch (J D 0, 1) intensities only and anisotropic
scattering (j D 2) to contribute to Q, P, R, O and S branch (J D 0, 1 and 2)
intensities. The summation condition on the lower rows of the 3-j symbols allows only
K D 0 transitions for 0 D 0 D 0. The triangular condition on the upper rows of the
3-j symbols allows intermediate or resonance values of Jr which differ by 1 or 0 from
Ji or Jf . These conditions on Jr lead to the existence of more than one route from Ji to
Jf in some branches. For example, for the Q branch (overall J D 0), three routes are
possible:
Ji ! Jr D Ji ! Jf D Jr D Ji
Ji ! Jr D Ji C 1 ! Jf D Jr 1 D Ji
and
Ji ! Jr D Ji 1 ! Jf D Jr C 1 D Ji
These routes may be described as QQ, RP and PR, respectively. The allowed values
of j in i
m fi for the Q, P, R, S and O branches and the various permitted routes for
these branches are given in Table 8.1. The permitted routes for a Q branch transitions are
illustrated in Fig. 8.1.
Table 8.1 Resonance rotational Raman scattering: allowed
j
j values in am
and permitted route(s) for the Q, P, R, S and
O branches.
Branch
Allowed j values in j

m
Permitted route(s)
Q
P
R
S
O
0, 1, 2
1, 2
1, 2
2
2
QQ, PR, RP
PQ, QP
RQ, QR
RR
PP
8.7 EVALUATION OF jajm fi j2

j
We now illustrate how the jm fi j2 are evaluated by considering the case of an S branch
for which J D C2, K D 0 and the resonant electronic transition is polarized along the
molecular z axis so that 0 D 0 D 0. For this case there is only one route in the scattering
amplitude namely of the type RR, that is Ji ! Jr D Ji C 1 ! Jf D Jr C 1 D Ji C 2,
so that Jr1 D Jr2 . Also, of course, Kf D Kr D Ki .

Jr = 3
Jr = 3
Jr = 3
|e r
|er
|er
J =2
Jr = 2
Jr = 2
r
J =1
r
J =1
r
J =1
r
J =0
r
J =0
r
J =0
277
J=3
J=3
J=3
|eg
|eg
|eg
J=2
J=2
J=2
J=1
J=1
J=1
J=0
J=0
QQ: J i = 1 J r = 1 1 J f = 1
J=0
PR: J i = 1 J r = 0 J f = 1
RP: J i = 1 J r = 2 J f = 1
Figure 8.1 QQ, PR and RP routes for one Q branch (J D 0) transition. The rotational
states Jr related to the intermediate electronic state jer i. The rotational states J relate to the
initial and final electronic state jeg i.
It follows from eq. (8.5.2) that for 2

m Ji C2,Ki :Ji ,Ki we have

2

2

D 1
2Jr C 12 2Ji C 12Jf C 1
m Ji C2,Ki :Ji ,Ki
h 2
Jr ,Kr
2
Jr
Jf
1
Ji
1
2
Jf
Kf
1 Jr
0 Kr
2
Jr
Kr
1 Ji
0 Ki
2
jp00 eg vi :er vr j2 jp00 er vr :eg vi j2

[er vr :eg vi C Jr Kr :Ji Ki 1 2 C e2r vr Jr Kr ]1
8.7.1
Formulae for the Wigner 3-j and 6-j symbols which occur in eq. (8.7.1) are given
r
i
r
i
below for the case of the RR route, with the relations J D J C 1, K D K ; and Jf D
r
i
r
i
J C 1 D J C 2 and Kf D K D K introduced at appropriate stages.

2
2
[Ji C 12 Ki ]
Jr 1 Ji
8.7.2
D
Kr 0 Ki
2Ji C 32Ji C 1Ji C 1
278
The Raman Effect

Jf
Kf
and
1 Jr
0 Kr

2
Jf
1
2
Jr
[Ji C 22 Ki ]
D
Ji
1
1
52Ji C 3
8.7.3
8.7.4
Using these results in eq. (8.7.1) we obtain

2
i
2
i2
i
2
i2
2

D 1 [J C 1 K ][J C 2 K ]
m Ji C2,Ki :Ji ,Ki
h 2

[er vr :eg vi C Jr Kr :Ji Ki 1 2 C e2r vr Jr Kr ]
This result is independent of m and since for j D 2 there are 5 values of m

2
2
2
G 2 i
D
j2
m Ji C2,Ki :Ji Ki j D 5jm Ji C2,Ki :Ji Ki j
i i i
J C2,K :J K
8.7.5
8.7.6
At this point it is convenient to express the denominator of eq. (8.7.5) in terms of

wavenumbers
Q . Thus an alternative form of eq. (8.7.5) is

2
i
2
i2
i
2
i2
2

D 1 [J C 1 K ][J C 2 K ]
m Ji C2,Ki :Ji Ki
hc0 2
8.7.7
[Q
er vr :eg vi C
Q Jr Kr :Ji Ki
Q 1 2 C e2r vr :Jr Kr ]
We now introduce into the denominator of eq. (8.7.7) the rotational term for a symmetric
Q v J, K for a prolate symmetric
top in the vibrational state v. We use the rotational term F
top given by eq. (6.7.15). This neglects centrifugal stretching terms and is repeated below,
in the notation used in this Chapter J D Ji , K D Ki
Q v J, K D Bv Ji Ji C 1 C Av Bv Ki2
F
8.7.8
The case under consideration involves an RR route with Jr D Ji C 1 and Kr D Ki .

Thus
Q Jr Kr :Ji Ki D
Q Ji C1,Ki :Ji Ki and the transition wavenumber
Q Ji C1,Ki :Ji Ki may be obtained
Q i C 1, Ki FJ
Q i , Ki using eq. (8.7.8). We use the result to express the whole
from FJ
denominator in the form
8.7.9
2R C R2
where
R D
Q er vr :eg vi C Bv0 Bv00 J2 C 3Bv0 Bv J C 2Bv0
8.7.10
and R is the lifetime of the intermediate state er vr Jr Kr . Both R and R as defined here
have units of cm1 .
279
The other irreducible transition tensor components can be evaluated in a similar way.
The full results are given in Table 8.2 for resonance involving one vibronic band with the
resonant transition polarized along the figure axis z0 of the symmetric top, so that 00 D z0 .
The results in Table 8.2 can be shown to reduce to the Placzek-Teller b factors for the
limiting case of no resonance. The procedure is to set P D Q D R so that the differing
dependencies of the denominators of the rotational energies of the intermediate state are
removed. When this simplification is introduced, all the antisymmetric j D 1 transition
tensor components vanish; all the isotropic components j D 0 reduce to 1/3 2J C 1 as
the sum over all possible intermediate levels Jr is complete; and the anisotropic j D 2
components now involve only the initial rotational level quantum numbers and reduce to
the Placzek-Teller factors in this case bJ2
f Ki :Ji Ki .
8.8 INTENSITIES AND DEPOLARIZATION RATIOS

Formulae for intensities and depolarization ratios can be obtained by using the above
j
expressions for m fi together with the following relationships

1
I/2; ?s , ?i D k
Q Nvi Q
s 4 30
10G 0 fi C 4G 2 fi I
8.8.1

1
5G 1 fi C 3G 2 fi I
8.8.2
I/2; jjs , ?i D k
Q Nvi Q
s 4 30
and
/2; ?i D
5G 1 fi C 3G 2 fi

10G 0 fi C 4G 2 fi
8.8.3
j
The Placzek invariants G j fi are related to the m fi as follows:

2
G j fi D
jj
m fi j
8.8.4
and
2
j
2
jj
m fi j D 2j C 1jm fi j
8.8.5
We shall not make the necessary substitutions as rather cumbersome formulae would
result; these are best handled numerically using a computer program. The results of such
calculations can be used to illustrate the properties of intensities and depolarization ratios
of typical resonance rotational and vibration-rotation bands. Some specific examples are
considered in the section which follows.
The expressions in Table 8.2 show that the relative rotational Raman intensities will
depend on the interplay between the terms and . The term measures the difference
between the resonance absorption wavenumber
Q er vr :eg vi C
Q Jr Kr :Ji Ki and 1 . This is often
referred to as the detuning from a particular resonance absorption frequency. If D 0
there is exact resonance and the denominator is equal to 2 . This is the smallest possible
value of the denominator and would correspond to a maximum in a plot of the relative
Chapter 5 and Reference Table 5.2(a) deal with these relationships.
J C K2 J K2 2J C 3J C 1 2

P C P2 1
30J3 2J 12J C 1

J2 K2 J C K C 1J K C 1
R P C P R
C
15JJ C 12J C 1
R P C P R 2 C R P 2 P R

K2 2J 1J C K C 1J K C 1
Q R C Q R

15JJ C 13
Q R C Q R 2 C Q R 2 Q R

K2 2J C 3J2 K2
Q P C Q P

15J3 J C 1
Q P C Q P 2 C Q P 2 Q R
(continued )
K4 2J C 12J 12J C 3 2

J2J 1J C K C 12 J K C 12 2
2 1

C

C
R C R2 1
Q
Q
30J3 J C 13
302J C 12J C 3J C 13
J2 K2 [J 12 K2 ] 2

P C P2 1
5JJ 12J 1
2
2
2
A. Anisotropic component j2
m Jf Ki :Ji Ki j jp0 eg vi :er vr j jp0 er vr :eg vi j
[J C 12 K2 ][J C 22 K2 ] 2

R C R2 1
5J C 1J C 22J C 3

J 1
2P Q C P Q
K2 J2 K2
J C 1
C
C
10J3
J 12P C P2 J C 12Q C Q2 P Q C P Q 2 C P Q 2 P Q

K2 J C K C 1J K C 1
J C 2
J
2Q R C R Q
C
C
10J C 13
J2Q C Q2 J C 22R C R2 Q R C R Q 2 C Q R 2 R Q
Rotational
transition
Table 8.2 Contributions of irreducible tensor components to intensities of resonance rotation Raman transitions. (only one vibronic
band involved; ef D eg and vf D vi , for an electronic transition polarized along z, the top axis, 0 D 0 D 00 D z0 ).
280
The Raman Effect
Rotational
transition
K4 2J C 1 2
J2 K2 J C 1 2
2 1

C
P C P2 1 C J[J C 12 K2 ]2 2R C R2 1
Q
6J C 13 J3 Q
6J3 2J C 1

K2 J2 K2
Q P C Q P
C
3J2 J C 1 Q P C Q P 2 C Q P 2 Q P

K2 [J C 12 K2 ]
Q R C Q R

3JJ C 13

J2 K2 [J C 12 K2 ]
P R C P R

3J C 1J2J C 1
P R C P R 2 C P R 2 P R

K2 J2 K2
2P Q C P Q
2
2 1
2
2 1

C

C

C

P
P
Q
Q
6J3
P Q C P Q 2 C P Q 2 P Q

K2 J C K C 1J K C 1
2R Q C R Q
2
2 1
2
2 1

C

C

C

Q
Q
R
R
6J C 13
R Q C R Q 2 C R Q 2 R Q
2
2
2
B. Antisymmetric component jj

281
2J C 1K4 2
J2 K2 2
[J C 12 K2 ]2
2 1
2
2 1

2 C R2 1

C

C
C

C
Q
P
3J2 J C 12 Q
3J2 2J C 1 P
3J2 J C 12 2J C 1 R

2K2 J2 K2
Q P C Q P
C
3J2 J C 1
Q P C Q P 2 C Q P 2 Q P

2K2 [J C 12 K2 ]
Q R C Q R
C
3JJ C 12

2J2 K2 [J C 12 K2 ]
P R C P R
C
32J C 1JJ C 1
P R C P R 2 C P R 2 P R
2
2
2
C. Isotropic component j0
R D
Q er vr :eg vi C Bv0 Bv00 J2 C 3Bv0 Bv00 J C 2Bv0 C [A0v Bv0 A00v Bv00 ]K2
Q 1
Note that in order to simplify the nomenclature, Ji in the text is replaced by J in this Table.
Q D
Q er vr :eg vi C Bv0 Bv00 JJ C 1 C [A0v Bv0 A00v Bv00 ]K2
Q 1
P D
Q er vr :eg vi C Bv0 Bv00 J2 Bv0 C Bv00 J C [A0v Bv0 A00v Bv00 ]K2
Q 1
The symbols P , Q and R are defined below for a prolate symmetric top. The corresponding expressions for an oblate symmetric top may be
obtained by replacing A0 by C0 and A00 by C00 .
Rotational
transition
282
The Raman Effect
283
intensity distribution of a particular vibrationrotation band against . As increases,

that is the excited state lifetime gets shorter, a larger number of resonance enhanced
rotational transitions are observed. In the limit when is very large, that is P,Q,R ,
the classical pattern of rotational Raman intensities is obtained. As decreases, that is the
excited state lifetime increases, the intensities of a fewer number of rotational transitions
appear enhanced. In addition the presence of antisymmetric contributions j D 1 to P,
R and Q branches means that anomalous depolarization ratios will be observed.
In some respects the calculation of relative resonance rotational intensities for a vibrational band is simpler than the calculation of resonance vibrational Raman intensities.
This is because the expressions for resonance rotational intensities do not involve vibronic
coupling parameters. We see from eq. (8.5.2) that once a particular rotational level charj
acterized by Ji Ki is chosen all the numerators of m fi are determined.
The pattern of intensities in a particular resonance rotational branch is a sensitive
function of and this is illustrated in the following section.
8.9 AN ILLUSTRATIVE EXAMPLE

We now consider as an illustrative example some aspects of the pure rotational and
vibrationrotation resonance Raman spectra of ammonia, NH 3 , taken from the work of
Ziegler (1986). Figure 8.2 shows the calculated relative intensities in the pure rotational
J D 0 resonance Raman spectrum of NH 3 for four choices of excited state parameters: (a) D 10 cm1 , D 0.26 ps; (b) D 28.5 cm1 , D 0.093 ps; (c) D 50 cm1 ,
D 0.053 ps; and (d) D 250 cm1 , D 0.011 ps. The parameter was taken to be
constant throughout the resonant vibronic band. The excitation radiation had a wavelength of 208.8 nm (47 890 cm1 which is in resonance with a vibronic feature in the
ultra-violet absorption spectrum of NH 3 associated with v02 D 2 where v02 refers to the
vibrational quantum of the symmetric bending vibration of NH 3 in the upper electronic
state. The absorption spectrum of NH 3 at 298K, arising from the transition from the
Q 00 , is shown in Fig. 8.3
Q 01 to the excited electronic state AA
ground electronic state XA
2
for the wavelength range 190220 nm. The resonance Raman process is illustrated in
Fig. 8.4. The calculated resonance Raman spectra in Fig. 8.2 have been appropriately
broadened using a Gaussian function to represent the spectral band pass of the detection
system. It is evident from Fig. 8.2 that the pattern of intensities is a sensitive function of
and . As the lifetime gets shorter and approaches a single vibrational period, more
rotational transitions appear with comparable intensity and the pattern of relative J 6D 0
intensities approaches the off-resonance pattern. The increasing number of rotational scattering transitions that appear with decreasing excited state lifetime reflects the larger
importance of the damping constants in the polarizability denominators as decreases.
An excellent fit to the observed resonance rotational Raman spectrum of NH 3 is obtained
for D 28.5 cm1 which determines the excited state lifetime to be 0.093 ps.
The formulae used throughout this chapter and the values given here relate to the definition of the full
width as 2 as discussed in Chapter 4, p. However, Ziegler (1986) uses as the full width so that his
values are twice those given here and his formulae reflect the different definition.
284
The Raman Effect

(a) t = 0.26 ps
Relative intensity
(b) t = 0.093 ps
(c) t = 0.053 ps
(d) t = 0.011 ps
200
100
0
1
n/cm
100
200
300
Figure 8.2 Calculated resonance rotational Raman spectrum (Stokes and anti-Stokes) with
208.0 nm excitation for four values of resonant excited state lifetime .
200
205
Wavelength/nm
212.8 nm
Absorbance
195
~
X
208.8 nm
~
A
204.2 nm
NH3 absorption
210
215
220
Figure 8.3 Absorption spectrum of NH 3 . The arrows indicate some excitation wavelengths
that can be used to obtain resonance Raman spectra.
285
v2 = 4
v2 = 3
v2 = 2
v2 = 1
v2 = 0
46,990 cm1
v2 = 4
v2 = 3
v2 = 2
v2 = 1
v2 = 0
X
(a) Vibronic levels involved in resonance for excitation at 212.8 nm. v 2 and v 2 are vibrational quantum
numbers in the X and A electronic states of the n2(a1) symmetric bending mode of NH3 (n2 = 960 cm1).
Jr = 3
Jr = 3
Jr = 3
|e r
|er
|er
Jr = 2
Jr = 2
Jr = 2
Jr = 1
Jr = 1
Jr = 1
J =0
J =0
Jr = 0
J=3
J=3
|eg
J=3
|eg
|eg
J=2
J=2
J=2
J=1
J=1
J=1
J=0
J=0
QR route for R branch J =1

Ji = 1 Jr = 1 J f = 2
RQ route for R branch J =1

Ji = 1 Jr = 2 J f = 2
J=0
RR route for S branch J = 2
Ji = 1 Jr = 2 J f = 3
(b) Examples of rotational levels involved in R and S branches:
Figure 8.4 Diagrammatic illustrations of (a) vibronic levels and (b) rotational levels
involved in the resonance rotation Raman process.
286
The Raman Effect
It is of particular interest to calculate the relative values of the anisotropic j D 2 and

antisymmetric j D 1 contributions to the intensity and these are shown in Fig. 8.5 for
the best fit case of D 28.5 cm1 ( D 0.093 ps). The antisymmetric j D 1 contribution
is 25% of the total intensity.
We now consider the effect of changing the excitation wavelength from 208.8 nm to
212.8 nm 46 990 cm1 which is in resonance with the vibronic feature associated with the
Q
Q electronic transition. At this excitation wavelength
v02 D 1 absorption band of the A
X
the observed rotational bands J 6D 0 in the vibrationrotation Raman spectrum of
NH 3 are fitted best with D 16 cm1 that is an excited state lifetime of 0.16 ps. In this
instance 29% of the total intensity arises from the antisymmetric j D 1 contribution.
This is larger than for the case of 208.8 nm excitation and indicates that the importance
of the antisymmetric j D 1 contribution increases as decreases (or increases).
The depolarization ratio, /2; ?i , of the J 6D 0 bands depends on the relative contributions of the anisotropic j D 2 and antisymmetric j D 1 terms in eq. (8.8.2). As
decreases ( increases) the antisymmetric contribution increases and f/2; ?i 3/4g
increases. As increases ( decreases) the antisymmetric j D 1 contribution decreases
and /2; ?i approaches 3/4, the limiting value when there is no antisymmetric contribution.
We have so far considered only the J 6D 0 transitions which are associated with the
R and S branches in the pure rotational Raman spectrum of a symmetric top. These can
only have anisotropic and/or antisymmetric contributions. We now consider the Q branch
J D 0 transitions which arise in rotationvibration Raman spectra. This can have
isotropic j D 0 contributions as well as anisotropic j D 2 and antisymmetric j D 1
contributions.
The intensity of a Q branch also shows appreciable sensitivity to (and ) but the
sensitivity of the depolarization ratio f/2; ?i g to (and ) depends on the range of
values in which lies. As Fig. 8.6 shows, when < 50 cm1 , /2; ?i is relatively
Relative intensity
(a) Antisymmetric ( j = 1) contribution
(b) Anisotropic ( j = 2) contribution
200
100
0
1
n/cm
100
200
300
Figure 8.5 Calculated relative antisymmetric (a) and anisotropic (b) contributions to the
best-fit resonance rotational Raman spectrum with 208.8 nm excitation (see Fig. 8.2(b),
D 0.93 ps).
287
.50
150
.40
120
.30
(a)
90
60
.20
(b)
Intensity/arbitrary units
r(p/2; i)
30
.10
0
10
100
0
1000
G /cm1
Figure 8.6 Calculated (a) depolarization ratio /2; ?i and (b) integrated Q branch intensity for a Q branch of the Raman spectrum of NH 3 with 208.0 nm excitation as a function
of .
insensitive to ; but, when lies between 50 and 150 cm1 , /2; ?i exhibits a
degree of dependence on . However, for values in the range 5500 cm1 , whereas
the intensity changes by a factor of a hundred, the depolarization ratio changes only
by a factor of three. When is large ( small) the antisymmetric j D 1 contribution
decreases and in the limit /2; ?i for the Q branch tends to a value of 1/8 for a
z-polarized electronic state and K D 0.
8.10 CONCLUDING REMARKS

This treatment of resonance rotational Raman scattering shows that when closure over the
rotational states Rr is prevented by appropriate resonances, new pathways for rotational
Raman scattering are opened. The resulting Raman spectra can be analysed using the
theory developed here to yield important information on lifetimes of excited states.
REFERENCE
Ziegler, L. D. (1986). J. Chem. Phys. 84, 6013.
Chapter 7, Section 7.7.2 deals with such a case.
289
9
Normal and Resonance
Electronic and Vibronic Raman
Scattering
Striking the electric chain wherewith we are darkly bound.
Byron
9.1 INTRODUCTION
This chapter considers first normal electronic and vibronic scattering and then resonance
electronic and vibronic Raman scattering. Many aspects of the treatment are similar to
that used for vibrational resonance Raman scattering in Chapter 7, but with the essential
difference that in electronic and vibronic Raman scattering the initial and final electronic
states are necessarily different. As in Chapter 7 we shall only consider in detail the
more simple cases so that the essential features of electronic Raman scattering can be
established. We shall find that electronic Raman scattering provides another powerful
method for studying excited electronic states of molecules.
9.2 NORMAL ELECTRONIC AND VIBRONIC RAMAN

SCATTERING
The essential formulae for electronic Raman scattering were developed in Chapter 4,
Section 4.9 using perturbation theory. They are reproduced here for convenience. The
290
The Raman Effect
component of the electronic transition polarizability tensor in the non-resonance case is

given by
9.2.1
ef vf :eg vi D AIII C BIII C CIII C DIII
where
AIII
BIII C CIII
DIII

p 0ef er p 0er eg
p 0ef er p 0er eg
C
hvf jvi i
9.2.2
er eg 1
er ef C 1
e 6De ,e

1 p kef er p 0er eg C p 0ef er p ker eg
D
h r g f k
er eg 1
e 6De ,e

p 0ef er p ker eg C p kef er p 0er eg
C
vf jQk jvi
9.2.3
er ef C 1

0
0
p kef er p ker eg
1 p kef er p ker eg
D
C
hvf jQk Qk 0 jvi i
h r g f k,k 0
er eg 1
er ef C 1
e 6De ,e
9.2.4
1
D
h r g f
We recall that these formulae have been derived on the basis that 1 , the frequency of
the exciting radiation is far removed from any absorption frequency er eg so that the vr
and the damping factors ier vr can be neglected in the denominators and hence closure
made over the intermediate states jvr i in each electronic state jer i.
The AIII term makes the dominant contribution to the intensity of electronic and vibronic
Raman scattering. The contribution from the term BIII C CIII is very much smaller, and
that from the DIII term much smaller again.
Diagrammatic representations of normal vibrational, electronic and vibronic Raman
transitions are given in Fig. 9.1 and illustrate their essential characteristics.
9.2.2 AIII -term scattering

For the AIII term defined in eq. (9.2.2) to be non-zero two conditions must be met. At
least one of the two products of transition dipole moment components p 0ef er p 0er eg
and p 0ef er p 0er eg must be non-zero; and also the vibrational overlap integral hvf jvi i
must be non-zero.
The first condition depends upon the symmetry of the electronic states jeg i, jer i and
f
je i. This is discussed in more detail in Section 9.3.2.
The second condition involves the properties of the vibrational wave functions in the
electronic states jeg i and jef i. If jvf i and jvi i are orthogonal then the condition vf D vi
is imposed and the AIII term will give rise to purely electronic transitions of the type
jeg ijvi i. If only the ground vibrational state is occupied vi D 0, the
jef ijvf D vi i
Raman spectrum will show just one band arising from the pure electronic transition
jeg ij0i.
jef ij0i
However, the orthogonality of the vibrational states will be removed if either the potential function has a different shape in the state jef i as compared with the state er i, or there
291
|e f
g
|e
(a)
(b)
(c)
Figure 9.1 Comparison of (a) vibrational, (b) electronic and (c) vibronic Raman transitions
excited under non-resonance conditions.
is a displacement of the excited state potential minimum with respect to that of the ground
state along any normal coordinate of the system.
The displacement of the potential minimum is usually the more important factor. As
explained in Chapter 7, symmetry conditions require that such a displacement can only
occur for totally symmetric modes, unless the molecular symmetry is changed in the
excited state. Thus AIII will be associated with electronic Raman scattering involving
quanta of totally symmetric vibrations; and the Raman spectrum will show a progression of
bands of the type jef ijvf i
jeg ij0i i, and also hot bands of the type jef ijvf i
jeg ijvi i
i
if vibrational states with v > 0 are occupied. The relative intensities of these bands
depend on the magnitudes of the overlap integrals hvf jvi i which are determined by the
magnitude of the displacement of the excited state potential minimum relative to that of
the ground state. If, as is normally the case, the excited electronic state is a low energy
jeg ij0i i
state this displacement will generally be small, and then only the jef ij0f i
electronic Raman band will be observed in the absence of resonance.
9.2.3 B III C C III -term scattering

Again, two conditions have to be met if the BIII C CIII term defined in eq. (9.2.3) is to be
non-zero. There must be non-zero products of components of transition dipole moments
and their derivatives, and also the vibrational transition integral hvf jQk jvi i must be nonzero. The first condition involves the symmetries of the electronic states involved. The
second condition imposes the condition that vf D vi 1, and if just the ground vibrational
state is occupied vi D 0 then only vf D 1 is allowed for a given normal coordinate Qk .
There is no symmetry restriction on Qk .
292
The Raman Effect
A transition dipole moment derivative, such as p ker eg defined in eq. (4.9.3), is very
much smaller than the corresponding term p 0er eg and thus it follows that the intensity of
electronic Raman scattering arising from the BIII C CIII term will be very much weaker
than that associated with the AIII term.
9.2.4 D III -term scattering

The DIII term involves products of transition dipole moment derivatives as for example
0
p kef er p ker eg , and so Raman scattering arising from this term will be extremely weak.
Thus the DIII term is of little significance. However we note that the presence of the vibrational integral hvf jQk Qk 0 jvi i means that DIII -term scattering must involve two vibrational
quanta, that is vk D 2 (overtones) when k D k 0 , or vk D 1, vk 0 D 1 (combinations) when k 6D k 0 .
9.2.5 Transition tensor symmetry

It is readily established that the AIII , BIII C CIII and DIII terms are not symmetric upon
exchange of the subscripts and . Thus fi 6D fi . This asymmetry in the transition polarizability tensor for electronic and vibronic Raman scattering has implications for
the symmetry-based selection rules and the polarization properties of the scattered radiation. Tensor asymmetry also arises in the case of resonance electronic and vibronic Raman
scattering. It is convenient to treat the two cases together, and this is done in Section 9.3.
9.3 RESONANT ELECTRONIC AND VIBRONIC RAMAN

SCATTERING
The formulae for resonance electronic and vibronic Raman scattering were developed
in Chapter 4, Section 4.11 using perturbation theory. They are also reproduced here for
convenience. The component of the electronic and vibronic transition polarizability
tensor in the resonance case is given by
ef vf :eg vi D AV C BV C CV C DV
9.3.1
where
hvff jvrr ihvrr jvig i
1
AV D p 0ef er p 0er eg
h
vr
BV D

1
heks er
0
0
p

p

r
g
f
s

ee
e e r s
h 2
e
e
vr
hvff jvrr ihvrr jQk jvig i
hekr es
1
C 2 p 0ef er
p 0es eg
er es
h
vr
9.3.2
9.3.3

CV D

1 hekf et
0
0
p

p

et er
er eg
h 2 ef et
vr
hekt eg hvff jvrr ihvrr jQk jvig i
1
C 2 p 0ef er p 0er et
eg et vr er vr :eg vi 1 ier vr
h
293
9.3.4
k0
hvff jQk jvrr ihvrr jQk 0 jvig i

hes er her es0
1
D D 3 p 0ef es
p 0es0 eg
er es er es0
h
vr
k
9.3.5
We recall that in obtaining these formulae we have assumed that only one electronic
manifold and only one sum over the vibrational states in that manifold need to be taken
into account; and also that terms with non-resonant denominators can be neglected.
9.3.2 AV -term scattering

We shall confine detailed consideration of AV -term scattering to molecules possessing
just one totally symmetric mode. The extension to molecules with more than one totally
symmetric mode will be on a qualitative basis only.
The AV term makes the dominant contribution to resonance electronic and vibronic
Raman scattering. In order that the AV term may be non-zero two conditions must be
fulfilled: the transition electric dipole moments p 0ef er and p 0er eg must both be nonzero; and the product of the overlap integrals hvff jvrr ihvrr jvig i must also be non-zero
for at least some v values.
The first condition requires electronic transitions between the resonant excited electronic
state jer i and both the initial electronic state jeg i and the final excited electronic state
jef i to be electric dipole allowed. Strong scattering will arise if the transition jer i
jeg i
is of the charge transfer or type, whereas ligand field type transitions, for example
dd or ff would give rise to weak scattering.
The second condition involves the degree of orthogonality of the vibrational wave
functions both in the electronic states jeg i and jer i and in the electronic states jer i and
jef i. This is determined by the shapes of the potential functions in the electronic states
jeg i, jer i and jef i and by the relative positions of their potential minima. Figure 9.2
shows a set of potential energy curves in which the potential minima of the electronic
states jer i and jef i are displaced by different amounts er eg and ef eg in the same sense
from the potential minimum of the state jeg i. In the case illustrated er eg > ef eg and
er eg D ef eg C er ef . The overlap integrals hvff j vrr i are related to the vibrational
frequencies vf and vr in the electronic states jef i and jer i and to ef er the displacement
of their potential minima; similarly the overlap integrals hvrr j vig i are related to the
vibrational frequencies vr and vg and to er eg . The functional dependence of overlap
integrals upon these parameters has been discussed in Chapter 7, Section 7.4, and the
formulae given there in eqs. (7.4.2) to (7.4.9) may be adapted readily to the present cases.
We now consider qualitatively the role of the overlap integrals in a few special cases.
If the shape of the potential is the same in the electronic states jeg i, jer i and jef i so that
294
The Raman Effect
| er
(resonant)
(c)
efeg
eref
Energy
ereg
(b)
|e f
|eg
(a)
Qg Q f
Qr
Figure 9.2 Potential energy curves for (a) the ground electronic state jeg i, (b) the final
electronic state jef i and (c) the resonant electronic state jer i. The potential minima of the
three electronic states are displaced relative to each other by the amounts shown.
g
k D kr D k and if also er eg D er ef D 0 (so that ef eg D 0), the only non-zero

overlap integrals will be those of the types hvff D vrr j vrr i and hvrr D vig j vig i.
This will result in an electronic Raman band consisting of the transition jef ij0f i
jeg ij0i i together with hot band transitions of the type jef ij1f i
jeg ij1i i, jef ij2f i
jeg ij2i i etc. However, if either er eg or er ef or both of them are non-zero then vibronic
bands of the type jef ij1f i
jeg ij0i i etc. become Raman active.
The intensity distribution within such a progression depends on the magnitudes of er eg
and er ef and, in general, the larger the displacements the greater the extent of the
g
f
progression. Similar considerations apply if kr and k are different from k . A situation
g
g
f
of interest is when kr 6D k and k 6D k but both er eg and er ef are zero, so that
there is no displacement of the potential minima relative to each other. In the harmonic
approximation the non-equality of the frequencies in the different electronic states can
only give rise to non-zero overlap integrals hvrr j vig i and hvff j vrr i if the sums
vff C vrr and vrr C vig are even numbers. Then the electronic Raman band
consists of a progression in which odd numbers of vibrational quanta are absent; that
jeg ij0i ief ij4f i
jeg ij0i i etc. However, in practice
is jef ij0f i
this is not a very probable situation because significant frequency changes usually imply
a significant displacement of the potential minimum.
295
Usually the electronic Raman transition jef i

jeg i is of the ligand field type and
g
f
jeg i
then k k and ef eg 0. However, the resonant electronic transition jer i
g
r
can be either of a charge transfer or a type. In the charge transfer case k 6D k
g
and er eg 6D 0, whereas in the case kr k and er eg 0. For resonance with
a charge transfer transition, er ef may have a significant value even if ef eg 0 (as
er ef D er eg ef eg ) and thus the observed electronic Raman bands will consist of a
progression in the totally symmetric fundamental. However for resonance with a ligand
field type transition both er ef and er eg are effectively zero and so the electronic Raman
spectrum is dominated by bands arising from the transition jef ij0f i
jeg ij0i i.
A comparison of normal and resonance electronic Raman spectra is of interest. Normal
electronic Raman spectra involve transitions of the type jef ij0f i
jeg ij0i i. As resonance
involving a charge transfer type transition is approached, these bands are enhanced and
progressions appear. However, if the resonance is of the ligand field type then, although the
bands involving transitions of the type jef ij0f i
jeg ij0i i are enhanced, no progressions
appear.
The concepts just discussed can be extended to molecules with more than one totally
symmetric mode. The overlap integrals in eq. (9.3.2) then become multi-dimensional and
the problem becomes rather difficult to treat unless it is assumed that the multi-dimensional
overlap integrals can be expressed as the products of one-dimensional integrals. If this
is so then it is possible to relate the forms of the electronic Raman spectrum to the
f
frequencies ki , kr and k , and the displacements ker eg and ker ef for each mode Qk
involved. The results of such an analysis based on this independent mode approximation
can be summarized as follows: (i) if ker eg and ker ef are zero for all the totally symmetric
jeg ij0i i electronic Raman bands will have appremodes involved only the jef ij0f i
k
k
ciable intensity; (ii) if er eg and er ef are non-zero for just one totally symmetric mode
the electronic Raman spectrum will consist of progressions in that mode; (iii) if ker eg and
ker ef are non-zero for two or more totally symmetric modes then the electronic Raman
spectrum will consist of overtone progressions in these modes and also combination
band progressions involving these modes. The relative magnitudes of the displacements
along each normal coordinate will determine the relative intensities and extents of these
progressions.
The independent mode approximation is generally valid for the state jef i because the
molecular geometry in this state is usually similar to that in the ground state jeg i. This
means that the molecular vibrations in the state jef i are essentially the same as those in
the state jeg i and thus where there is vibrational structure associated with the electronic
band it may be assigned on the basis of ground-state selection rules. The situation in
respect of the resonant state jer i can be more involved. A transition jer i
jeg i may
produce an appreciable change of molecular geometry, particularly if the transition is of
the charge transfer type and the independent mode approximation breaks down. This is
particularly the case when the molecular point group of the state jer i is different from that
of jeg i. In such cases the AV term active modes are those which are totally symmetric in
the point group formed by symmetry operations common to the states jer i and jeg i. The
symmetry species of corresponding modes in the state jeg i may be found using correlation
296
The Raman Effect
tables. The ethene molecule provides an example of this situation which also arises for
AVI -term scattering associated with vibrational resonance Raman scattering. This case is
discussed in Chapter 7, Section 7.5.
If the state jer i is degenerate, AV -term activity can arise for some non-totally symmetric
modes through excited state JahnTeller coupling. Again, this situation also arises for AVI
term scattering associated with vibrational resonance Raman scattering. It is discussed
and examples are given in Chapter 7, Section 7.5. The influence of JahnTeller excited
state coupling on electronic Raman bands may be summarized as follows. In the case
of small distortion, vibrational progressions involving totally symmetric modes plus one
quantum of a JahnTeller active mode are observed. In the case of large distortion, totally
symmetric modes plus multiple quanta of JahnTeller active modes and/or overtone and
combination tone progressions of the JahnTeller active modes are observed.
We consider finally the situation where the state jef i is degenerate. If jer i is nondegenerate then the overlap integrals are non-zero only for totally symmetric vibrations; and, although overlap integrals hvff jvrr i may become non-zero for non-totally
symmetric modes as a result of their JahnTeller activity in the degenerate state jef i, the
product hvff jvrr ihvrr jvig i must necessarily be zero. If both jer i and jeg i are degenerate the products of overlap integrals, hvff jvrr ihvrr jvig i will be non-zero for those
non-totally symmetric modes that are JahnTeller active in the state jer i. In this case the
influence on the electronic Raman spectrum will be very similar to the situation when
the state jef i is non-degenerate. Exceptionally, if the states jef i and jer i have the same
symmetry and the JahnTeller distortions are identical in both states, then the JahnTeller
active vibrations will not appear in the electronic Raman spectrum.
9.3.3 B V -term scattering

Several factors play a role in determining the magnitude of BV -term electronic Raman scattering: the two products of electric dipole transition moment components po ef es po er eg
and po ef er p oes eg ; the vibronic coupling integral heks er and er es the frequency separation of the states jer i and jes i; and the two products of vibrational overlap integrals and
vibrational transition integrals, hvff jQk jvrr ihvrr jvig i and hvff jvrr ihvrr jQk jvig i.
These factors are generally similar to those arising in BV -term scattering for vibrational
resonance Raman scattering except that jef i 6D jei i.
Clearly the transition moments involved in the first factor must be non-zero for the
BV term to be non-zero. For the vibronic coupling integral to be non-vanishing the irreducible representation of the vibrational fundamental with normal coordinate Qk must be
contained in the direct product of the irreducible representations of the states jer i and jes i.
Thus the BV term can be non-zero for both totally symmetric and non-totally symmetric
fundamentals, but in the case of a totally symmetric mode the states jer i and jes i must
have the same symmetry.
The sums over the states vr of the products hvff jQk jvrr ihvrr jvig i and the products
ff
jvrr ihvrr jQk jvig i can each be reduced to a single term by arguments which
hv
parallel those used in Chapter 7, Section 7.3.2. In essence, if the upper vibrational levels
297
of the kth mode are not populated and both er eg and er ef are zero, so that the potential
minima of the states jer i and jef i are not displaced relative to the state jeg i, then the
expression for BV assumes the following form:

k
f jQ j0r ih0r j0i i
1
h
h1
s
r
k
ee
p 0er eg
BV D 2 p 0ef es
er es
er 0rk :eg 0i 1 ier 0r
h

1
hekr es
h1f j1r ih1r jQk j0i i
0
0
C 2 pp ef er
p es eg
9.3.6
er es
er 1rk :eg 0i 1 ier 1r
h
Here we have simplified the nomenclature somewhat and written 0i for vig when
vig D 0, 1f for vff when vff D 1, and so on. It is evident that this expression
is only non-zero for fundamental transitions and thus the BV term gives rise to elec0i bands only. However, even when er eg and
tronic Raman scattering involving 1f
er ef are zero, AV -term scattering is still allowed for 0f
0i electronic Raman scatV
V
tering. As the B term is much smaller than the A term, because of the smallness of
the vibronic coupling integral heks er , the electronic Raman band will consist of a strong
jef ij0f i
jeg ij0i i component and a weak jef ij1f i
jeg ij0i i component. If er eg and
er ef are non-zero the jef ij1f i
jeg ij0i i component is dominated by the AV term.
9.3.4 C V -term scattering

This term relates to vibronic coupling of either the ground state jeg i or the final electronic
state jef i to a higher energy state jet i. As in the case of vibrational resonance Raman scattering, the CV term is usually negligible because of the large energy separation between
the states jeg i and jef i and any excited state jet i.
9.3.5 D V -term scattering

This term gives rise to vibronic Raman bands of the type jef ij2f i
jeg ij0i i, that is first
V
overtones, and also binary combination bands. The D term is much smaller than the
BV term. Even in systems where it is an important scattering mechanism the vibronic
Raman bands which derive intensity from it will be much weaker than the BV type bands
jef ij1f i
jeg ij0i i or the AV type bands jef ij0f i
jeg ij0i i.
9.4 SELECTION RULES IN ELECTRONIC RAMAN SPECTRA

9.4.1 General symmetry considerations
The general symmetry conditions for Raman activity have been discussed in Chapter 7,
Section 7. For convenience we restate the general symmetry condition given there
R D ei vi ef vf \ T2
9.4.1
298
The Raman Effect
This condition assumes the BornOppenheimer approximation applies. The application

of this condition to electronic Raman spectra is generally similar to its application
to vibrational resonance Raman spectra and in many cases leads to similar conclusions.
It should be noted, however, that in the case of electronic Raman scattering, even
under non-resonance conditions, the transition tensor may be asymmetric, and as a consequence the electronic Raman selection rules are less restrictive than in the case of
vibrational Raman scattering. For example in the case of the point group Oh , T2
spans the representations A1g C Eg C F2g for a symmetric transition tensor and only
vibrations corresponding to these symmetry species are Raman active except under resonance conditions. However, the asymmetric tensor components span the representation
F1g so that electronic Raman transitions of symmetry A1g , Eg , F1g and F2g are then
allowed.
In certain systems, as for example the lanthanide ions, spinorbit interaction is important and it is necessary to take into account spin degeneracy. Then we have
spin orbit D spin orbit
9.4.2
The application of this relationship involves double groups.

The treatment of the electronic Raman spectra of the lanthanide ions is most conveniently effected using the transition polarizability tensor expressed in terms of irreducible
spherical tensor components. This is because the eigenfunctions of the lanthanide ions
are spherical harmonics which are characterized by the following quantum numbers: n,
the principal quantum number; L, the total orbital angular momentum quantum number;
S, the total spin quantum number; and J, the total angular momentum quantum number,
where J D L C S, L C S 1, L S; and JZ , the projection of J on the z axis. The various
electronic configurations give rise to spectroscopic terms which are labelled so that they
incorporate the quantum numbers, S, L and J. The notation for the spectroscopic terms
has the form2SC1 LJ with S, P, D, F . . . used for L D 0, 1, 2, 3 . . ., respectively. The spin
multiplicity is (2S C 1).
The orbital parts of these eigenfunctions for the lanthanide ions for the initial and final
states i and f can be written as

aim Ym
9.4.3
i0 D
m
and
0
D
f
afm0 Ym0
9.4.4
m0
0
where aim and afm0 are coefficients and Ym and Ym0 are spherical harmonics. The symbols
m and m0 are used in eqs. (9.4.4) and (9.4.5) because the use of the traditional symbols m
To be consistent with the notation of earlier chapters we use F for triply degenerate representations,
although the symbol T is often employed in electronic spectroscopy to avoid possible confusion with atomic
F terms.
Chapter 14, Section A14.5 deals with irreducible spherical tensor components.
299
j
and m0 is pre-empted by the use of m in m . The transition polarizability for an electronic

Raman transition f
i is then given by
0 j
0
j
m fi D hf jm ji i

0

D
afm0 aim hYm0 jj
m jYm i
m
9.4.5
m0
j
where m is an irreducible spherical tensor component. Using the WignerEckart theorem

this may be expressed as
0

j
j
m
0
j
0
afm aim 1
9.4.6
m fi D
0 m m h jjm jji
m
0
m
where
0
m0
j
m
j
is a Wigner 3-j coefficient and h0 jjm jji is a reduced matrix element.

For the lanthanide ions we can set l D J, l0 D J0 , m D Jz and m0 D J0z . Thus the Wigner
3-j coefficient in eq. (9.4.6) becomes
0

J
j J
9.4.7
J0z m Jz
For the Wigner-3j coefficient to be non-zero, the following conditions must be satisfied:
J j J0 J C j
9.4.8
J0z D Jz C m
9.4.9
jJj D jJ0 Jj j
9.4.10
jJz j D jJ0z Jz j j
9.4.11
and
Thus
and
As j D 2, 1 or 0, it follows that
and
J 2
9.4.12
Jz 2
9.4.13
We illustrate these selection rules by considering the Eu3C ion as an example. This ion
has six 4f electrons so that n D 6. Its ground state is 7 F0 , that is L D 3, 2S C 1 D 7 and
J D 0. The lower energy excited states of this ion are 7 F1 and 7 F2 . Figure 9.3 shows the
300
The Raman Effect

4
3
5D
J
21 000
19 000
17 000
Wavenumber / cm1
6 000
7F
J
6
4 000
5
4
2 000
3
2
1
0
J
Figure 9.3 The 7 FJ and the 5 DJ electronic energy levels of Eu3C .
and also the 5 DJ energy levels of Eu3C . The energy difference of the J manifold of
levels of a given term is a measure of spin-orbit coupling and the energy difference of F
and D terms is related to the repulsive forces between the f electrons.
The general selection rules for electronic Raman scattering may be summarized as
follows: J 2, L 2, S D 0. We consider first the transition 7 F1 7 F0 , for which,
L D 0, S D 0. For the ground state J D 0, Jz D 0 and for the excited state J0 D 1,
(Jz D 1, 0) so that J D 1. It follows from the properties of the 3-j symbol discussed
above that j D 1 and m D 1, 0. Consequently, the tensor components involved are 11 ,
10 and 11 all of which are antisymmetric tensor components. For the transition 7 F2 7 F0
we have J0 D 2, (Jz D 2, 1, 0) so that J D 2. Thus j D 2 and hence m D 2, 1
or 0. The tensor components involved are therefore 22 , 21 , 20 , 21 and 22 , all of which
are symmetric tensor components.
In certain circumstances these selection rules are contravened. For example in a strong
crystal field, J states can be mixed and electronic transitions can be observed for which
J > 2. The Eu3C ion, for which electronic Raman transitions from 7 F0 to 7 F3 , 7 F4 ,
7 F and 7 F have been observed is a notable example of such behaviour.
5
6
7F
J
301
9.5 INTENSITIES AND POLARIZATION PROPERTIES OF

ELECTRONIC RAMAN SCATTERING
9.5.1 Intensities: general considerations
The intensities of electronic Raman bands are generally of the same order as those of
vibrational Raman bands. However, electronic Raman bands are much broader than the
corresponding vibrational Raman bands. There are two reasons for this. The first is the
superposition of hot bands, that is electronic transitions for which the initial state is a
higher vibrational level of the ground electronic state. The second reason is that relaxation
from electronically excited states is faster than from vibrational levels of the ground state.
As a result of this, even when spectra are recorded from samples at low temperatures
electronic Raman bands are usually still broader than vibrational Raman bands.
9.5.2 Excitation profiles

Far from resonance the intensity of electronic Raman scattering is proportional to
1 fi 4 . Under resonance conditions the intensity varies much more strongly as
1 changes. If there is resonance involving one electronic transition, then for AV term
scattering the intensity of an electronic Raman band will be determined by terms like
V
jAxx eg vf :eg vi j2 where
R

2
hvff jvrr hivrr jvig i

2
AV

D 1 px 0f r px r g
ee
xx ef vf :eg vi
e e
r
2vr C e2r vr
h 2
k
v
jvig ivrk vr 0 C er vr er vr 0
r 0 r
C2
hvff jvrr ihvrr jvig ihvff jv

vr <vr
2vr
k
r 0 r
ihv
e2r vr 2r 0
v
e2r vr
9.5.1
with
vr D er vr :eg vi 1
k
and vr 0 D er vr 0 :eg vi 1
k
9.5.2
Except that jef i 6D jeg i, eq. (9.5.1) is very similar to eq. (7.4.11) for resonance vibrational
Raman scattering and similar considerations apply. In particular, the second term consists
of a double summation of terms which have cross-products of vibrational overlap integrals
0
involving pairs of vibrational levels vr and vr of the intermediate state jer i. As a result the
second term can be positive or negative leading to an increase or a decrease, respectively,
in the scattered intensity relative to that arising from the first term which is always positive.
The second term will become negative if one of the overlap integrals is negative or if
either vrk or vr 0 is negative, a situation which arises when 1 is greater than er vr :eg vi
k
k
k
or er vr 0 :eg vi , respectively.
k
302
The Raman Effect
In principle, eq. (9.5.1) can be used to calculate excitation profiles and, by fitting
calculated and experimental profiles, values of the displacements of the potential minima,
er eg and er ef , may be obtained.
If the values of v are small, the excitation profile and the absorption spectrum will
show vibrational structure. However, if the values of v are larger than the vibrational
intervals in the excited state then both the excitation profile and the absorption spectrum
will be smooth and structureless, and their maxima will be in approximately the same
position.
9.5.3 Depolarization ratios

The characteristic properties of depolarization ratios for electronic Raman scattering
follow closely the patterns found for vibrational and vibrational resonance Raman scattering.
Far from resonance the ranges of depolarization ratio values are as follows:
0 /2; ?i < 3/4 for bands arising from totally symmetric electronic Raman
transitions; /2; ?i D 3/4 for bands arising from non-totally symmetric electronic
Raman transitions; 3/4h/2; ?i i 1 for electronic Raman transitions associated with
anti-symmetric tensor components.
Close to resonance the depolarization ratios of totally symmetric electronic Raman
transitions depend on the degeneracy of the resonant excited state jer i. This dependence
has been discussed in some detail in Chapter 7, Section 7. We give a summary of the
results here: /2; ?i D 1/3 for a non-degenerate state jer i; /2; ?i D 1/8 for a
doubly degenerate state jer i; and /2; ?i D 0 for a triply degenerate state jer i.
If the ground state is degenerate the depolarization ratios of totally symmetric modes
can have so-called anomalous values, that is values which do not conform to those given in
the previous paragraph. This situation has also been discussed in some detail in Chapter 7,
Section 7. Here, we shall just recall that the anomalous depolarization ratios arise when the
ground electronic state jeg i is degenerate. The symmetry species of the ground state is then
determined by the species eg eg which can contain irreducible representations other
than the totally symmetric one. A special case is when the ground state is not orbitally
degenerate but has spin degeneracy. Strong spinorbit coupling can then result in the
ground electronic state acquiring some degenerate character which produces anomalous
depolarization ratios.
303
10
Rayleigh and Raman Scattering
by Chiral Systems
When a man hath done, then he beginneth.
Ecclesiastes XVIII 6
10.1 INTRODUCTION
The treatments of Rayleigh and Raman scattering given so far have assumed that the
O can be adequately expressed in the form
interaction hamiltonian H
O DH
Op
H
10.1.1
O p involves only the electric dipole operator. We now consider a more general
where H
situation in which the interaction hamiltonian has the form
OmCH
Oq
O DH
OpCH
H
10.1.2
O q the electric quadrupole operator.

O m involves the magnetic dipole operator and H
where H
O
O
The operators Hm and Hq are non-zero only if the electric field strength of the perturbing
electromagnetic radiation is not constant over the molecule; that is, the electric field
gradient is non-zero. In the visible range of the spectrum the wavelength of electromagnetic radiation is about 1000 times greater than the dimensions of small molecules. Thus
the assumption that the electric field strength is constant over the molecule is normally
a very good one. However, if this assumption is not made and eq. (10.1.2) is used as
the definition of the interaction hamiltonian instead of eq. (10.1.1), some new aspects of
304
The Raman Effect
Rayleigh and Raman scattering are revealed. These include Rayleigh and Raman scattering by chiral molecules and consideration of these optically active scattering phenomena
forms the main substance of this chapter.
O m and H
O q which are
O p, H
It is useful to reproduce here the general definitions of H
developed in Chapter A13:
O p D pO E 0
H
10.1.3
where
pO D
ei ri
10.1.4
Also
where
O m D m
H
O B 0
10.1.5
ei
ri pi
m
O D
2mi
i
10.1.6
O
O q D 1
H
3 E 0
10.1.7
and
where
O D

1
2
ei 3ri ri ri2
10.1.8
In these definitions the particle i at ri (components ri , ri , ri ) has charge ei , mass mi and
momentum pi (components pi ) ; and the summations are over all particles. Summation
over repeated Greek subscripts is also implied and the subscript zero indicates that the
quantity in question is taken at the origin of the molecular system. In the developments that
follow we shall generally omit this subscript to avoid an over-embellished nomenclature,
but it must be understood that these quantities are always taken at the origin of the
molecular system unless otherwise stated. Here, of course, we need only consider the
displacement of the electrons, and so we set ei D e and mi D me in the above equations.
O q are real, the operator
O p and H
It is important to appreciate that, although the operators H
O
Hm is pure imaginary, because for a component of momentum pi the corresponding
operator is
ih
or ih
2 ri
ri
In Section 10.2 we give the additional treatment which is needed when the interaction
hamiltonian is defined as in eq. (10.1.2). Formulae for the intensities and polarization
properties of Rayleigh and Raman scattering are then presented and discussed in subsequent sections. For some parts of the development we shall take the wave functions to be
The use here of p for both dipole components and momentum components cannot be avoided but is
unlikely to cause confusion as the context will clarify the situation.
305
real. However we shall take the amplitudes of the electric and magnetic field of the electromagnetic radiation to be complex so that the various states of polarization are included.
We shall also assume, for most of the treatment, that the frequency of the exciting radiation is far removed from any absorption frequency of the scattering system. We shall
refer to this as the non-resonant case.
10.2 OUTLINE OF THE THEORETICAL TREATMENT

Our initial objective is to obtain expressions for appropriate induced transition moments
for a transition from an initial molecular state jii to a final molecular state jfi when
the system is subject to a perturbation for which the interaction hamiltonian is the more
general one defined by eq. (10.1.2).
To the first order, the perturbed time-dependent wave function has the usual form
i0 D i0 C i1
but i1 is now defined by
i1 D
1
fap1
C am
C aq1
gr0
ir
ir
ir
10.2.1
10.2.2
r6Di
The corresponding expression for f0 is

f0 D f0 C f1
with
f1 D
1
fap1
C am
C aq1
gr0
fr
fr
fr
10.2.3
10.2.4
r6Df
O p , as in eq. (10.1.1), only the coefficients ap1

If the interaction operator were simply H
ir
1
1
to
and ap1
would
contribute
and

.
The
additional
coefficients
that
occur
in
i
f
fr
O q in the more general interaction hamiltoO m and H
eqs. (10.2.2) and (10.2.4) arise from H
nian defined in eq. (10.1.2).
Using time-dependent perturbation theory, expressions may be obtained for the coefficients involved in eqs. (10.2.2) and (10.2.4). The results for the coefficients in eq. (10.2.2)
are given in full below. To emphasize their complex nature the coefficients are now
1 and a
labelled as aQ p1
, aQ m
Q q1
:
ir
ir
ir

1
hrjpO jii Q i
hrjpO jii Q i
1
E exp iri 1 t C
E exp iri C 1 t
aQ pir D
2h ri 1 0
ri C 1 0
10.2.5

1 hrjm
hrjm
O jii Q i
O jii Q i
1
aQ m
D
B0 exp iri 1 t C
B exp iri C 1 t
ir
2h ri 1
ri C 1 0
10.2.6
Chapter 2, Section 2.4 treats this case.
306
aQ q1
ir
1
D
6h
The Raman Effect

O jii
O jii
hrj
hrj
i
i
exp
i
t
C
exp iri C 1 t
EQ
EQ
ri
1
0
0
ri 1
ri C 1
10.2.7
In these general formulae, summation over repeated Greek subscripts is implied
1
1 and a
throughout. The results for aQ pfr
, aQ m
Q q1
have similar forms. Here EQ i 0 is the
fr
fr
component of the complex amplitude of the electric field vector, BQ i 0 is the component
i
is the component
of the complex amplitude of the magnetic induction vector and EQ
0
of the complex amplitude of the electric field gradient tensor. The perturbation arises
from plane wave electromagnetic radiation of frequency 1 . The properties of such a
wave, its propagation direction and state of polarization will determine which components
i
EQ i 0 , BQ i 0 and EQ
are non-zero and also their interrelation.
0
The relations between the complex electric field components, and the real electric field
components of the incident radiation of frequency 1 are of the form
Q i 0 C EQ i 0 g
Ei 0 D 12 fE
10.2.8
On introducing the time-dependence explicitly, this becomes
Ei 0 D 12 fEQ i 0 exp i1 t C EQ i 0 exp i1 tg
10.2.9
These relationships refer to the origin of the molecular system and the full nomenclature
EQ i 0 0 is used to emphasize this. There are corresponding expressions for other real field
components such as Bi 0 and Ei 0 . Also, BQ i and EQ i are related to components of EQ i
as follows:
1
Q i
BQ i D ni0 E
10.2.10
c0
and
Q i D
E
i1 i i
Q
n E
c0 0
10.2.11
In these equations the direction of propagation of the wave is defined by the unit vector
n0i , with ni0 the magnitude of its component along the cartesian axis, and ni0 the
magnitude of its component along the axis.
It will be instructive to illustrate how the general formulae of eqs. (10.2.5) to (10.2.7),
(10.2.10) and (10.2.11) operate when applied to a particular case. We consider the case of
plane polarized incident radiation with n0i D ezi , Exi0 6D 0 and Eyi 0 D 0. This defines D z
and D x in eq. (10.2.10). Thus D zx and for this to be non-zero we must have
D y. Hence D yzx D 1; also, ni0 D ni0z D 1. Thus for n0i D ezi .
Byi D
Chapter A17 treats monochromatic plane waves.
Eix
c0
307
and Bxi D 0 and the alternating tensor has generated the physically correct result! Turning
to eq. (10.2.11), D x, D z and ni0 D ni0z D 1 so that
Eizx D
i1 i
E
c0 x
which again is physically evident. As a result aQ p1

has only one non-zero term which
ir
1 has only one non-zero term which involves B

Q yi and BQ yi , and
involves EQ xi0 and EQ xi0 , aQ m
ir
0
0
has only one non-zero term which involves EQ zxi 0 and EQ zxi 0 .
aQ q1
ir
In order to calculate the intensity of the scattered radiation we must next obtain expressions for the relevant first-order induced time-dependent transition moments. At the level
of approximation being considered here there are three of these to be considered: the
Q fi ;
electric dipole transition moment p
Q fi ; the magnetic dipole transition moment m
O
Q fi now includes terms
and the electric quadrupole transition moment fi . Here p
1
1
1 , a
1 , a
O DH
O p ; and
generated by the coefficients aQ m
Q
Q
and
a
Q
which
are
zero when H
mfr
qir
qrf
ir
O fi are also zero if H
O DH
O p.
of course the transition moments m
Q fi and
The first order time-dependent induced electric dipole transition moment is given by
O i0 i C hf0 jpj
O i1 i
pQ 1 fi D hf1 jpj
10.2.12
This is inherently complex, even with real wave functions, because of the forms of its
exponential time dependent parts; and the real induced transition electric dipole moment
pQ 1 fi is defined as
p 1 fi D pQ 1 fi C pQ 1 fi
10.2.13
We now insert in eq. (10.2.12) the expression for i1 obtained by combining eqs. (10.2.2)
and (10.2.5) to (10.2.7) and proceed similarly for f1 . We then collect the terms in the
resulting expression for p 1 fi which have the correct time-dependence for Rayleigh
and Raman scattering. It is convenient to give the results in component form, and for the
component of p 1 fi we have
1
Q 1
exp is t C pQ 1
exp is tg
p1
fi D 2 fp
0 fi
0 fi
10.2.14
Here, pQ 1
is the complex transition amplitude of p1
fi ; and s is the frequency of
0
the scattered radiation which is given by s D 1 fi (with fi D f i ). Rayleigh
scattering corresponds to the case when f D i (so that s D 1 ) and Raman scattering
to the case when f 6D i (so that s D 1 fi D 1 M ).
We consider first the case of optically active Rayleigh scattering for which we shall give
a reasonably detailed treatment. The case of optically active Raman scattering is more
complicated and will not be treated in such detail. Some of the characteristic properties
of optically active Raman scattering can be inferred from the results for optically active
See Chapter 4, second footnote on page 50.

Chapter 4, Section 4.4 discusses the Rayleigh scattering tensor and the distinction between i0 i (i0 degenerate with i, but i0 6D i) and ii (i0 degenerate with i, and i0 D i). Here we consider only ii .
308
The Raman Effect
Rayleigh scattering and in certain special cases the Placzek polarizability theory can be
extended to optically active Raman scattering.
For optically active Rayleigh scattering, in the most general case when the wave functions of the system are taken to be complex, pQ 1
0 ii may be expressed in the general
form
i
Q ii BQ i C 1 AQ , ii EQ
pQ 1 ii D Q ii EQ i C G
10.2.15
0
where summation over repeated Greek subscripts is implied. Here Q ii is the compoQ ii is the component of a complex
nent of the complex polarizability tensor aQ ii , G
Q ii , and AQ , ii is the , component of another complex
optical activity tensor G
Q
optical activity tensor Aii . We now introduce the simplification that the wave functions
are real and use the perturbation treatment outlined above to obtain expressions for the
tensor components which appear in eq. (10.2.15). For two of the tensors the operators
O , are real. Hence we now have real tensors aii and Aii .
involved, namely pO and
Their components are found to be as follows:
2
ri
RefhijpO jrihrjpO jihg
10.2.16
ii D
2
h r6Di ri 12
and
A, ii D
2
ri
O jii
RefhijpO jrihrj
2
h r6Di ri 12
10.2.17
For convenience in subsequent developments, these equations for tensor components

and others that follow are given in the alternative form developed in Chapter 4. Equation (10.2.16) is identical with eq. (4.4.18) and eq. (10.2.17) has the same general form.
It should be noted that
10.2.18
ii D ii
and
A, ii D A, ii
10.2.19
Equation (10.2.16) can be regarded as representing an electric dipole perturbation followed

by an electric dipole emission, and eq. (10.2.17) as representing an electric quadrupole
perturbation followed by an electric dipole emission.
Q ii whose component G
Q ii appears in eq. (10.2.15)
The case of the complex tensor G
needs special consideration. We first write
Q ii D G ii iG0 ii
G
10.2.20
following the sign choice defined in Chapter 4, Section 4.4. For this tensor the operators
involved are of the types pO and m,
O the latter being pure imaginary. Thus when the wave
Q ii D iG0 ii so that G
Q ii is now pure
functions are real G ii D 0, and hence G
imaginary.
Chapter A15 deals with the optical activity tensors, G, G , A and A

A.
309
From the perturbation treatment it is found that G0 ii which is real is given by
G0 ii D
2
1
O jii
ImhijpO jrihrjm
2
h r6Di ri i2
10.2.21
This equation can be regarded as representing a magnetic dipole perturbation followed

by an electric dipole emission.
By analogy with eq. (10.2.12) the component of the time-dependent induced magnetic
dipole transition moment can be obtained from the relation
m
Q ii D hi1 jm
O ji0 i C hi0 jm
O ji1 i
10.2.22
where m
O is the component of the magnetic dipole operator defined by eqs. (10.1.5)
and (10.1.6). We now substitute in eq. (10.2.22) an expression for ji1 i obtained by
1 and a
Q q1
have
combining eq. (10.2.1), a truncated form of eq. (10.2.2) in which aQ m
ir
ir
been set equal to zero, and eq. (10.2.5). We also proceed similarly for hi1 j, neglecting
1 and a
Q q1
. The component of the complex amplitude of the term with the correct
aQ m
ir
ir
time-dependence for Rayleigh scattering is then found to be given by the general formula
mQ 0 ii D GQ ii EQ i 0
10.2.23
Proceeding as before, we separate the real and imaginary parts of the new complex tensor
component GQ ii and write
GQ ii D G ii iG 0 ii
10.2.24
As the operators pO and m

O are again involved it follows that for real wave functions
0 so that GQ is pure imaginary. From the perturbation treatment it
GQ D iG
ii
ii
is found that
2
1
0
ii D
10.2.25
Imhijm
O jrihrjpO jii
G
2
h r6Di ri 12
This equation represents an electric dipole perturbation followed by a magnetic dipole
1 and a
emission. It should be noted that the neglect of terms like aQ m
Q q1
in the perturbation
ir
ir
0
process means that G ii does not have terms involving transition moment products of
O jii which are much smaller than products of the
the type hijmjrihrj

O
mjii
O
and hijmjrihrj
O
type hijmjrihrj
O
pjii.
O
Using hermiticity properties it is readily shown that for real wave functions
G 0 ii D G0 ii
10.2.26
m
Q 0 ii D iG 0 ii EQ i 0 D iG0 ii EQ i 0
10.2.27
and so eq. (10.2.23) now becomes
310
The Raman Effect
Again, by analogy with eq. (10.2.12) the component of the time-dependent induced
electric quadrupole transition moment can be obtained from the relation
O ji0 i C hi0 j
O ji1 i
ii D hi1 j
10.2.28
Proceeding as for the case of the magnetic dipole, and making the same approximations
for ji1 i and hi1 j the component of the complex amplitude of the term with the
correct time-dependence for Rayleigh scattering is found to be
Q 0 ii D A , ii EQ i

0
where for real wave functions
ri
2
O jrihrjO jii
A , ii D
Refhij
2
h r6Di ri i2
10.2.29
10.2.30
Equation (10.2.30) can be regarded as representing a scattering process in which the

system suffers an electric dipole perturbation by the incident electric field and then undergoes an electric quadrupole transition with the emission of radiation. The neglect of terms
1 and a
Q q1
in the perturbation process means that A , ii does not have transuch as aQ m
ir
ir
O jrihrj
O jii which are much
O jrihrjmjii
O
and hij
sition moment products of the type hij
O
O
smaller than products of the type hijjrihrjpjii.
Using hermiticity properties it is readily shown that
A , ii D A, ii
10.2.31
Q 0 ii D A, ii EQ i

0
10.2.32
so that eq. (10.2.29) becomes
Each term in eq. (10.2.15) involves a different property of the incident electromagnetic
radiation, namely EQ i 0 , BQ i 0 and EQ 0 . However this equation can be recast in terms of the
electric field strength only, by using the relations given by eqs. (10.2.10) and (10.2.11).
We then obtain

1
i
1
1
Q ii C
n0 AQ , ii EQ i 0
10.2.33
pQ 0 ii D Q ii C n0 G
c0
3c0
Equations (10.2.27) and (10.2.32) are already expressed in terms of components of EQ i 0
and no further rearrangement is needed.
10.3 INTENSITIES OF OPTICALLY ACTIVE RAYLEIGH

SCATTERING
10.3.1 General Considerations
We see from the foregoing that optically active Rayleigh scattering at 1 arises from three
sources within a molecule: an induced transition electric dipole in which electric dipole,
311
magnetic dipole and electric quadrupole perturbations have been taken into account; an
induced transition magnetic dipole; and an induced transition electric quadrupole. As
all three multipoles are properties of one and the same molecule, the electric fields they
radiate are phase related and thus their amplitudes must be combined appropriately before
the scattered intensity can be calculated. This situation is discussed in Chapter A 20,
Section A 20.5.
We now consider the case of a single space-fixed molecule at the origin O of a spacefixed cartesian axis system x, y, z with the incident radiation a plane electromagnetic
wave travelling along the positive z direction n0i D ez and having a specific linear
polarization EQ 0 6D 0 D x, y. Initially we take the direction of observation along the
positive x direction n0s D ex . This is the illuminationobservation geometry used in
early experimental work on optically active Rayleigh and Raman scattering. Subsequently
observation along the positive or negative z directions (n0s D ez or n0s D ez , respectively),
so-called forward and back scattering, has also been used and we shall consider these
cases later. When n0s D ex , EQ s0 , the total amplitude of the component of the electric
field of the scattered radiation at a distance x along Ox is given by

k
1
i1 Q
pQ 0 ii xa mQ a0 ii
x0 ii
10.3.1
EQ s0 D
x
c0
3c0
where can be y or z, and
kD
12
4 0 c02
10.3.2
We now substitute for pQ 0 ii using eq. (10.2.33) and adapt eqs. (10.2.23) and (10.2.29)
to substitute for ma0 ii and x0 ii , respectively. The result can be written in the form
k
EQ s0 D Q ii EQ i 0
x
where
10.3.3

1
i
1
Q ii xa GQ a ii ] C
[AQ ,z ii AQ ,x ii ]
Q ii D Q ii C [ z G
c0
3c0
10.3.4
Here Q ii is the component of a new complex scattering tensor cQ ii which
has optical activity contributions in addition to the non-optically active contributions
from Q ii . This scattering tensor plays the same role for chiral systems as ii ,
the polarizability transition tensor, does for non-chiral systems. Note that in introducing
eq. (10.2.33) since n0s D ex we have set D z so that D x or y and is determined by
z .
Equation (10.3.4) relates to the general case where the wavefunctions are complex. We
now proceed to the special case where the wavefunctions are real. Then the components of
a, A and A
A are real, whereas the components of G and G are pure imaginary. Furthermore
we can invoke the relations given by eqs. (10.2.26) and (10.2.31) and express components
of the scattering tensor (c )ii solely in terms of components of aii , G 0 ii and Aii .
312
The Raman Effect
The intensity Iy of the scattered radiation, along the positive x direction, and linearly
Q sy 6D 0 for a single space-fixed molecule is given by
polarized with E
0 c0 Q s Q s
E E
10.3.5
2 y0 y0
The scattered intensity Iz linearly polarized with Esz 6D 0 has a corresponding dependence
on Esz0 .
We now consider some specific cases of polarization of the incident and
scattered radiation which we define using the intensity symbol I; ps , pi and the
illuminationobservation geometry which were developed in Chapter 5.
When the incident radiation is linearly polarized with Exi0 6D 0 and Eyi 0 D 0 it follows
from using eqs. (10.3.3) and (10.3.4) with D y and D x that
Iy D
k
EQ ys0 D Q yx ii EQ xi0
x
with
Q yx ii D
10.3.6

i
i1
0
0
yx ii [Gyy ii Gxz ii ] C
[Ay,zx ii Ax,yx ii ]
c0
3c0
10.3.7
Also, with D z, D x, it follows that

k
EQ zs0 D Q zx ii EQ xi0
x
with
Q zx ii D
10.3.8

i
i1
0
0
zx ii [Gzy ii C Gxy ii ] C
[Az,zx ii Ax,zx ii ]
c0
3c0
10.3.9
Likewise, when the incident radiation is linearly polarized with EQ y0 6D 0 and EQ xi0 D 0, then
using eqs. (10.3.3) and (10.3.4) with D D y we obtain
k
EQ ys0 D Q yy ii EQ yi 0
x
with
Q yy ii D
10.3.10

i
i1
0
0
yy ii C [Gyx ii C Gyz ii ] C
[Ay,zy ii Ay,yx ii ]
c0
3c0
10.3.11
Also, with D z and D y we obtain

k
EO zs0 D Q zy ii EQ yi 0
x
with
Q zy ii D

i
i1
zy ii C [G0zx ii G0yy ii ] C
Az,zy ii Ay,zx ii
c0
3c0
10.3.12

10.3.13
To obtain intensities for the case of N freely rotating molecules of an ideal gas with
Nvi in the initial state vi we must proceed as follows: form quadratic products of the type
313
EQ s0 EQ s0 as required by eq. (10.3.5); then take istotropic averages of the resulting quadratic
products of the tensor components; and finally introduce the population factor Nvi .
It is clear from eqs. (10.3.6) to (10.3.13) that forming quadratic products of the type
s
Q
E0 EQ s0 will produce a considerable number of terms. However some of them can be
disregarded as insignificant and others have zero isotropic averages. It is expedient to
anticipate these simplifications. We consider first the question of relative magnitudes.
Forming products of the type EQ s0 EQ s0 will now lead, in general, to six distinct types of
quadratic products of tensor components which we may represent symbolically as being
of the forms , G0 , A, G0 A, G0 G0 and AA. This is to be contrasted with the situation
O m and H
O q are ignored and only quadratic products of the type arise. However
when H
the magnitudes of quadratic products of the type are of the order of 103 times greater
than the magnitudes of quadratic products of the types G0 and A and of the order of
106 greater than the magnitudes of quadratic products of the types G0 A, G0 G0 and AA. We
shall therefore neglect the quadratic products of the types G0 A, G0 G and AA in obtaining
intensity formulae.
We consider next the three kinds of isotropic averages involved which we may represent
symbolically as hi, hG0 i and hAi. The isotropic averages of the type hi are treated
in Chapter A14. They have been used frequently in earlier chapters and need no special
comment here. However the isotropic averages of the types hG0 i and hAi which are
treated in Chapter A15 appear for the first time in this chapter and their general properties
call for mention.
Isotropic averages of the type hG0 i are similar to those of the type hi. Thus for
hG0 i the non-zero isotropic averages are of the types hxx G0xx i, hxx G0yy i and hxy G0xy i,
just as for hi they are of the types hxx xx i, hxx yy i and hxy xy i. Further, the isotropic
averages of the type hG0 i may be expressed in terms of the same types of tensor invariants
as used for isotropic averages of the type hi, namely an isotropic one aG0 and an
anisotropic one G2 0 . Isotropic averages of the type hAi follow a different pattern. The
only non-zero isotropic averages have forms of the type hzx Azzy i, and in this case there
is no isotropic invariant, only an anisotropic one, A2 . These invariants aG0 , G2 0 and A2
are treated and tabulated in Chapter A15. It should be noted that the definition of A2
incorporates .
For Rayleigh scattering the tensor invariants should strictly be written as aG0 ii , G2 0 ii
and A2 ii to distinguish them from the corresponding tensor invariants for Raman scattering which would be written as aG0 fi , G2 0 fi and A2 fi . However, this fuller notation
is rather cumbersome and as far as possible we shall rely on the context to identify which
tensor invariants are involved.
We conclude this section by emphasizing that symmetry conditions also have to be
satisfied for components of hG0 i and hAi to exist for optically active Rayleigh and
Raman scattering. These considerations are dealt with later in Section 10.5. It will emerge
that only molecules belonging to the chiral point groups meet the necessary symmetry
conditions.
314
The Raman Effect
10.3.2 Intensity formulae

We are now in a position to obtain intensities for various cases of polarization of the
incident and scattered radiation. It must be remembered that for Rayleigh scattering the
formulae will apply only to the case of an ideal gas at a low pressure.
The procedure is similar to that used in Chapter 5 but the scattering tensor components
Q ii defined in eqs. (10.3.6) to (10.3.13) are now used instead of fi .
We shall consider the particular cases where the incident and scattered radiation have
the following characteristics. The incident radiation is along z with n0i D ez and is either
right or left circularly polarized. The time-dependent electric field of the incident electromagnetic radiation is then given by
or
1
E R D p E0 ex iey exp i1 t kz
2
10.3.14
1
E L D p E0 ex C iey exp i1 t kz
2
10.3.15
and for either polarization

I D 12 c0 0 E02
10.3.16
The scattered radiation is observed along x with n0s D ex and is analysed for its
linearly polarized components jjs EQ zs0 or ?s EQ ys0 . Thus we require expressions for
R i , I /2; ?s ,
L i , I /2; jjs ,
R i and I /2; jjs ,
L i . These cases are of
I /2; ?s ,
particular importance for chiral systems for reasons that will emerge shortly.
R i and I /2; ?s ,
L i .
To illustrate the procedure we take as examples I /2, ?s ,
s
We use eq. (10.3.5) and set D y. We see that there are two contributions to EQ y0 , one
from EQ xi0 given by eqs. (10.3.6) and (10.3.7) and one from EQ yi 0 given by eqs. (10.3.10)
and (10.3.11). These contributions to EQ ys0 must be combined appropriately according as
R i or
L i . The necessary substitutions are then made in eq. (10.3.5), and terms
pi D
of the types G0 G0 , G0 A and AA are neglected because their contributions are very small
Q zs for which
R i and I /2; jjs ,
L i involve E
as already explained. Similarly, I /2; jjs ,
0
i
Q
there are two contributions, one from Ex0 given by eqs. (10.3.8) and (10.3.9), and one
from EQ yi0 by equations (10.3.12) and (10.3.13).
For reasons which will soon emerge the results will be given in the form of the sums
R i I /2; ?s ,
L i and I /2; jjs ,
R i I /2; jjs ,
L i .
and differences I /2; ?s
We consider the case of N molecules with Nvi molecules in the initial state vi . The
molecules are assumed to be freely rotating so that isotropic averages can be taken. In the
formulae given below those quadratic terms whose space averages are zero are omitted.
Thus we have
R i CI /2; ?s ,
L i D kQ Q s4 Nvi fh2yy i C h2yx igI
I /2; ?s ,
Chapter A20 deals with circularly polarized radiation.
10.3.17

R i I /2; ?s ,
L i
I /2; ?s ,

1
kQ Q s4 Nvi
0
0
2 hyx Gyx i C hyy Gyy i C
hyx Ay,yz i I
D
c0
3
R i C I /2; jjs ,
L i D kQ Q s4 Nvi fh2zx i C h2zy igI
I /2; jjs ,
s
315
10.3.18
10.3.19
R I /2; jj ,
L
I /2; jj ,
4

1
kQ Q s Nvi
2 hzx G0xx i C hzy G0zy i
hzx Az,zy i hzy Az,zx i I 10.3.20
D
c0
3
On introducing the expressions for the isotropic averages in terms of the invariants of
the chiral tensor given in Table A15.1 of Chapter A15 the following results are obtained
for intensities of Rayleigh scattering:

45a2 C 7 2
s
i
s
i
4
R C I /2; ? ,
L D kQ Q s Nvi
10.3.21
I
I /2; ? ,
45

0 C 7 2 C 2
2
45aG
0
A
G
R i I /2; ?s ,
L i D kQ Q s4 Nvi
I /2; ?s ,
I 10.3.22
c0
45

2
R i C I /2; jjs ,
L i D kQ Q s4 Nvi 2
I /2; jjs ,
10.3.23
I
15

4 3G2 0 A2
s
i
s
i
4
R I /2; jj ,
L D kQ Q s Nvi
10.3.24
I /2; jj ,
I
c0
45
The reasons for the use of intensity sums and differences now become clear. We see
from eqs. (10.3.21) to (10.3.24) that the intensity sums involve only normal scattering
tensor invariants, whereas the intensity differences involve only optically active tensor
invariants. It is the cross-terms G0 and A (and hence aG0 , G2 0 and A2 ) which give
rise to a slightly different response of the scattering system to right and left circularly
polarized incident radiation. We shall deal shortly with the symmetry conditions that must
be satisfied for these cross-terms to exist.
It is convenient to introduce at this point the conventional, normalized circular intensity
differential (CID) which for a general scattering angle is denoted by ; ps , i where
the symbol i indicates that the incident radiation is circularly polarized. It is defined as
; ps , i D
R i I; ps ,
L i
I; ps ,
R i C I; ps ,
L i
I; ps ,
10.3.25
Expressions for this CID are given below for the cases of ps D jjs , ps D?s and ps D
?s C jjs when D /2:
/2; ?s , i D
245aG0 C 7G2 0 C A2

c0 45a2 C 7 2
10.3.26
Note that the definition of A2 used here differs from that used in my earlier book (Long 1977). See Chapter
A15, p 500.
316
The Raman Effect
23G2 0 A2
10.3.27
3c0 2
245aG0 C 13G2 0 A2
/2; jjs C ?s , i D
10.3.28
c0 45a2 C 13 2
The treatment of optically active light scattering has so far been restricted to the case of
incident circularly polarized radiation and scattered linear polarized radiation with a 90
scattering angle. The first experimental observations of optically active light scattering
were based on this case but more recent work has used other combinations of polarization
states of the incident and scattered radiation and other scattering angles. The cases that
have been studied are summarized below.
/2; jjs , i D
(a) Incident circular polarization (ICP)

I; ps , pi for D 0, /2,
R i or
L i
with ps D?s , jjs or ?s C jjs and pi D
in any combination
(b) Scattered circular polarization (SCP)
I; ps , pi for D 0, /2,
R s or
L s and pi D?i , jji or ?i C jji
with ps D
in any combination
(c) Dual circular polarization (DCP)

This involves simultaneous synchronous circular polarization modulation of the incident and scattered radiation, which may be in-phase or out-of-phase:
(i) in-phase
I; ps , pi for D 0, /2,
R s and pi D
R i
with ps D
L s and pi D
L i
or ps D
(ii) out-of-phase
I; ps , pi for D 0, /2,
L s and pi D
R i
with ps D
R s and pi D
L i
or ps D
The labels in and out are added to the intensity symbols when it is necessary to indicate
the phase relationship between the scattered and incident circular polarization states.
Additional CIDs can be defined for the new cases, namely b, c(i) and c(ii). For scattered
circular polarization we introduce
R s , pi I;
L s , pi
I;
10.3.29
; s , pi D
R s , pi C I;
L s , pi
I;
For the case of dual circular polarization two further CIDs can be identified. When the
incident and scattered circular polarizations are in phase we have

in ; s , i D
317
R s,
R i I;
L s,
L i
I;
R s,
R i C I;
L s,
L i
I;
10.3.30
For the out-of-phase case we have

out ; s , i D
L s,
R i I;
R s,
L i
I;
L s,
R i C I;
R s,
L i
I;
10.3.31
10.3.3 Stokes parameters

This section gives the Stokes parameters for expressing intensities of Rayleigh scattering
by chiral molecules for D /2, 0 and . They are presented in terms of a general
polarization of the incident radiation and so may be used to obtain various combinations
of intensities and polarizations of the scattered radiation for any chosen polarization of
the incident radiation.
For incident radiation propagating along the z axis, n0i D ez with E0i the amplitude
of its electric field strength and partially polarized with a degree of polarization P, an
azimuth
and an ellipticity it is found that the Stokes parameters of the radiation
scattered at 90 to z, along the x direction n0s D ex are given by
S0s /2 D

kQ Q s4 Nvi
x2
45a2 C 13 2 45a2 C 2 P cos 2 cos 2 C 2/c0 45aG0 C 13G2 0 A2 P sin 2
90
E0i 2
(10.3.32)
S1s /2 D

kQ Q s4 Nvi
x2
45a2 C 2 1 P cos 2 cos 2 C 2/c0 45aG0 C G2 0 C 3A2 P sin 2

90
E0i 2
(10.3.33)
S2s /2 D 0
S3s /2 D

kQ Q s4 Nvi
x2
2/c0 45aG0 C 13G2 0 A2 2/c0 45aG0 C G2 0 C 3A2 P cos 2 cos 2
90
(10.3.34)
E0i 2
(10.3.35)
The superscript s denotes Stokes parameters for scattered radiation.
318
The Raman Effect
We recall that Stokes parameters are linear functions of the squares of the amplitudes
of the radiation with which they are associated. Thus eqs. (10.3.32) to (10.3.35) relate the
squares of the electric field amplitudes of the scattered radiation at a distance x from the
scattering molecules to the squares of the electric field amplitudes of the incident radiation.
Subsequently we shall adapt these formulae to give intensities of scattered radiation in
terms of the irradiance of the incident radiation.
The Stokes parameters for radiation scattered at 0 that is along the positive z axis with
nzs D ez are given by

4N i
2 C 7 2 C 4/c 45aG0 C 2 2 P sin 2
k

Q
45a
0
0
Q s v
A
G
S0s 0 D
E0i 2
z2
45
S1s 0 D
kQ Q s4 Nvi
z2
kQ Q s4 Nvi
S2s 0 D
z2

10.3.36
45a2 C 2 P cos 2 cos 2

45
45a2 C 2 P cos 2 sin 2
45
E0i 2
10.3.37
E0i 2
10.3.38
2
2
2
2
0
45a 5 P sin 2 C 45aG C G0 A
4N i
k

Q
c

Q
v
s
0
E0i 2
S3s 0 D
z2
45
180 ,
10.3.39
that is along the negative z axis
The Stokes parameters for radiation scattered at

with nzs D ez are given by
2
2
2
2
3G0 C A P sin 2
4 N i 45a C 7 C
k

Q
c0
Q s v
s
E0i 2
S0 D
z2
45

kQ Q s4 Nvi 45a2 C 2 P cos 2 cos 2
S1s D
E0i 2
z2
45

4N i
2 C 2 P cos 2 sin 2
k

Q
45a
Q s v
S2s D
E0i 2
z2
45
2 5 2 P sin 2 C 8 3 2 C 2
45a
A
G0
kQ Q s4 Nvi
c0
s
E0i 2
S3 D
z2
45
10.3.40
10.3.41
10.3.42
10.3.43
It is the Stokes parameters S0s and S1s that determine the intensities of the components
polarized parallel and perpendicular to the scatter plane. In the three sets of equations
given above any contributions to these intensities arising from the additional terms aG0 ,
319
G2 0 and A2 depend on P sin 2. Thus these contributions are zero if the incident radiation
is unpolarized P D 0 or linearly polarized P D 1, D 0, sin 2 D 0 but not otherwise.
Further, as the intensity of any circularly polarized component is determined by S3s we
see that the additional terms give rise to a circularly polarized component because there
are always terms in S3s which are independent of P, and . For forward and backward
scattering there is no change in azimuth, whereas for 90 scattering the azimuth is always
perpendicular to the scatter plane.
Finally we see from eq. (10.3.35) that the inclusion of the additional terms gives rise
in the case of 90 scattering to scattered radiation which is circularly polarized when the
incident radiation is linearly polarized P D 1, D 0, cos 2 D 1. For such scattering the
degree of circularity C /2; pi is a useful parameter. For circularly polarized scattered
radiation, which arises for the cases of pi D jji and pi D?i , then as
C ; pi D
S3
S0
10.3.44
we have
23G2 0 A2
3c0 2
245aG0 C 7G2 0 C A2
C /2; ?i D
c0 45a2 C 7 2
C /2; jji D
10.3.45
10.3.46
Comparison of eq. (10.3.45) with eq. (10.3.27) and eq. (10.3.46) with eq. (10.3.26) shows
that for optically active Rayleigh scattering C /2; jji D /2; jjs , i and C /2;
?i D /2; ?s , i . However, this is not a general result.
The Stokes parameters for the scattered radiation may be used to obtain intensities,
intensity sums and differences, and polarization properties of the scattered radiation. The
procedure is given in Chapter A21, Section A21.2.4. It involves adapting the definitions of
Stokes parameters given in Chapter A21, Section A21.2.3 and the definitions of changes
of polarization given in Chapter A21, Section A21.4.
Using these procedures with S0s /2 and S1s /2 the intensity components
I /2; jjs , pi and I /2; ?s , pi are found to be

2 C 4/c 3 2 2
6
0
0
A
G
P sin 2 I
10.3.47
I /2; jjs , pi D kQ Q s4 Nvi
90
and
I /2; ?s , pi D

kQ Q s4 Nvi
45a2 C 7 2 45a2 C 2 P cos 2 cos 2
90

2
0
2
2
C 45aG C 7G0 C A P sin 2 I
c0
See Chapter A21, eqs. (A21.4.5) and (A21.4.8).
10.3.48
320
The Raman Effect
These results may be used to confirm for D /2 the intensity sums and differences
given by eqs. (10.3.21) to (10.3.24) and the CIDs given by eqs. (10.3.26) to (10.3.28).
In a similar manner the corresponding intensity sums and differences and CIDs may be
calculated for D 0 using eqs. (10.3.36) and (10.3.37) and for D using eqs. (10.3.40)
and (10.3.41). The formulae for intensity sums and differences for circularly polarized
incident radiation and linearly polarized scattered radiation are presented in Reference
Tables 10.1 D /2, 10.2 D 0 and 10.3 D The corresponding normalized
circular intensity differentials are given in Reference Table 10.4 for D /2, 0 and .
Using S1s and S3s the intensity sums and differences for circularly polarized scattered radiation may be calculated for D /2, 0 and . Specific intensity formulae are
given in Reference Table 10.5; and the corresponding normalized circular intensity differentials in Reference Table 10.6. Formulae for specific intensity sums and differences for
dual circular polarization of the incident and scattered radiation are presented in Reference Table 10.7 for D /2, 0 and . The corresponding normalized circular intensity
differences are given in Reference Table 10.8.
It must be emphasized that these formulae apply only when the scattering can be
treated using the five tensor invariants, a2 , 2 , aG0 , G2 0 and A2 . As we have seen, this is
the case for Rayleigh scattering and real wave functions. We shall see in Section 10.4
that these formulae can be applied to Raman scattering if certain special conditions are
satisfied.
Examination of Reference Tables 10.1 to 10.8 leads to a number of interesting conclusions. For example it can be seen that 0; ?s C jjs , i D 0; s , ?i C jji , out ; ?s
C jjs , i D out ; s , ?i C jji and out /2; s , i D out 0; s , i D out ;
s , i . Also for scattering with D 0 the contribution to the intensity from the invariant
aG0 is very much larger than that from the invariants G2 0 and A2 . However for D there
is no contribution to the intensity from aG0 , and G2 0 makes a larger contribution than A2 .
For both D 0 and D the use of dual circular polarization (in phase) of the incident
and scattered radiation gives rise to scattering with the least amount of normal scattering
relative to optically active scattering. Thus this method is intrinsically the most effective
for observing optically active scattering. A number of relationships between DCP, ICP and
R s,
R i I /2;
L s,
L i for DCPin
SCP scattering also emerge. For example I /2;
s
s
i
s
s
i
s
R I /2, ? C jj ,
L or I /2;
R
L s , ?i C jji .
is equal to I /2; ? C jj ,
These tables also show that by using appropriate intensity measurements or combinations thereof it is possible to determine values of the individual isotropic invariants. For
example the invariant combination 96G2 0 C 32A2 c01 can be obtained from the in-phase
R s,
R i I ;
L s,
L i . Also the invariant combinaDCP intensity difference I ;
1
tion 24G2 0 8A2 c0 can be obtained from either of the following: the ICP intensity
R i I /2; jjs ,
L i ; or the SCP intensity difference I /2;
R s
difference I /2; jjs ,
2
2
s
i
L , jj . A knowledge of these two invariant combinations enables G0 and A to be calcu
lated. Figure 10.1 illustrates the decomposition of the optically active Raman spectrum
of C-trans-pinane.
Counts/ 104
(e)
I g A2
5
0
5
(d)
Counts / 104
321
2
I gG
5
0
5
Counts / 104
(c)
2 g2
I 3g G
5
0
5
Counts / 104
(b)
2 + g2
I 3g G
5
0
5
Counts / 107
(a)
(+) - trans-pinane
I g A2
4
2
0
700
900
1100
1300
1500
Wavenumber/cm1
Figure 10.1 Optically active Raman spectra of (C)-trans-pinane, excited with 488 nm
R i C I /2; jjs ,
L i proportional to 2 . (b) DCPin ; I ;
R s,
R i
radiation: (a) I /2; jjs ,
s L i
2
2
s R i
s L i
L
I ; , proportional to 3G0 C A . (c) I /2; jj , I /2; jj , proportional to 3G2 0 A2 . (d) Pure magnetic dipole anisotropic invariant spectrum, proportional
to G2 0 , calculated by combining (b) and (c) appropriately. (e) Pure electric quadrapole
anisotropic invariant spectrum, proportional to A2 , calculated by combining (b) and (c)
appropriately.
10.4 INTENSITIES OF OPTICALLY ACTIVE RAMAN

SCATTERING
The appropriate formulae for the intensity of optically active Raman scattering can be
obtained by developing expressions for pQ 01 fi following the procedures already given
322
The Raman Effect
for pQ 01 ii . Although the result will be broadly similar to those for optically active
Rayleigh scattering there are some important differences, and we now examine how
these arise.
For optically active Raman scattering eq. (10.3.3) is replaced by
k
10.4.1
EQ s0 D Q fi EQ i 0
x
and the component of the new complex scattering tensor
Q fi is given by
1
Q fi xa GQ a fi ] C i [1 AQ ,z fi s AQ ,x fi ]
[z G
c0
3c0
10.4.2
It should be noted that both 1 and s now appear in eq. (10.4.2).
For the general case where the wave functions are taken to be complex, time-dependent
perturbation treatment yields the following general expressions, including r , for the five
complex tensors which appear on the right-hand side of eq. (10.4.2).

Q fi D
C
10.4.3
h r6Di,f ri 1 ir
rf C 1 C ir

1 hfjpO jrihrjm
O jrihrjpO jii
O jii hfjm
Q
G fi D
C
10.4.4
h r6Di,f ri 1 ir
rf C 1 C ir

1 hfjm
O jrihrjpO jii hfjpO jrihrjm
O jii
Q
G
D
C
10.4.5
fi
h r6Di,f ri 1 ir
rf C 1 C ir

hfjp
O jii hfjO jrihrjpO jii
1
O
jrihrj

C
10.4.6
AQ , fi D
h r6Di,f ri 1 ir
rf C 1 C ir

O jrihrjpO jii hfjpO jrihrjO jii
1 hfj
AQ , fi D
C
10.4.7
h r6Di,f ri 1 ir
rf C 1 C ir
Q fi D Q fi C
We now discuss the isotropic invariants associated with these optically active Raman
transition tensors. As we wish to compare them with those associated with optically active
Rayleigh scattering we shall use the full notation for the invariants. If we include the
cases where the tensors can have both symmetric and antisymmetric parts, the number
of isotropic invariants which can arise is very considerable. There are 13 isotropic
invariants (five of them antisymmetric) associated with the real parts and 13 (five of
them antisymmetric) associated with the imaginary parts making a total of 26 in all.
If we take the wave functions to be real there are then only 13 isotropic invariants,
and if further we disregard the five antisymmetric parts that remain we are left with
only eight isotropic invariants: four isotropic invariants involving the real parts of the
2
Q fi and A
Q fi , namely a2 fi , 2 fi , A2 fi and A
A
complex tensors aQ fi , A
fi and
The full notation for the tensor invariants was introduced earlier on page 313.
323
Q fi
four isotropic invariants involving the imaginary parts of the complex tensors G
2
2
0
0
Q
and G fi , namely aG fi , G0 fi , aG fi and G 0 fi . This may be contrasted with
normal Rayleigh and Raman scattering where the number of isotropic invariants is only
two when the wave functions are real and the tensor is symmetric; and also with optically
active Rayleigh scattering where the number of isotropic invariants is only five when the
wave functions are real and the tensor is symmetric, as we have seen in Section 10.3. In
this latter case the hermitian properties of the operators involved in the tensors lead to the
relations between the components of the tensors Gii and G ii and the components of
the tensors Aii and A
A ii which are given in eqs. (10.2.26) and (10.2.31), respectively.
As a result there are the following three relationships between the isotropic invariants:
aG0 ii D aG 0 ii
10.4.8
G2 0 ii
A2 ii
10.4.9
D
D
G2 0 ii
2
A
ii
10.4.10
and the number of independent isotropic invariants is reduced from eight to five. For
optically active Raman scattering these relationships do not hold because i 6D f, so that
jii 6D jfi, ri 6D rf and f 6D i .
Reference Table 10.9 gives intensity sums and differences for optically active Raman
scattering for both ICP and SCP in the case when the wave functions are real, the transition tensor is symmetric and there are eight isotropic invariants. The formulae reduce
to the corresponding formulae in Reference Tables 10.1 to 10.8 when the relations in
eqs. (10.4.8) to (10.4.10) are invoked. We shall not analyse these much more complicated formulae here. However, two interesting consequences should be noted: for a given
vibrational transition the circular intensity differential is now not equal to the degree of
circularity; and the circular intensity differential for the Stokes transition is not equal to
that for the anti-Stokes transition and similarly for the degree of circularity.
In certain circumstances however these more general, but rather intractable, results
for optically active Raman scattering may be reduced to the same forms as those for
optically active Rayleigh scattering. Let us assume that the final state jfi is not very
different from the initial state jii so that ri rf and i f . This restricts the
resulting approximate formulae to low-frequency vibrational and low-energy electronic
transitions. We also assume that 1 > ri and r can be neglected. This further restricts
the approximate formulae to normal non-resonant cases. With these assumptions the three
relationships between the isotropic invariants for optically active Rayleigh scattering given
by eqs. (10.4.8) to (10.4.10) may be considered to apply to a good approximation to optically active Raman scattering. Thus only five isotropic invariants need to be considered
and the formulae given in Reference Tables 10.1 to 10.8 may also be applied to optically active Raman scattering. This situation is discussed further in Section 10.4.3 which
considers the Placzek polarizability theory in relation to optically active scattering.
10.4.2 Discussion of intensities and isotropic invariants

The full theory of optically active Raman scattering provides, in principle, the basis for
many new insights into the structure and configuration of chiral molecules. In particular
324
The Raman Effect
structural information about different parts of a chiral molecule can be obtained from the
properties of bands associated with characteristic group vibrations in the optically active
Raman spectrum. However the exploitation of optically active Raman scattering is rather
demanding experimentally.
The discussion in this section is based on the simplified theory which involves only
five isotropic invariants, a2 , 2 , aG0 , G2 0 and A2 . We now introduce a ratio of intensity
differences R /2, defined as
R /2 D
R i I /2; ?s ,
L i
I /2; ?s ,
R i I /2; jjs ,
L i
I /2; jjs ,
10.4.11
and, using eqs. (10.3.22) and (10.3.24), we have

R /2 D
45aG0 C 7G2 0 C A2

23G2 0 A2
10.4.12
It is usually a good approximation to neglect the isotropic term aG0 . If we assume a

purely magnetic dipole mechanism for the scattering, so that A2 D 0, then R /2 D 7/6.
If instead we assume a purely electric quadrupole mechanism, so that G2 0 D 0, then
R /2 D 1/2. If G2 0 D A2 then R /2 D 2; and if G2 0 D A2 , then R /2 D 3/4.
In general, an observed value of R /2 less than 7/6 but positive indicates an electric
quadrupole contribution of opposite sign to the magnetic dipole contribution.
Experimentally determined values of R /2 can provide useful structural information.
For example, if all the bonds in a molecule have cylindrical symmetry then a relatively
simple bond polarizability theory predicts that for such molecules aG0 D 0 and G2 0 D A2
so that R /2 D 2. On this basis, observation of significant deviations from R /2 D 2
would be expected to be associated with deformation vibrations of non-axially symmetric
bonds and would indicate that either G0 or A, is dominant. An example of this is
provided by the bands in the optically active Raman scattering of -pinene associated
with the deformation of the non-axially symmetric alkenic group. However, significant
deviations from R /2 D 2 have been observed for molecules in which all the bonds
are axially symmetric. The optically active Raman spectrum of menthol provides an
example with some bands having R /2 1. These deviations have been attributed to
large electric quadrupole contributions induced by excited state interactions involving
O or S Rydberg p orbitals on other parts of the molecule. In general the presence of
low-lying excited electronic states connected to the ground state by magnetic dipole or
electric quadrupole transition moments with the appropriate components fully allowed
can selectively boost G0 or A, respectively.
10.4.3 Placzek polarizability theory and optically active scattering

The assumptions made in Section 10.4.1 in order to reduce the number of isotropic invariants to five are effectively those invoked in the Placzek theory of normal Raman scattering.
These assumptions result in the general transition tensors given by eqs. (10.4.3) to (10.4.7)
being reduced to vibrational transition tensors in which the electronic part of the tensor is
325
regarded as an operator. This operator is a function of the normal coordinates of vibration

Q and brings about the vibrational transitions. Thus optically active Raman scattering can
also be treated using the Placzek theory and we now explore this approach.
We recall that vf vi , the component of the Placzek vibrational transition polarizability, is a function of the normal coordinates Q of the system and is defined by
Expanding O Q as a Taylor series in Q, we have

Qk
O Q D 0 C
Q
k
0
k
10.4.13
10.4.14
where derivatives higher than the first have been neglected. Introducing this expansion
into eq. (10.4.13) gives

f
i
hvf jQk jvi i
10.4.15
vf vi D hv j 0 jv i C
Q
k
0
k
Analogous expressions can be obtained for the components of the vibrational transition
optical activity tensors G0 vf vi and A, vf vi .
The first term in eq. (10.4.15) is associated with Rayleigh scattering. In the harmonic
approximation it is zero when vf 6D vi , and equal to 0 when vf D vi . The intensity
of Rayleigh scattering from non-optically active molecules is thus proportional to terms
of the type
10.4.16
jhvi j 0 jvi ij2 D j 0 j2
By analogy the additional contributions to the intensity of Rayleigh scattering associated
with chiral molecules will involve terms of the type
jhvi j jvi ihvi jG0 jvi i D 0 G0 0
10.4.17
jhvi j jvi ihvi jA, jvi i D 0 A, 0
10.4.18
and
The second term in eq. (10.4.15) is associated with vibrational Raman scattering. It is
non-zero provided that /Qk 0 is non-zero and also, in the harmonic approximation,
that vf D vi 1. It follows that the intensity of normal Raman scattering associated with
the kth normal coordinate is proportional to terms of the type

2 f
jhv jQk jvik ij2
10.4.19
Qk 0 k
f
with vk D vik 1.
By analogy the additional contributions to the intensity of Raman scattering from chiral
systems will involve terms of the types

0
G

f
jhvk jQk jvik ij2
10.4.20
Qk 0 Qk 0
326
The Raman Effect

and

Qk

0
A,
Qk

f
jhvk jQk jvik ij2
10.4.21
where again vk D vik 1. Also, just as non-zero normal Raman scattering requires that
/Qk 0 is non-zero, so for the chiral contributions to be non-zero the associated
derivatives G0 /Qk 0 and A, /Qk 0 which appear in eqs. (10.4.20) and (10.4.21),
respectively, must also be non-zero.
10.5 SYMMETRY CONSIDERATIONS

The symmetry requirements for optically active Rayleigh and Raman scattering can be
conveniently discussed in terms of the following equations:
hvf j jvi ihvf jG0 jvi ia D vf vi G0 vf vi
f
hv j jv ihv jA, jv i D vf vi A, vf vi
10.5.1
10.5.2
For Rayleigh scattering vf D vi and for Raman scattering vf D vi 1 in the harmonic

approximation.
The symmetry arguments we shall use are a straightforward extension of those used
for normal Rayleigh and Raman scattering. We consider first eq. (10.5.1). For this to be
non-zero, then for optically active Rayleigh scattering the same components of the tensors
a and G 0 , for example and G0 , must each span the totally symmetric representation;
and for optically active Raman scattering and G0 must each span the irreducible
representation of the normal coordinate under consideration.
The case of eq. (10.5.2) might seem at first to be a little more complicated because
A, itself does not transform like G0 . However , A, has transformation
properties identical to those of G0 and it is this combination which always appears
in association with . Thus for optically active Rayleigh scattering pairs of components such as and A, must each span the totally symmetric representation; and
for optically active Raman scattering pairs such as and A, must each span
the irreducible representation of the normal coordinate under consideration. Clearly all
Raman-active vibrations in chiral molecules should show Raman optical activity.
These symmetry requirements can only be met in a limited number of point groups
which, in consequence, are termed the chiral point groups. The point groups in question
are Cn , Dn , T, O and I. We now consider as an example the point group C2 , the character
table for which is given below.
C2
A
B
E
1
1
C2
1
1
z
x, y
Rz
Rx , Ry
Chapter 5, Section 5 discusses these symmetry arguments.

This stems from the tensor contraction implicit in , A, . See Chapter A10, Section A10.4.5.
327
Now the second-rank polar tensor components xx . . . xy . . . transform as the cartesian

products xx . . . xy . . ., and the second-rank axial tensor components G0xx . . . G0xy . . . transform as xRx . . . xRy . . . . Thus xx , yy , zz , and xy belong to A as do G0xx , G0yy , G0zz
and G0xy . However xz , yz , G0xz and G0yz belong to B . Thus optically active Rayleigh
scattering can arise through the tensor component products xx G0xx , yy G0yy , zz G0zz and
xy G0xy . Optically active Raman scattering can arise through the same four products for
vibrations belonging to A and through the products xz G0xz and yz G0yz for vibrations
belonging to B . These statements also apply to the products A, because they
transform as G0 .
10.6 CONCLUDING REMARKS

The theory of optically active Raman scattering can be expanded, as in the case of normal
Raman scattering, to cover resonance situations. The treatment follows that developed in
Chapter 4 using the BornOppenheimer approximation and HerzbergTeller vibronic
coupling. The resulting formulae are very complicated and will not be given here.
Two other extensions of the concepts developed in this chapter deserve mention. The
first concerns magnetic Rayleigh and Raman optical activity. All molecules in a static
magnetic field should show Rayleigh and Raman optical activity. The origin of this
scattering lies in the perturbation of the polarizability by an external magnetic field and
is treated by writing the polarizability as a power series in the magnetic induction B:
Q D i0 C B i0 B C . . .
10.6.1
It is found that at transparent frequencies the main contribution to magnetic Rayleigh

and Raman optical activity arises from interference between components of unperturbed
and components of 0 perturbed to first order in the magnetic induction, together
with interference between components of perturbed to first order in the magnetic
induction and components of unperturbed 0 . The treatment can be developed in detail
for both non-resonant and resonant cases but the results will not be given here.
The second extension concerns electric Rayleigh and Raman optical activity. The origin
of this scattering lies in the perturbation of the polarizability and optical activity tensors
and transition tensors by an external electric field. Electric Rayleigh optical activity
depends on interference between the components of the unperturbed tensor a and the
components of the tensors G 0 plus A perturbed to first order in the electric field, together
with interference between the components of the unperturbed tensors G 0 plus A and the
components of the tensor a perturbed to first order in the electric field. For electric Raman
optical activity transition tensors are involved in the same patterns of interference. The
formulae are not reproduced here.
REFERENCE
Long, D.A. (1977). Raman Spectroscopy, McGraw Hill: New York.
328
The Raman Effect

The formulae for scattered intensities and normalized circular intensity differentials
developed in this chapter are brought together in these reference tables for ease of consultation.
The Reference Tables 10.1 to 10.8 are based on five tensor invariants
a2 , 2 , aG0 /c0 , G2 0 /c0 , A2 /c0 . Tables 10.1 to 10.4 relate to the intensities and normalized
circular intensity differentials for incident polarized radiation (ICP) for D /2, 0 and
. Tables 10.5 and 10.6 relate to normal intensities and normalized circular intensity
differentials for scattered circularly polarized radiation (SCP) for D /2, 0 and .
Tables 10.7 and 10.8 relate to intensities and normalized circular intensity differentials
for dual circularly polarized radiation (DCP) in-phase and out-of-phase for D /2, 0
and .
The Reference Table 10.9 gives various intensity sums and differences for ICP
and SCP with D /2, 0 and for the case of eight isotropic invariants
2
a2 , 2 , aG0 /c0 , G2 0 /c0 , A2 /c0 , aG 0 /c0 , G2 0 /c0 , A
/c0 .
Reference Table 10.1
Intensities for incident circularly polarized radiation (ICP) and D /2.
General formulae:

kQ Q s4 Nvi
45a2 C 7 2 45a2 C 2 P cos 2 cos 2
90

P
2
I
C 45aG0 C 7G2 0 C A2 sin 2
c0
2

kQ Q s4 Nvi
4
s
i
2
2
2
I /2; jj , p D
6 C 3G0 A P sin 2 I
90
c0
I /2; ?s , pi D
R i,
L i
pi D
For specific cases, intensity sums and differences can be expressed in the form
kQ Q s4 Nvi
90

aG0 G2 0 A2

f a2 , 2 ,
,
,
c0 c0 c0

I
where the function f is a linear combination of the stated invariants whose coefficients are
given in the Table that follows:
329
Intensity sums and differences
Coefficients of
R i C I /2; ?s ,
L i
I /2; ?s ,
a2
2
aG0
c0
G2 0
c0
A2
c0
90
14
R I /2; ? ,
L
I /2; ? ,
180
28
R i C I /2; jjs ,
L i
I /2; jjs ,
12
R i I /2; jjs ,
L i
I /2; jjs ,
24
8
R i C I /2; ?s C jjs ,
L i
I /2; ?s C jjs ,
90
26
R i I /2; ?s C jjs ,
L i
I /2; ?s C jjs ,
180
52
4
Intensities for incident circularly polarized radiation (ICP) and D 0.
General formulae:
I0; ?s , pi D
kQ Q s4 Nvi
f45a2 C 7 2 C 45a2 C 2 P cos 2 cos 2
90
4
45aG0 C G2 0 A2 P sin 2gI
c0

kQ Q s4 Nvi
I0; jjs , pi D
45a2 C 7 2 45a2 C 2 P cos 2 cos 2
90

4
C 45aG0 C G2 0 A2 P sin 2 I
c0
C
R i,
L i
pi D
kQ Q s4 Nvi
90

aG0 G2 0 A2

f a2 , 2 ,
,
,
c0 c0 c0

I
330
The Raman Effect
Coefficients of
a2
2
aG0
c0
G2 0
c0
A2
c0
R i C I0; ?s ,
L i
I0; ?s ,
R i I0; ?s ,
L i
I0; ?s ,
90
14
360
8
R i C I0; jjs ,
L i
I0; jjs ,
s R i
s L i
I0; jj , I0; jj ,
90
14
360
8
R i C I0; ?s C jjs ,
L i
I0; ?s C jjs ,
R i I0; ?s C jjs ,
L i
I0; ?s C jjs ,
180
28
720
16
16
Intensities for incident circularly polarized radiation (ICP) and D .
General formulae:
kQ Q s4 Nvi
f45a2 C 7 2 C 45a2 C 2 P cos 2 cos 2 C 83G2 0 C A2 gI
90
kQ Q s4 Nvi
f45a2 C 7 2 45a2 C 2 P cos 2 cos 2 C 83G2 0 C A2 P sin 2gI
I ; jjs , pi D
90
R i,
L
pi D
I ; ?s , pi D

2
0
2
kQ Q s4 Nvi
2
2 aG G0 A
I
f a , ,
,
,
90
c0 c0 c0
Coefficients of
a2
2
aG0
c0
G2 0
c0
A2
c0
R i C I ; ?s ,
L i
I ; ?s ,
R i I ; ?s ,
L i
I ; ?s ,
90
14
48
16
R i C I ; jjs ,
L i
I ; jjs ,
R i I ; jjs ,
L i
I ; jjs ,
90
14
48
16
R i C I ; ?s C jjs ,
L i
I ; ?s C jjs ,
s
s R i
s
s L i
I ; ? C jj , I ; ? C jj ,
180
28
96
32
331
Reference Table 10.4 Normalized circular intensity differentials for incident circularly polarized
radiation (ICP) and D /2, 0, .
General formulae:
R i I; ps ,
L i
I; ps ,
R i C I; ps ,
L i
I; ps ,
s
s
s
s
s
p D? , jj or ? Cjj
; ps , i D
D /2, 0 or
; ps , i

/2
/2; ?s , i
245aG0 C 7G2 0 C A2

c0 45a2 C 7 2
/2; jjs , i
23G2 0 A2
3c0 2
/2; ?s C jjs , i
245aG0 C 13G2 0 A2

c0 45a2 C 13 2
0; ?s , i
445aG0 C G2 0 A2

c0 45a2 C 7 2
0; jjs , i
445aG0 C G2 0 A2

c0 45a2 C 7 2
0; ?s C jjs , i
445aG0 C G2 0 A2

c0 45a2 C 7 2
; ?s , i
83G2 0 C A2
c0 45a2 C 7 2
; jjs , i
83G2 0 C A2
c0 45a2 C 7 2
; ?s C jjs , i
83G2 0 C A2
c0 45a2 C 7 2
332
The Raman Effect

and .
Intensities for scattered circularly polarized radiation (SCP) for D /2, 0
The intensities have the general forms

kQ Q s4 Nvi 2
45aG0 C 13G2 0 A2
90c0

45aG0 C G2 0 C 3A2 P cos 2 cos 2 I
kQ Q s4 Nvi
R sC
L s , pi D
I;
45a2 C 13 2 45a2 C 2 P cos 2 cos 2
90

2
45aG0 C 13G2 0 A2 P sin 2 I
C
c0
pi D?i , jji , ?i C jji
R s
L s , pi D
I;
D /2, 0,
For specific cases the intensities can be expressed in the form
kQ Q s4 Nvi
90

aG0 G2 0 A2

f a2 , 2 ,
,
,
c0 c0 c0

I
I; s , pi
Coefficients of
a2
2
aG0
c0
G2 0
c0
A2
c0
R sC
L s , ?i
I /2;
R s
L s , ?i
I /2;
90
14
180
28
R sC
L s , jji
I /2;
R s
L s , jji
I /2;
12
24
8
R sC
L s , ?i C jji
I0;
R s
L s , ?i C jji
I0;
180
28
720
16
16
R sC
L s , ?i , C jji
I ;
s
R
L s , ?i C jji
I ;
180
28
96
32
333
Reference Table 10.6 Normalized circular intensity differentials for scattered circularly polarized
radiation (SCP) and D /2, 0 and .
General formula:
; s , pi D
R s , pi I;
L s , pi
I;
R s , pi C I;
L s , pi
I;
pi D?i , jji or ?i C jji

D /2, 0,
; s , pi

/2; s , ?i
245aG0 C 7G2 0 C A2

c0 45a2 C 7 2
/2; s , jji
23G2 0 A2
3c0 2
/2; s , ?i C jji
245aG0 C 13G2 0 A2

c0 45a2 C 13 2
0; s , ?i C jji
445aG0 C G2 0 A2

c0 45a2 C 7 2
; s , ?i C jji
83G2 0 C A2
c0 45a2 C 7 2
/2
Reference Table 10.7 Intensities for dual circularly polarized radiation (DCP), in-phase and
out-of-phase, for D /2, 0 and .
The intensity sums and differences can be expressed in the form

2
0
2
kQ Q s4 Nvi
2
2 aG G0 A
I
f a , ,
,
,
90
c0 c0 c0
334
Phase
The Raman Effect

Coefficients of
a2
2
aG0
c0
G2 0
c0
A2
c0
DCPin
R s,
R i C I /2;
L s,
L i
I /2;
R s
R i I /2;
L s,
L i
I /2;
45
13
180
52
4
DCPout
L s,
R i C I /2;
R s,
L i
I /2;
L s,
R i I /2;
R s,
L i
I /2;
45
13
DCPin
R s,
R i C I0;
L s,
L i
I0;
R s,
R i I0;
L s,
L i
I0;
180
720
16
16
DCPout
L s,
R i C I0;
R s,
L i
I0;
s
i
s
L ,
R I0;
R ,
L i
I0;
24
DCPin
R s,
R i C I ;
L i,
L i
I ;
R s,
R i I ;
L s,
L i
I ;
24
96
32
DCPout
L s,
R s C I ;
R s,
L i
I ;
i
i
s
L ,
R I ;
R ,
L i
I ;
180
Reference Table 10.8 Normalized circular intensity differentials for dual circularly polarized
radiation (DCP), in-phase and out-of-phase, for D /2, 0 and .
General formulae:
R s,
R i I;
L s,
L i
I;
s
i
s
R ,
R C I;
L ,
L i
I;
L s,
R i I;
R s,
L i
I;
out ; s , i D
L s,
R i C I;
R s,
L i
I;
in ; s , i D
in/out ; s , i
in /2; s , i
out /2; s , i
in 0; s , i
out 0; s , i
in ; s , i
out ; s , i
445aG0 C 13G2 0 A2

c0 45a2 C 13 2
0
445aG0 C G2 0 A2
c0 45a2 C 2
0
43G2 0 C A2
3 2
0
335
Reference Table 10.9 Intensity sums and differences for ICP and SCP with D /2, 0 and
for the case of eight isotropic invariants:
aG0 G2 0 A2 aG 0 G2 0
2
,
, ,
,
and A
a2 , 2 ,
c0 c0 c0 c0 c0
c0
The various intensity combinations are all expressed in the form

0
2
2
0
2
2
aG

a

G
kQ Q s4 Nvi
0
0
I
f a2 , 2 ,
, G , A,
, G , A
90
c0 c0 c0 c0 c0 c0
where the function f is a linear function of the stated invariants whose coefficients are given
in the Table that follows:
Coefficients of
a2
2
aG0
c0
G2 0
c0
A2
c0
aG 0
c0
G2 0
c0
A2
c0
R i C I /2; ?s ,
L i
I /2; ?s ,
R i I /2; ?s ,
L i
I /2; ?s ,
90
14
180
28
R i C I /2; jjs ,
L i
I /2; jjs ,
R i I /2; jjs ,
L i
I /2; jjs ,
12
24
8
R sC
L s , ?i
I /2;
s
R
L s , ?i
I /2;
90
14
180 28
R sC
L s , jji
I /2;
R s
L s , jji
I /2;
12
R i C I0; ?s C jjs ;
L i
I0; ?s C jjs ;
R i I0; ?s C jjs ;
L i
I0; ?s C jjs ;
180
28
360
56
R sC
L s , ?i C jji
I0;
R s
L s , ?i C jji
I0;
180
28
360
40
R i C I ; ?s C jjs ,
L i
I ; ?s C jjs ,
R i I ; ?s C jjs ,
L i
I ; ?s C jjs ,
180
28
360
56
360 40
C24
R sC
L s , ?i C jji
I ;
R sC
L s , ?i C jji
I ;
180
28
360
40
24 360 56
24
8
40
24
24 360 56
360
337
Part Two
APPENDICES
A compendium of mathematical and physical tools relevant to the

treatments of light-scattering phenomena developed in Part One.
339
Introduction
Man is a tool-using animal.
Without tools he is nothing,
with tools he is all.
Thomas Carlyle
Doubtless there will be quite a few readers who are not familiar with all of the physics
and mathematics necessary for the treatment of light-scattering phenomena at the level
of sophistication used here. Certainly the author has found it necessary to make good
some gaps in his knowledge in the course of writing this book. Thus, one reason for this
Appendix is to provide the reader with the necessary tools, so that he is all.
A second reason is the need to bring together careful definitions of the various physical
quantities, their symbols and units. Mere tabulation would not suffice as some discussion is
usually necessary to explain and justify particular choices. In some instances it has proved
convenient to deal with the necessary definitions in the course of a short review of the
relevant topic in physics or mathematics. Thus the justification for those sections, which
might be considered rather elementary, is that they have been included as much to facilitate
the development of the notation as to instruct. An essential adjoint to these Appendices
is the IUPAC publication, Quantities, Units and Symbols in Physical Chemistry, prepared
by Mills et al. (1993).
The Appendices are organized into chapters, with sections and subsections, as in the
main text, but their numbers are prefixed by an A. For example: Chapter A5, Section A5.1,
sub-section A5.1.1 etc. Following the system used in the main text, equations are numbered
serially according to the section; but figures and tables are numbered serially according
to the chapter. Thus, for example, eq. (A5.2.1), but Fig. A5.27 and Table A5.12.
340
The Raman Effect
The chapters are ordered so that as far as possible a given chapter needs only a
knowledge of material in earlier chapters and groups of consecutive chapters cover
related material. A few chapters deal with concepts that will probably be new to some
readers, especially those with a background of chemistry rather than physics. For example,
ClebschGordan coefficients and Wigner 3-j coefficients might fall into this category.
The power and the beauty of techniques of this kind is such that the effort of mastering
them will be repaid amply.
REFERENCE
Mills, I., Cvitas, T., Homann, K., Kallay, N., and Kuchitsu, K. (1993). Quantities, Units and
Symbols in Physical Chemistry, Blackwell Science: Oxford.
341
A1
The Right-handed Cartesian
Axis System and Related
Coordinate Systems
. . . all causes of natural effects have to be
expressed by means of lines, angles
and figures. . . .
Robert Grosseteste (Chancellor of Oxford
University, d. 1253)
A1.1 INTRODUCTION
To express physical phenomena in mathematical form we require in general a reference
frame and within this frame a coordinate system. Formally, such a frame involves a
mathematical representation which assigns to each event in the physical world a unique
place in three-dimensional Euclidean space and a unique instant on a real line. However,
we are concerned with the mathematical representation of physical objects and physical
laws in three-dimensional Euclidean space within a fixed frame. For this purpose we
need only a system of axes and various coordinate systems related to these axes. We now
proceed to define our chosen axis system and the several related coordinate systems that
are useful in the treatment of light-scattering phenomena.
342
The Raman Effect
A1.2 THE RIGHT-HANDED CARTESIAN AXIS SYSTEM
Along the positive z axis
The right-handed cartesian axis system is shown in Fig. A1.1. In this figure O is called
the origin and the mutually perpendicular directed lines Ox, Oy and Oz are called the
positive x, y and z axes, respectively. To avoid ambiguity we explain that the statements light is scattered along the positive z axis, looking along the positive z axis
or looking up the z axis mean along the positive z axis in the direction away from
the origin O. Conversely, along the negative z axis or looking down the z axis mean
along the negative z axis away from the origin O. These statements are illustrated in
Fig. A1.1.
The cartesian axis system shown in Fig. A1.1 is described as right-handed, because the
arrangement of the axes is such that, if we perform a rotation about the z axis in the sense
that would cause a right-handed screw to advance along the positive z axis, the x axis
moves towards the original position of the y axis. If we look along the positive z axis
this right-handed screw motion appears to be clockwise. Conversely, looking along the
negative z axis, this motion appears to be anti-clockwise. An alternative but equivalent
definition of a right-handed axis system is that if the thumb of the right hand is set to
point along the positive z axis, the curl of the fingers indicates the movement of the x
axis towards the original y axis. Figure A1.2 illustrates these alternative descriptions of
a right-handed axis system.
z axis
y axis
x axis
Along the negative z axis
Figure A1.1 The right-handed cartesian axis system.
Sometimes called a right-handed rectangular system.
The Right-handed Cartesian Axis System and Related Coordinate Systems
343
z axis
y axis
x axis
(a)
Figure A1.2(a) For a right-handed cartesian axis system a right-handed screw motion about
the z axis appears to be a clockwise motion when viewed along the positive z axis (looking
up the z axis) and causes the x axis to move towards the original position of the y axis.
z axis
y axis
x axis
(b)
Figure A1.2(b) For a right-handed cartesian axis system, when

the thumb of the right hand is set to point along the positive z axis
the curl of the fingers indicates the movement of the x axis towards
the orginal y axis.
Throughout this book we shall employ the right-handed cartesian axis system. For
conciseness we shall use descriptions like a set of cartesian axes x, y, z or the axis system
x, y, z, a right-handed system always being implied. The use of free and dummy subscripts
as cartesian axis labels is considered in Chapter A2.
We shall refer to an axis system fixed at the centre of the nuclear mass of a molecule,
and parallel to the laboratory axis system as a space-fixed system and an axis system
fixed in the nuclear frame-work, so that it rotates with the molecule as a molecule-fixed
system. As we frequently consider axis systems with different orientations in space we
shall find it convenient to introduce the symbol S for an axis system and add appropriate
distinguishing labels as required.
344
The Raman Effect
A1.3 CARTESIAN COORDINATE SYSTEM

The point P in Fig. A1.3 may be represented by an ordered triple of real numbers (x, y, z)
which are called the cartesian coordinates of P. The first number x, or the x coordinate,
is the directed perpendicular distance from the yz plane to P. Thus x is positive if P lies
on the same side of the yz plane as the positive x axis, negative if it lies on the opposite
side, and zero if P lies in the yz plane. The other two cartesian coordinates, y and z,
are defined in an analogous manner. These definitions of the x, y and z coordinates are
illustrated in Fig. A1.3.
z axis
x
y
P
z
y axis
x axis
Figure A1.3 Cartesian coordinates. The directed distance from the yz plane to P is a
measure of the x coordinate which in this case is a positive number; and similarly for the y
and z coordinates which in this case are also positive numbers.
In the cartesian coordinate system an element of volume d is given by

d D dx dy dz
A1.3.1
and an element of area dA in, say, the xy plane by

dA D dx dy
A1.3.2
A1.4 CYLINDRICAL COORDINATE SYSTEM

In the cylindrical coordinate system the position of the point P in Fig. A1.4 is specified
by , and z. The coordinate is the perpendicular distance of P from the z axis; is
the azimuth angle of the plane containing P and the z axis, measured in the right-hand
screw sense; and z is the directed perpendicular distance from the xy plane to P. The
definitions of , and z are illustrated in Fig. A1.4.
345
z axis
r
P
z
O
y axis
f
x axis
Figure A1.4 The cylindrical co-ordinate system.
The cartesian coordinates are related to the cylindrical coordinates as follows:

x D cos
y D sin
A1.4.1
zDz
The inverse relationships for and are
D jx 2 C y 2 1/2 j
y
y
D tan1
D sin1 2
x
jx C y 2 1/2 j
A1.4.2
In the cylindrical coordinate system an element of volume d is given by

d D d d dz
A1.4.3
A1.5 POLAR COORDINATE SYSTEM

In the polar coordinate system the position of the point P in Fig. A1.5 is specified by r,
and where r is the distance from the origin O, to the point P, is the angle between
the z axis and r, and is the azimuthal angle, defined as the angle between the x axis
and the projection of r onto the xy plane.
The cartesian coordinates are related to the polar coordinates as follows:
x D r sin cos
y D r sin sin
A1.5.1
z D r cos
Often called spherical polar coordinates. However, we use polar coordinates in order to distinguish them
from spherical coordinates, which are complex. These are introduced later in Chapter A6.
346
The Raman Effect

z axis
dr
P
r
dq
r sin q df
rdq
O
y axis
df
f
r sin q
r sin q df
x axis
Figure A1.5 The polar coordinate system.
The inverse relationships are:

r D jx 2 C y 2 C z2 1/2 j
jx 2 C y 2 1/2 j
z
D tan1
D cos1 2
2
z
jx C y C z2 1/2 j
y
y
D tan1
D sin1 2
x
jx C y 2 1/2 j
A1.5.2
In polar coordinates an element of volume d is given by

d D r 2 sin dr d d
A1.5.3
as may be seen by reference to Fig. A1.5.

For the special case where the point P is confined to the xy plane, so that D /2 and
z D 0, we have
x D r cos
A1.5.4
y D r sin
The inverse relationships are:
r D jx 2 C y 2 1/2 j
y
y
D sin1 2
D tan1
x
jx C y 2 1/2 j
A1.5.5
It is also useful to note the following relationships between the polar and cylindrical
coordinates:
D r sin
A1.5.6
z D r cos
347
For the special case where D /2 and z D 0 an element of area dA in the xy plane is
given by
dA D r dr d
A1.5.7
A1.6 COMPLEX COORDINATE SYSTEMS

For some of the matters considered in this book, particularly those which involve treatment
of rotation about an axis, complex coordinates can prove useful. Coordinate systems of
this kind are more conveniently considered later in Chapter A6 which deals with many
aspects of complex numbers and quantities.
349
A2
The Summation Convention
And though thou hadst small Latin,
and less Greek.
Ben Jonson
A2.1 GENERAL DEFINITIONS

We may illustrate the Einstein summation convention by considering the following set of
linear equations
px D xx Ex C xy Ey C xz Ez
py D yx Ex C yy Ey C yz Ez
A2.1.1
pz D zx Ex C zy Ey C zz Ez
These equations define the relationships between the two sets of variables px , py , pz
and Ex , Ey , Ez through the coefficients xx , xy etc. The physical significance of these
equations is dealt with in Chapter A13. They occur in various forms in the theory of light
scattering.
A more condensed, but readily understandable, notation for the above set of equations is

E
A2.1.2
p D
D x, y, z
D x, y, z
By adopting the convention due to Einstein that when a Greek subscript appears twice in
the same term, summation with respect to that subscript is to be understood, eq. (A2.1.2)
A term can consist of a product involving two symbols with Greek subscripts as in eq. (A2.1.3) or a
single symbol with Greek subscripts as in eq. (A2.1.4).
350
The Raman Effect
can be written in a still more condensed form with the summation sign omitted:
p D E
A2.1.3
In such equations is called the free subscript and the dummy subscript. As occurs
twice in the term on the right-hand side of eq. (A2.1.3) summation over is implied.
The free subscript can be either x, y or z so that p is a shorthand notation for the
three quantities px , py and pz . The dummy subscript can also, in this case, be either x,
y or z. It must be carefully noted that is not a shorthand notation for all of the nine
coefficients xx , xy , . . .. If 6D then is a shorthand notation for the six coefficients
xy , yx , yz , zy , xz and zx , but is not a shorthand notation for xx , yy or zz . This
is because the subscript occurs twice in the term and a sum is implied, so that
D xx C yy C zz
A2.1.4
Care is therefore needed when going from the general to the specific case when Greek
subscripts are involved. The distinction between terms such as and is particularly
striking in the case of the Kronecker delta as illustrated for example in Chapter A3. When
it is felt that the conciseness
of the Einstein summation convention might lead to obscurity
we shall use the standard
form.
It will be apparent that with this convention it is necessary to distinguish between free
or dummy subscripts, and specific subscripts. We shall adopt the following convention
throughout this book: Greek letters will be used for free or dummy subscripts; English
letters or numerals will be used for subscripts which denote specific labels, as for example
x, y or z.
The dummy subscript is so named because it can be replaced by another Greek
letter without changing the value of the sum. Thus for example on replacing by
in eq. (A2.1.3) we have
p D E D E
A2.1.5
The replacement of one dummy subscript by another can prove useful in some mathematical developments.
351
A3
Direction Cosines
Ubi materia, ubi geometria.
Johannes Kepler
A3.1 INTRODUCTION
It is frequently necessary to specify the relative orientation of two sets of axes. This
information is needed, for example, in order to obtain the relationships between such
quantities as vectors and tensors in two sets of axes. One method of defining the relative
orientations of axes is to use direction cosines and we now consider their definition and
properties.
A3.2 DEFINITIONS AND PROPERTIES OF DIRECTION COSINES

Consider two sets of cartesian axis systems x, y, z and x 0 , y 0 , z0 with a common origin
O (Fig. A3.1), but different orientations, which we shall label systems S and S0 , respectively. These systems will sometimes be referred to as the unprimed and primed systems,
respectively. The relative orientation of the two sets of axes, S and S0 may be specified
by using direction cosines. A direction cosine is the cosine of the angle between the
positive directions of any two axes. We shall adopt a condensed nomenclature and write
lx0 z for the cosine of the angle between the positive directions of the x 0 and z axes, and
so on. Conversely, cos1 lx0 z is the angle between the x and z axes. Clearly there are
nine direction cosines and nine angles involved in specifying the relative orientation of
the two sets of axes. The nine direction cosines are not, however, all independent. We
The order of writing the subscripts is immaterial since lx0 z D lzx0 and cos1 lx0 z D cos1 lzx0 .
352
The Raman Effect

z
z
cos1lz z
cos1lz y
cos1lz x
Figure A3.1 Two sets of cartesian axis systems x, y, z (S) and x 0 , y 0 , z0 , (S0 ) with a common
origin O but different orientations. The angles between the z0 axis and the x, y and z axes
are shown in terms of inverse direction cosines.
shall now show that there are six relationships between the direction cosines so that only
three of them can be assigned independently.
The geometrical considerations set out in Fig. A3.2 lead to the result that the cosine of
the angle between two position vectors r and r 0 drawn from the origin O to points P1
and P2 , respectively, is given by
cos D lrx lr 0 x C lry lr 0 y C lrz lr 0 z
A3.2.1
where lrx is the direction cosine between r and the x axis and so on. If r and r 0 are
perpendicular to each other eq. (A3.2.1) yields an orthogonality relationship, namely
lrx lr 0 x C lry lr 0 y C lrz lr 0 z D 0
A3.2.2
If r and r 0 are coincident then eq. (A3.2.1) yields normalization relationships, namely
l2rx C l2ry C l2rz D 1
l2r 0 x C l2r 0 y C l2r 0 z D 1
A3.2.3
We now apply these orthonormality conditions to the direction cosines that specify the
relative orientation of the two sets of axes S and S0 . Applying the normalization relationship to x 0 , y 0 and z0 in turn, we obtain
l2x0 x C l2x0 y C l2x0 z D 1
l2y 0 x C l2y 0 y C l2y 0 z D 1
l2z0 x C l2z0 y C l2z0 z D 1
A3.2.4
Direction Cosines
353
Cartesian
coordinates of P1
x 1 = lr x
1
y1 = lry
P1
z1 = lr z
1
r1
distance between P1 and P2

= |{(x 1 x 2)2 + (y 1 y 2)2 + (z1 z 2)2}1/2|
Cartesian
coordinates of P2
q
O
x 2 = lr x
P2
r2
y 2 = lr2y
z 2 = lr2z
Setting |r 1| = |r2| = 1 for simplicity
cos q = 1 12 [(x 1 x2)2 + (y1 y 2)2 + (z1 z 2)2] = x1x2 + y1y2 + z1z2
Therefore
cos q = lr x lr x + lr y lr y + lr z lr z
1
Figure A3.2 Relationship between cos and direction cosines.
Further, as x 0 , y 0 and z0 are mutually perpendicular we can use the orthogonality relationship of eq. (A3.2.2) to obtain
lx0 x ly 0 x C lx0 y ly 0 y C lx0 z ly 0 z D 0
ly 0 x lz0 x C ly 0 y lz0 y C ly 0 z lz0 z D 0
A3.2.5
lz0 x lx0 x C lz0 y lx0 y C lz0 z lx0 z D 0

An equivalent set of six orthonormality relations can be obtained by considering the other
set of axes x, y and z, when we obtain
l2x0 x C l2y 0 x C l2z0 x D 1
l2x0 y C l2y 0 y C l2z0 y D 1
A3.2.6
l2x0 z C l2y 0 z C l2z0 z D 1

and
lx0 x lx0 y C ly 0 x ly 0 y C lz0 x lz0 y D 0
lx0 y lx0 z C ly 0 y ly 0 z C lz0 y lz0 z D 0
A3.2.7
lx0 z lx0 x C ly 0 z ly 0 x C lz0 z lz0 x D 0

If we now introduce the Kronecker delta defined by

0 when i 6D j
ij D
1 when i D j
A3.2.8
354
The Raman Effect
then, using the Einstein summation convention, we may write the orthonormality relationships in the condensed forms
l0 l 0 D 0 0
A3.2.9
l0 l0 D
A3.2.10
where the primed Greek subscripts refer to the S0 system and the unprimed Greek
subscripts to the S system. It should be noted carefully that, in writing terms like l0 ,
summation is not implied because 0 and refer to different axis systems.
Some examples will help to clarify this compact notation. Using 0 D 0 D x 0 in
eq. (A3.2.9) we obtain the first of the eqs. (A3.2.4) and using D D x in eq. (A3.2.10)
we obtain the first of the eqs. (A3.2.6). However, using 0 D x 0 , 0 D y 0 in eq. (A3.2.9)
we obtain the first of the eqs. (A3.2.5). Note that the introduction of the appropriate
Latin subscripts means that the Kronecker delta in eqs. (A3.2.9) and (A3.2.10) becomes
x0 x0 D 1, whereas 0 0 D 3.
Because of the six relationships implied in such orthonormality conditions, only three
of the nine direction cosines involved in specifying the relative orientation of two sets
of cartesian axes, can be assigned independently. A representation of a rotation of axes
using three independent parameters is treated in Chapter A5, which deals with the Euler
angles.
A3.3 DEFINITIONS OF DIRECTION COSINES IN OTHER

COORDINATE SYSTEMS
An alternative definition of direction cosines in terms of complex coordinates is considered
in Chapter A9.
In such situations some authors deliberately use two different sets of Greek indices instead of and 0 ,
and 0 so that there can be no possibility of ambiguity concerning terms like l0 . We feel that once having
drawn attention to this problem the introduction of special nomenclature is unnecessary.
355
A4
Isotropic Averages of Products of
Direction Cosines
Die ganzen Zahlen hat Gott gemacht;
alles anderes ist Menschenwerk.
Leopold Kronecker
A4.1 INTRODUCTION
Isotropic averages of products of direction cosines are needed, for example, when the
relationships between quadratic products of tensor components in two axis systems have
to be averaged over all orientations of the one axis system relative to the other. The
calculation of such averages is based on the general result that
1
for n D 2k
n
A4.1.1
hcos i D 2k C 1
0
for n D 2k C 1
where k D 0, 1, 2, 3 . . . and the symbol h. . .i indicates an isotropic space average, that
is an average in which all orientations are taken to be equally probable. The classical
calculations are relatively tedious and we refer the reader to the literature for the details.
Useful treatments have been given by, for example, Woodward (1972) and Barron (1982).
Here, we present a summary of the relevant results.
Following Chapter A1 we define the two cartesian axis systems as follows: x, y, z is
a space-fixed axis system with associated Greek subscripts , , , ; and x 0 , y 0 , z0 is a
An alternative method for finding the isotropic averages of second rank tensors is given in Chapter A14,
Section A14.7.
356
The Raman Effect
molecule-fixed axis system with associated Greek subscripts 0 , 0 , 0 , 0 . The isotropic

averages of the products of direction cosines are obtained by averaging over all orientations of the molecule-fixed axis system relative to the space-fixed axis system.
A4.2 SPECIFIC ISOTROPIC AVERAGES OF PRODUCTS OF

DIRECTION COSINES
For the isotropic average of a single direction cosine, for example, it follows directly
from eq. (A4.1.1) that, taking the angle between the x axis and the x 0 axis to be ,
hlxx0 i D hcos i D 0
A4.2.1
The same result applies to any single direction cosine and thus we may write
hlx0 i D hly0 i D hlz0 i D 0
0 D x 0 , y 0 , z0
A4.2.2
For the isotropic average of a product of two direction cosines we have two cases to
consider: those where the two direction cosines are the same, as for example hl2xx0 i; and
those where they are not the same, as for example hlxx0 lyx0 i. The first case also follows
directly from eq. (A4.1.1). For example
hl2xx0 i D hcos2 i D
1
3
A4.2.3
This result applies to any pair of identical direction cosines so that we have
hl2xx0 i D hl2yx0 i D hl2zx0 i D hl2xy 0 i D hl2yy 0 i D hl2zy 0 i D hl2xz0 i D hl2yz0 i D hl2zz0 i D
1
3
A4.2.4
These results can also be obtained by isotropic averaging of the normalization relationships
given by eqs. (A3.2.4) and (A3.2.6). For example, using the first of the equations (A3.2.6)
we can write
hl2xx0 i C hl2xy 0 i C hl2xz0 i D 1
A4.2.5
and as the three averages are equal each has the value 1/3. This procedure can be applied to
the orthogonality relationships given by eqs. (A3.2.5) and (A3.2.7) to obtain the isotropic
averages of pairs of different direction cosines. For example, using the first of the equations
(A3.2.7) we have
hlxx0 lyx0 i C hlxy 0 lyy 0 i C hlxz0 lyz0 i D 0
A4.2.6
and as the three averages are equal they must each be separately zero. The results for
the isotropic averages of products of two direction cosines can be brought together in the
condensed form
1
hlx0 lx0 i D hly 0 ly 0 i D hlz0 lz0 i D
A4.2.7
3
Isotropic Averages of Products of Direction Cosines
357
where is the second-rank tensor invariant defined in Chapter A10, and all other types
of averages are equal to zero. For example if we put D D x we have three of the
isotropic averages implied in eq. (A4.2.4). If we put D x, D y 6D we have the
three zero isotropic averages implied in eq. (A4.2.6).
For the isotropic average of a product of three direction cosines we have
hlxy 0 lyz0 lzx0 i D hlxz0 lyy 0 lzx0 i D hlxz0 lyx0 lzy 0 i D hlxx0 lyz0 lzy 0 i
A4.2.8
1
D hlxx0 lyy 0 lzz0 i D hlxy 0 lyx0 lzz0 i D
6
All other types of isotropic averages are zero. These results can be expressed in the
condensed form
1
A4.2.9
hlx0 ly 0 lz 0 i D 0 0 0
6
where 0 0 0 is the third-rank tensor invariant defined in Chapter A10 and 0 , 0 , 0 D
x 0 , y 0 , z0 .
For the isotropic average of a product of four direction cosines there are several cases
to consider. If all four direction cosines are the same we have such averages as
1
A4.2.10
5
If the four direction cosines consist of products of pairs of identical direction cosines then
we have such averages as
1
hl2xx0 l2xy 0 i D
15
A4.2.11
1
2 2
hlxx0 lyx0 i D
15
when both pairs have one axis in common. However, if the pairs have no axes in common
we have averages of the type
2
hl2xx0 l2yy 0 i D
A4.2.12
15
Finally, if the direction cosines do not occur in pairs the only non-zero averages are of
the type
1
hlxx0 lyx0 lxy 0 lyy 0 i D
A4.2.13
30
These non-zero terms are characterized by having the axes occurring in pairs.
These results can be summarized in the condensed forms
hlx0 lx 0 lx 0 lx0 i D hly0 ly 0 ly 0 ly0 i D hlz0 lz 0 lz 0 lz0 i
A4.2.14
1
D 0 0 0 0 C 0 0 0 0 C 0 0 0 0
15
and
hlx0 lx 0 ly 0 ly0 i D hly0 ly 0 lz 0 lz0 i D hlx0 lx 0 lz 0 lz0 i
A4.2.15
1
D 40 0 0 0 0 0 0 0 0 0 0 0
30
hl4xx0 i D
358
The Raman Effect
where 0 0 0 0 etc. are fourth-rank tensor invariants defined in Chapter A10. All other
type of averages are zero. For example if we put 0 D 0 D x 0 and 0 D 0 D y 0 in
eq. (A4.2.14) then the Kronecker delta term becomes 1/15 and we obtain eq. (A4.2.11).
The isotropic average of the product of five direction cosines has been treated by
Kielich (1968/69) and by Boyle and Matthews (1971). The procedure involved is rather
complicated and will not be considered here. The results are included in section A4.3.
A4.3 GENERAL FORMULAE

The isotropic averages of products of direction cosines can be presented in very compact
forms if the properties of the Kronecker delta and the Levi-Cività or alternating tensor
are fully exploited. The results given below are based on Hecht and Nafie (1991). Their
symmetry is appealing and their space requirements economic but their unravelling is
demanding!
hl0 i D 0
1
hl0 l 0 i D 0 0
3
1
hl0 l 0 l 0 i D 0 0 0
6
1
hl0 l 0 l 0 l0 i D [ 40 0 0 0 0 0 0 0 0 0 0 0
30
C 0 0 0 0 C 40 0 0 0 0 0 0 0
A4.3.1
A4.3.2
A4.3.3
C 0 0 0 0 0 0 0 0 C 40 0 0 0 ] A4.3.4

1
hl0 l 0 l 0 l0 l
0 i D
0 0 0 0
0 C
0 0 0
0 0
30
C
0 0 0 0
0 C

0 0 0 0 0
C
0 0 0
0 0 C
0 0
0 0 0
C

0 0 0 0 0 C
0 0
0 0 0
C
0
0 0 0 0 C
0
0 0 0 0
A4.3.5
Chapter A10, Section A10.4.3 deals with the Levi-Cività tensor.
REFERENCES
Barron, L. D. (1982). Molecular Light Scattering and Optical Activity, Cambridge University
Press: Cambridge.
Boyle, L. L. and Mathews, P. S. C. (1971). Int. J. Quant. Chem. 5, 381.
Hecht, L. and Nafie, L. A. (1991). Molecular Physics 72, 441.
Kielich, S. (1968/69). Bulletin de la Societe des Amis des Sciences et des Lettres de Poznan B21, 47.
Woodward, L. A. (1972). Introduction to the Theory of Molecular Vibrations and Vibrational Spectroscopy, Clarendon Press: Oxford.
359
A5
The Euler Angles and the
Rotation Operator
Revolutions are not made with rose-water.
Edward Bulwer-Lytton
A5.1 INTRODUCTION
Any rotation of a rigid body may be defined by three real numbers. The most useful
method of defining these numbers is that due to Euler and involves the so-called Euler
angles. A clear and unambiguous definition of the Euler angles is necessary if subsequent
ambiguities are to be avoided. As there is no accepted convention for choosing the Euler
angles we now set out our choice of definition in detail and couple this with a warning that
care is required when using the literature in which other definitions are not infrequently
employed.
A5.2 DEFINITIONS OF THE EULER ANGLES AND THE

ROTATION OPERATOR
The general displacement of a rigid body due to a rotation about a fixed point may be
obtained by performing three rotations about two of three mutually perpendicular axes
fixed in the body. In accord with the definition given in Chapter A1 we assume a righthanded axis system and we define a positive rotation about a given axis to be one which
A useful and salutary account of the inconsistencies in the literature is given by Silver (1976) in the first
chapter of his book, Irreducible Tensor Methods.
360
The Raman Effect

z
Figure A5.1 Positive rotations about the x, y and z axes of a right-handed cartesian axis
system.
would cause a right-handed screw to advance along the positive direction of that axis. For
example, a rotation about the z axis which carries the x axis towards the original position
of the y axis would be considered to be positive. Figure A5.1 illustrates positive rotation
about the x, y and z axes of a right-handed cartesian axis system.
To facilitate the description of the three rotations we use S to represent the original
(and hence space-fixed) set of cartesian axes x, y, z; S1 to represent the set of cartesian
axes, x1 , y1 , z1 after the first rotation; S2 to represent the set of cartesian axes x2 , y2 , z2
after the second rotation and S3 to represent the set of cartesian axes x3 , y3 , z3 after the
third rotation.
With the help of Fig. A5.2 we now consider the positive rotations which are required
for the general rotational displacement described above. These are to be performed successively in the following order:
(a) A rotation 0 < 2 about the z axis, of the axis system S (x, y, z ) to give a new
axis system S1 (x1 , y1 , z1 with z1 D z). The axis of this rotation is commonly called
the vertical (Fig. A5.2(a))
(b) A rotation 0 < about the y1 axis of the axis system S1 , to give a new
axis system S2 (x2 , y2 , z2 with y2 D y1 ). This rotation axis is called the line of nodes
(Fig. A5.2(b))
(c) A rotation 0 < 2 about the z2 axis of the axis system S2 , to give a new axis
system S3 (x3 , y3 , z3 with z3 D z2 ). This rotation axis is called the figure axis and its
position depends on the previous rotations and (Fig. A5.2(c)) It should be noted
that although the possible values of , and are restricted, we do not have a one
to one correspondence between rotations and parameters for all possible rotations; for
example, a rotation symbolized by 0 is identical with that symbolized by 0 0 0 ,
if C D 0 C 0 (Fig. A5.2(c))
Note that the rotation is limited to angles less than , whereas the rotations and are limited to
angles less than 2. These ranges for , and are sufficient to allow any final position to be attained as
may be verified quite readily.
The Euler Angles and the Rotation Operator
361
z, z1
z, z1
z, z1
z2
z2, z3
y3
x1
y1
Dz (a)
Rotation about
the vertical axis (z)
(a)
y
y y1 2
x1
x2
x1
x2
x3
Dy1 (b)
Dz2 (g)
Rotation about the

line of nodes (y1)
Rotation about the

figure axis (z2)
(b)
y1 y2
(c)
Figure A5.2 The three successive rotations (a), (b) and (c), described on page 360 and
represented by Dz , Dy1 and Dz2 respectively.
A practical advantage of the Euler angles is that the polar coordinates and (defined
with respect to the original axis system S) of the z axis in its final position, are identical
with the Euler angles and , respectively.
In the description of the general rotation just given, the rotations and have been
defined with respect to the intermediate axis systems S1 and S2 . In many applications it
is convenient to refer rotations to the original fixed axis system S and we now consider
how this may be effected.
First it will be convenient to introduce an operator notation for describing a rotation
about an axis. A rotation about an axis z will be represented by the operator Dz
and so on. Operators corresponding to successive rotations are ordered from right to left.
Thus the three successive rotations (a), (b) and (c) described above would be represented
by Dz2 Dy1 Dz .
Using this rotation operator notation the result we seek can be expressed as follows:
Dz2 Dy1 Dz Dz Dy Dz
A5.2.1
The reversal of the order of the angles , and should be noted. The above result
states that the final result of performing the three rotations about the axes z, y1 and z2 can
also be achieved by performing three appropriate rotations successively about the axes z,
y and z of the original space-fixed system S (x, y, z). We now outline the proof of this
equivalence, and thereby illustrate the usefulness of rotation operators.
The proof is based on a general equivalence relationship of the type
Dy1 Dz Dy Dz
A5.2.2
This states that the effect of a rotation of about the y1 axis is equivalent to first moving
the y1 axis back to y with a rotation of about the z axis; then rotating about the y axis
362
The Raman Effect
through an angle ; and finally moving the y axis back to y1 with a rotation of about
the z axis. This may be regarded as reasonably self-evident.
It follows from this that also
Dz2 Dy1 Dz1 Dy1
A5.2.3
We now substitute eq. (A5.2.2) into eq. (A5.2.3) and note that Dz1 D Dz , that rotations about the same axis commute and that Dz Dz D I. We then find that
Dz2 Dz Dy Dz Dy Dz
A5.2.4
Finally, if we substitute eqs. (A5.2.4) and (A5.2.2) into the left-hand side of eq. (A5.2.1)
we obtain, after a little manipulation, the right-hand side of eq. (A5.2.1), as desired.
The equality given in eq. (A5.2.1) is useful because it is often easier to consider a
coordinate transformation as a series of rotations about a fixed coordinate system.
A5.3 THE RELATIONSHIP OF THE EULER ANGLES TO THE

POLAR COORDINATES
The polar coordinates and , referred to the original axis system S of the z axis in its
final position, are identical to the Euler angles and , respectively. We now introduce
a new label for , namely and so complete an alternative set of labels for the Euler
angles, , and such that , and . The set , and will be used
henceforth.
A5.4 DIRECTION COSINES AND EULER ANGLES

The nine direction cosines which relate the initial or space-fixed cartesian axis system
S (x, y, z) and a general rotated cartesian axis system S0 x 0 , y 0 , z0 can be expressed in
terms of the three Euler angles. These relationships are presented in Table A5.1 which
gives the elements of a direction cosine matrix which we label Eu D0 . In this notation
D is used to indicate a direction cosine matrix; the subscripts 0 indicate that D relates
the cartesian coordinates of the rotated system S0 x 0 , y 0 , z0 to the cartesian coordinates of
the initial system Sx, y, z; and the presuperscript Eu indicates that the direction cosines
l0 are expressed as functions of the Euler angles , and . Thus an entry in row y 0
and column x gives the direction cosine ly 0 x in terms of , and .
For completeness we give in Table A5.2 the elements of the direction cosine matrix
c D 0 . Here the presuperscript c indicates that the direction cosines are given in the basic

cartesian forms l0 which involve the cosine of the angle between the axes 0 and .
This notation for direction cosine matrices will be developed further in later chapters
of this Appendix.
This set of angles is also used by Wilson, Decius and Cross (1955), and Stanton (1973). Bhagavantam
(1966) uses instead of for the Euler angle .
The Euler Angles and the Rotation Operator
363
Table A5.1 Elements of the direction cosine matrix
Eu
D 0

Eu
D 0
0
x
y0
z0
cos cos cos sin sin

cos cos sin sin cos
sin cos
cos sin cos C cos sin

cos sin sin C cos cos
sin sin
sin cos
sin sin
cos
Table A5.2 Elements of the direction cosine matrix c D0

c
D 0
0
x
y0
z0
lx 0 x
ly 0 x
lz 0 x
lx 0 y
ly 0 y
lz 0 y
lx 0 z
ly 0 z
lz 0 z
REFERENCES
Bhagavantam, S. (1942). Scattering of Light and the Raman Effect, Chemical Publishing Company:
New York.
Silver, B. L. (1976). Irreducible Tensor Methods, Academic Press: New York.
Stanton, L. (1973). J. Raman Spectrosc. 1, 53.
Wilson, E. B., Decius, J. C. and Cross, P. C. (1955). Molecular Vibrations, McGraw-Hill: New
York.
365
A6
Complex Numbers and
Quantities
Imaginary numbers are a fine and
wonderful refuge of the divine spirit,
almost an amphibian between
being and non-being.
Gottfried Leibnitz
A6.1 INTRODUCTION
Terms such as imaginary transition tensor, imaginary third-order susceptibility, often
puzzle newcomers to the field. To clarify such difficulties this chapter presents a short
account of the basic properties of complex numbers and indicates how their use can
facilitate the description of certain physical quantities.
A complex number differs from a real number in much the same way as a vector
differs from a scalar. Thus a complex number is an ordered pair of real numbers which
are subject to special rules for the basic mathematical operations; and a vector in threedimensional space is an ordered triplet of three numbers which are subject to another set
of special rules for the basic mathematical operations.
A6.2 DEFINITIONS AND OPERATIONS

Let the complex number aQ 1 , be represented by the ordered pair of real numbers, x1 and
y1 , that is
A6.2.1
aQ 1 D x1 , y1
366
The Raman Effect
and the complex number aQ 2 be represented by the ordered pair of real numbers (x2 , y2 ),
that is
A6.2.2
aQ 2 D x2 , y2
The tilde is used throughout to indicate a complex quantity. The complex conjugate of
aQ 1 , is designated by aQ 1 , and is defined to be
aQ 1 D x1 , y1
A6.2.3
Similarly, the complex conjugate of aQ 2 is

aQ 2 D x2 , y2
A6.2.4
It should be noted that the pair x1 , 0 is called the real number x1 , and the pair 0, y1
is called the imaginary number y1 .
By definition the sum of aQ 1 and aQ 2 is
aQ 1 C aQ 2 D x1 C x2 , y1 C y2
A6.2.5
aQ 1 aQ 2 D x1 x2 , y1 y2
A6.2.6
aQ 1 aQ 2 D x1 x2 y1 y2 , x1 y2 C x2 y1
A6.2.7
the difference is
the product is
and the quotient is
aQ 1
D
aQ 2
Also
and
x1 x2 C y 1 y 2 y 1 x2 y 2 x1
,
x22 C y22
x22 C y22
dQa1
D
d
dx1 dy1
,
d d
aQ 1 d D
A6.2.8
A6.2.9
x1 d,
y1 d
A6.2.10
where a1 , x1 and y1 are functions of the variable .

It can be seen that each mathematical operation on an ordered pair produces another
ordered pair.
We now set
A6.2.11
aQ 1 D x1 C iy1
and
aQ 2 D x2 C iy2
A6.2.12
where i2 D 1. Clearly the complex conjugates of aQ 1 , and aQ 2 are obtained simply by
changing the sign of i. Thus
aQ 1 D x1 iy1
A6.2.13
aQ 2
A6.2.14
D x2 iy2
Complex Numbers and Quantities
367
It should be noted that the definitions given in eqs. (A6.2.11) to (A6.2.14) have involved
a choice of the sign of i associated with aQ 1 , and aQ 1 (and aQ 2 and aQ 2 ). The choice made
here is that traditionally used in mathematical texts. However, in physics, the freedom of
choice may be constrained by other considerations, as will emerge later.
It is readily seen that with aQ 1 and aQ 2 defined by eqs. (A6.2.11) and (A6.2.12), respectively, the mathematical operations considered above are all duplicated. Thus
aQ 1 C aQ 2 D x1 C x2 C iy1 C y2
A6.2.15
aQ 1 aQ 2 D x1 x2 C iy1 y2
A6.2.16
aQ 1 aQ 2 D x1 x2 y1 y2 C ix1 y2 C x2 y1

x1 x2 C y 1 y 2
y 1 x2 y 2 x1
aQ 1
D
Ci
aQ 2
x22 C y22
x22 C y22
dx1
dy1
dQa1
D
Ci
d
d
d

aQ 1 d D x1 d C i y1 d
A6.2.17
A6.2.18
A6.2.19
A6.2.20
In the definition of aQ 1 given by eq. (A6.2.11), x1 is called the real part of the complex
number aQ 1 and y1 is called the imaginary part. We can express this by writing
It is useful to note that
RefQa1 g D x1
A6.2.21
ImfQa1 g D y1
A6.2.22
1
RefQa1 g D fQa1 C aQ 1 g
2
A6.2.23
and
i
1
A6.2.24
ImfQa1 g D fQa1 aQ 1 g D fQa1 aQ 1 g
2
2i
The two alternative forms for ImfQa1 g should be noted. They illustrate the sign change
when i is moved from the numerator to the denominator.
It is important to appreciate that y1 , which is the coefficient of i in eq. (A6.2.11), is a
real number, but iy2 is of course an imaginary quantity.
A complex number aQ 1 becomes pure real when the imaginary part is zero, that is
y1 D 0; and pure imaginary when the real part is zero, that is x1 D 0. Also, aQ D 0 implies
x1 D 0 and y1 D 0. It follows that aQ 1 D aQ 2 means that x1 D x2 and y1 D y2 . The identity
of two complex numbers is similar to the case of the identity of two vectors A1 and A2
in three-dimensional space. The identity A1 D A2 requires three equalities between the
three corresponding components of A1 and A2 .
A6.3 GRAPHICAL REPRESENTATION OF COMPLEX NUMBERS

A complex number aQ , can be represented graphically as follows. If the real part of aQ is
taken as the abscissa, and the imaginary part of aQ as the ordinate in a two-dimensional
cartesian coordinate system, then aQ can be represented as a point in a plane. The x axis
368
The Raman Effect

y
Imaginary
axis
a~
Real axis
x
~
a*
Figure A6.1 Graphical representation (or Argand diagram) of a complex number aQ and its
conjugate complex aQ using cartesian coordinates and polar coordinates.
is termed the real axis, the y axis the imaginary axis, and the set of points representing
all possible values of aQ the complex plane. This is illustrated in Fig. A6.1.
Such a geometrical representation of complex numbers as points with coordinates x
and y is equivalent to their representation as vectors with components x and y. When
complex numbers are represented by vectors in the complex plane the resulting figure
is often called an Argand diagram. In the complex plane the addition and subtraction of
complex numbers correspond to the addition and subtraction of vectors. This is illustrated
in Fig. A6.2.
y
x1 + x2
a~1 + a~2
a~2
y1 + y2
y2
a~1
y1
x1
y1 y2
x2
x1 x2
a~1 a~2
Figure A6.2 Addition and subtraction of complex numbers. The two vectors with components x1 , y1 and x2 , y2 are indicated by the arrows
. The resultant vector from addition
of these vectors has components x1 C x2 , y1 C y2 and is indicated by
; and the resultant
vector from subtraction has components x1 x2 , y1 y2 and is indicated by
.
369
The length of the vector representing a complex number aQ is called the modulus or the
absolute value of a and is denoted as jaj. It is always positive and real. Clearly
jQaj D jx 2 C y 2 1/2 j
A6.3.1
If we introduce aQ , the complex conjugate of aQ , then we have

jQaj D jaQ aQ 1/2 j
A6.3.2
As Fig. A6.1 shows, the operation of forming the complex conjugate corresponds to a
reflection with respect to the real axis, x.
A6.4 COMPLEX NUMBERS AND POLAR COORDINATES

If we use the polar coordinates r, instead of the cartesian coordinates (x, y) for the
point representing aQ in the complex plane, it can be seen from Fig. A6.1 that
aQ D rcos C i sin
A6.4.1
aQ D rcos i sin
A6.4.2
and
In these so-called polar forms of aQ and aQ
r D jQaj
The angle is called the phase of aQ and its principal value is given by
y
<
D tan1
x
We now introduce the identity
A6.4.3
A6.4.4
ei D cos C i sin
A6.4.5
ei D cos i sin
A6.4.6
and its conjugate complex

These relationships are readily established by writing series expansions of ei , ei , cos
and sin in terms of . It should be noted that ei and ei are complex numbers because
each is the sum of two terms, one real and the other imaginary.
Using eq. (A6.4.5) we can write eq. (A6.4.1) as
aQ D rei
A6.4.7
aQ D rei
A6.4.8
from which follows
is sometimes called the argument of aQ and r the amplitude of aQ .
370
The Raman Effect
The use of eq. (A6.4.5) to give eq. (A6.4.7) for aQ is imposed by our choice for the
definition of aQ as in eq. (A6.2.11). We note that using either eqs. (A6.4.1) and (A6.4.2),
or eqs. (A6.4.7) and (A6.4.8) gives
aQ 1 aQ 1 D r 2
A6.4.9
In polar form, multiplication and division are performed quite simply. Thus if aQ 1 D r1 ei
and aQ 2 D r2 ei
and
aQ 1 aQ 2 D r1 r2 ei1 C2 D r1 r2 fcos1 C 2 C i sin1 C 2 g
A6.4.10
aQ 1
r1
r1
D ei1 2 D fcos1 2 C i sin1 2 g
aQ 2
r2
r2
A6.4.11
These equations have a straightforward interpretation. When aQ 1 is multiplied by aQ 2 the

vector represented by aQ 1 is changed so that its new length is r2 times its original length,
and is turned anticlockwise through 2 ; and when aQ 1 is divided by aQ 2 it is changed so
that its new length is r21 times its original length and is turned clockwise through 2 .
The product aQ 1 aQ 2 is given by
aQ 1 aQ 2 D r1 r2 ei1 2 D r1 r2 fcos1 2 C i sin1 2 g
A6.4.12
The real part of (A6.4.12) is the scalar product of the vectors aQ 1 and aQ 2 , and the imaginary
part their vector product, since (1 2 ) is the angle between the two vectors.
The reciprocals of aQ and aQ are given by
and
1
1
D fcos i sin g
aQ
r
A6.4.13
1
1
D fcos C i sin g
aQ
r
A6.4.14
A6.5 COMPLEX QUANTITIES AND PHYSICAL PHENOMENA

We have so far described a number of the basic mathematical properties of complex
numbers or ordered pairs. Complex quantities are much used in the description of physical
phenomena because of their convenience. The uninitiated are urged to reject any ghostly
images conjured up by the terms real and imaginary, and to treat these terms strictly as
labels which distinguish two quantities in an ordered pair, there being defined rules of
combination for such pairs.
The use of complex quantities is particularly advantageous when treating physical
phenomena which can only be described in terms of a pair of related quantities. The
following example illustrates this. When an oscillating, sinusoidal electric field is incident
on a material system, an induced electric polarization is produced whose measure is the
induced electric dipole moment per unit volume. This polarization is proportional to the
electric field strength, provided the electric field is not too strong, and the constant of
371
proportionality is the first-order electric susceptibility. The polarization is not necessarily

in phase with the electric field and in general may have one part which is in phase with
the field and another part which is /2 out of phase with the field. Thus two quantities
are needed to define the first-order electric susceptibility. We can say that the first-order
electric susceptibility is complex and represent it by an ordered pair, thus:
Q D R , Im
R
A6.5.1
Im
where
is the real part of the complex susceptibility Q and
is the imaginary part.
If we adopt the polar description of a complex quantity based on eq. (A6.4.5) we see
that a phase change of /2 corresponds to Ci and so in this particular case we can write
Q in the following form
A6.5.2
Q D R C i Im
Many examples of physical quantities which are conveniently represented as complex
quantities occur in this book. These include the wave vector kQ , the polarizability aQ , the
electric field amplitude EQ 0 , the refractive index nQ and the electric permittivity eQ . The
forms which must be used for such complex quantities and their conjugate complexes are
discussed in later chapters of this Appendix.
A6.6 SPHERICAL COORDINATES

A set of complex coordinates usually known as spherical coordinates is particularly
useful when dealing with some of the properties of vectors and tensors under rotation.
In spherical coordinates the position of a point P in a cartesian axis system is specified by the three coordinates s1 , s0 and s1 which are defined in terms of the cartesian
coordinates of P as follows:
1
A6.6.1
s1 D p x C iy
2
s0 D z
A6.6.2
1
A6.6.3
s1 D p x iy
2
These definitions involve specific choices of the phase and normalization factors for s1
and s1 . The reasons for these choices are related to the need for consistency with the
conventions of Condon and Shortley (1935). The implications of this are explained in
Chapter A9.
The inverse relationships are
1
A6.6.4
x D p s1 s1
2
i
y D p s1 C s1
A6.6.5
2
z D s0
A6.6.6
Sometimes called circular coordinates, particularly in the earlier literature.
372
The Raman Effect
The spherical coordinates are related to the polar coordinates, defined in Chapter A1,
as follows:
1
s1 D p r sin ei
A6.6.7
2
A6.6.8
s0 D r cos
1
s1 D p r sin ei
A6.6.9
2
For rotation about the z axis, only the angle changes, and r and remain constant.
Thus the spherical coordinates s1 and s1 do not mix as a result of this rotation, whereas
the cartesian coordinates x and y do. This valuable property of spherical coordinates is
discussed further in Chapters A7 and A9 when dealing with the matrix representation of
rotations.
REFERENCE
Condon, E. U. and Shortley, G. H. (1935). The Theory of Atomic Spectra, Cambridge University
Press: Cambridge.
373
A7
Some Properties of Matrices
All that isnt Belgrave Square is
Strand and Picadilly.
W.S. Gilbert
A7.1 INTRODUCTION
Matrices are invaluable tools for the effective treatment of many of the mathematical
developments used in this book. In this section we define our nomenclature and consider
certain special properties of matrices which are relevant to the theoretical treatments we
deal with. We shall, however, assume a knowledge of the basic properties of determinants
and matrices and refer the uninitiated to standard texts.
A7.2 NOMENCLATURE
We shall always use bold type to represent matrices. In the general treatment considered
in this chapter we shall use bold, italic, upper case Latin letters, for example A. In
the main text where specific quantities are being considered other bold type will often
be appropriate. An example is the matrix of polarizability tensor components which is
represented by a.
The elements of a general matrix, will be denoted by italic upper case Latin letters,
with two subscripts which denote the row and the column whose intersection defines the
location of the matrix element. For example, Aij represents the element of the general
matrix A which is located at the intersection of the ith row and jth column. Similarly,
ij represents the ij element of the specific matrix a. When it is necessary to designate a
component of a matrix formed by rearrangement of another matrix, the use of brackets is
374
The Raman Effect
Q formed by interchanging the rows and columns

necessary. For example, for the matrix A
Q
Q ij D Aji , that is the ij component of A
of the matrix A, we write for the components A
is the ji component of A.
of the square matrix A we write det A; for the trace of the square
For the determinant
matrix A, defined as i Aii we write trfAg; and for the unit matrix, which has Aij D 1 if
i D j and Aij D 0 if i 6D j, we write I . An alternative definition of I is a matrix whose
elements are given by the Kronecker delta ij .
One of the valuable properties of matrices is that relatively complicated mathematical relationships may be expressed in simple equations in which only the symbols for
matrices are used, e.g. C D AB. We shall use the term compact matrix notation to describe
equations written in this form.
When the elements of a matrix are set out as an array they will be enclosed in square
brackets; and the order (or dimensions) of an array will be given in round brackets with
the numbers of rows first, followed by the number of columns. The following examples
illustrate these conventions. A column matrix with n elements is an n 1 array; and a
row matrix with m elements is a 1 m array.
A square matrix which is a 2 2 array of elements would be written

A11 A12
A21 A22
and a column matrix B which is a 2 1 array as

B11
B21
and a row matrix C which is a 1 2 array as
[ C11
C12 ]
A7.3 SOME SPECIAL MATRICES

Q and is formed by interchanging the rows
Transposed matrix. This is represented by A
and columns of the matrix A. Thus, using the bracket notation already explained we
Q ij D Aji . It should be carefully noted that when a product of
have for the components, A
matrices is transposed the sequence of the matrices forming the product must be reversed.
Thus
D BQ A
Q
A7.3.1
AB
This rule applies whatever the number of matrices in the product. Two special cases of
transposed matrices are of interest. When AQ D A the matrix A is said to be symmetric.
We avoid jAj for the determinant of A. We use jaj for the modulus of a, see p. 369.
Unlike some authors we do not usually use special symbols to distinguish row and column matrices from
other matrices, although this practice will occasionally be convenient. Section A7.4 provides an example of
this.
375
Q D A the
Such a matrix has Aij D Aji for all i, j and is necessarily square. When A
matrix A is said to be skew symmetric. Such a matrix has Aij D Aji for all i, j and is
necessarily square with all components on the leading diagonal zero.
Matrix of the cofactors. This is represented by A0 and is the matrix of the cofactors of
the elements of A. The cofactor element A0 ij is 1iCj times the value of the complementary minor of A, that is the determinant formed from the elements of A remaining
after the row i and column j in which the element Aij appears have been struck out.
Adjoint matrix . This is represented by AQ 0 and is formed by taking the transpose of A0 ,
the matrix of the cofactors of the elements of A. Thus AQ 0 ij D A0ji .
Singular matrix.
This is a square matrix A for which

det A D 0
A7.3.2
A matrix whose determinant is non-zero is said to be non-singular.

Inverse matrix. This is represented by A1 and defined such that
AA1 D A1 A D I
A7.3.3
Q 0 , the adjoint of A, by the determinant of A;
The inverse of A is formed by dividing A
that is
Q 0 / det A
A7.3.4
A1 D A
Only square matrices have inverses. When forming the inverse of a product of matrices
AB 1 D B 1 A1
A7.3.5
This rule applies whatever the number of matrices in the product.

Orthogonal matrix. A square matrix A, with real elements is said to be orthogonal if its
inverse is equal to its transpose. That is
Q or AA
Q DI
A1 D A
A7.3.6
Q ij D Aji
A1 ij D A
A7.3.7
and for the elements

It follows from eqs. (A7.3.4) and (A7.3.6) that for an orthogonal matrix A
detA D 1
A7.3.8
The columns, and also the rows, of an orthogonal matrix form an orthonormal set of
vectors. That is

Asi Asj D
Aik Ajk D ij
A7.3.9
s
These relationships are the necessary and sufficient conditions that a matrix be orthogonal.
The name adjoint is sometimes used for the matrix which we call conjugate transpose. The reader should
beware of possible confusion when comparing different texts.
376
The Raman Effect
When some or all of the elements of a matrix are complex further types of special
matrices arise and we now consider these.
Complex conjugate matrix. This is represented by A and is formed by taking the
complex conjugate of each complex element of A. Thus A ij D Aij . It should be noted
that in forming the complex conjugate of a product of matrices, the sequence of the
matrices is unaltered. That is
A7.3.10
AB D A B
Conjugate transpose matrix . This is denoted by A and is formed by transposing the
complex conjugate matrix. Thus
Q
A7.3.11
A D A
and for the elements
A ij D AQ ij D A ji
A7.3.12
Unitary matrix. A square matrix A is said to be unitary if its conjugate transpose matrix
A is also its inverse. That is
A7.3.13
A D A1
Hermitian matrix .
property that
A square matrix is said to be hermitian if it possesses the special

A D A
A7.3.14
Such a matrix has Aij D Aji for all i, j and in consequence its diagonal elements are real.
Skew hermitian matrix. A square matrix A is said to be skew hermitian if it possesses
the special property that
A D A
A7.3.15
Such a matrix has Aij D Aji for all i, j and in consequence its diagonal elements are
zero or pure imaginary.
It should be noted that a real symmetric matrix is in fact a special case of an hermitian
Q iBQ .
matrix. Suppose that H D A C iB is hermitian with both A and B real, then H D A
But by definition H D H and hence the real part is symmetric and the imaginary part
is skew symmetric. Thus a real hermitian matrix is also symmetric. Similarly, a real
orthogonal matrix is unitary, for if U D A C iB is unitary, by definition U D U 1 and
Q C iBQ A iB D I . Then since B D BQ D 0 for a real matrix, AA
Q D
U U D I , hence A
I . However, a complex symmetric matrix is not hermitian nor is a complex orthogonal
matrix unitary.
The construction of symmetric and skew-symmetric matrices. Any square matrix with
real elements may be written as the sum of a symmetric matrix and a skew-symmetric
matrix since
Q C 1/2A A
Q
A D 1/2A C A
A7.3.16
Sometimes called adjoint, hermitian conjugate or associate matrix.

Sometimes called self-adjoint.
377
If a square matrix has complex elements it may be written as the sum of a hermitian
matrix and a skew-hermitian matrix since
A D 1/2A C A C 1/2A A
A7.3.17
A7.4 MATRIX REPRESENTATION OF SIMULTANEOUS LINEAR

EQUATIONS
Simultaneous linear equations are very conveniently treated using matrix methods. Let us
consider the following set of simultaneous linear equations
a11 x1 C a12 x2 C . . . aij xj C . . . C a1n xn D b1
a21 x1 C a22 x2 C . . . a2j xj C . . . C a2n xn
..
..
..
..
.
.
.
.
D b2
..
.
A7.4.1
an1 x1 C an2 x2 C . . . anj xj C . . . ann xn D bn

where the aij and the bi are coefficients.
These equations may be written in compact matrix form as follows:
Ax D b
A7.4.2
where A is a square matrix n n of the coefficients aij , x is a column matrix n 1

of the xj , and b is a column matrix n 1 of the coefficients bi .
We may make a formal solution of these equations by premultiplication of both sides
of eq. (A7.4.2) by A1 . We then obtain
x D A1 b
A7.4.3
Thus the equations are solved when A1 is known. We must note that A1 exists only
if det A 6D 0; and that, otherwise, the solution of these equations is the trivial one with
x D 0.
A7.5 EIGENVALUES AND EIGENVECTORS

We now suppose that b D x so that eq. (A7.4.2) becomes
Ax D x
A7.5.1
where , a scalar quantity, is called an eigenvalue of A, and x is an eigenvector. If A

has dimensions n n there are, in general, n eigenvalues and n eigenvectors.
We may then rewrite eq. (A7.5.1) in the form
A I x D 0
This is a case where it is helpful to use distinctive symbols for column matrices.
Sometimes called the characteristic root and the characteristic vector, respectively.
A7.5.2
378
The Raman Effect
This implies n simultaneous equations as follows:

a11 x1 C a12 x2 . . . C a1j xj C . . . a1n xn D 0
a21 x1 C a22 x2 . . . C a2j xj C . . . a2n xn D 0
..
..
..
..
.
.
.
.
A7.5.3
an1 x1 C an2 x2 . . . C anj xj C . . . ann xn D 0

For a non-trivial solution we must have
detA I D 0
A7.5.4
Solution of eq. (A7.5.4) gives the values of the n roots 1 , 2 . . . n . The eigenvectors
can then be determined to within a normalization factor from eq. (A7.5.1). We now
introduce a matrix notation for the set of n eigenvectors and n eigenvalues. We define xj
to be a n 1 column matrix of the components of the eigenvector associated with the
eigenvalue j , where xj has components Nj xij i D 1, 2, . . . n and Nj is a normalization
constant.
We can assemble the n column matrices xj j D 1, 2 . . . n to form a square matrix X
of order n n, and the n eigenvalues to form a diagonal matrix 3 with entries j on
the diagonal. We can then write
A7.5.5
AX D X 3
This matrix equation embraces the n equations of the type given by eq. (A7.5.1). If we
multiply both sides of eq. (A7.5.5) on the left by X 1 (which presumes det X 6D 0) we
obtain
X 1 AX D 3
A7.5.6
Thus the matrix X enables the matrix A to be cast in diagonal form. A transformation
of this general type is called a similarity transformation. If X is complex and unitary,
X 1 D X (where X is the conjugate transpose) then eq. (A7.5.6) becomes
X AX D 3
A7.5.7
and the transformation is called a congruent transformation. If X is real and unitary, that
is real orthogonal, X 1 D XQ (where XQ is the transpose) and then eq. (A7.5.6) becomes
XQ AX D 3
A7.5.8
and the transformation is called a real orthogonal transformation.

The eigenvectors can be normalized to unit length using, as appropriate, the relationships
X X D I
A7.5.9
or
XQ X D I
A7.5.10
379
The quadratic nature of eqs. (A7.5.9) and (A7.5.10) means that we can determine only
the magnitudes of the normalization constants and so a choice of sign must be made for
each normalization constant.
We note, without proof, that a matrix may be diagonalized if: (a) all its eigenvalues
are distinct, or (b) it is hermitian or symmetric, or (c) it is unitary. The process of diagonalization is unique except for the order of writing the j and the associated xj .
A7.6 EXAMPLE OF DIAGONALIZATION OF A MATRIX

We consider as an example the diagonalisation of a matrix R, which is defined as

cos sin
A7.6.1
RD
sin cos
Q Also det R D 1.
This matrix is unitary and real, that is real orthogonal, and so R 1 D R.
We seek first the eigenvalues 1 and 2 of R. These are found from the determinantal
equation

cos
sin

D0
A7.6.2
sin
cos
whose characteristic roots j are given by
1 D ei D cos i sin
i
2 D e D cos C i sin
We seek next the eigenvectors of R using the equations

cos sin
Nj aj
Nj aj
D j
sin cos
Nj bj
Nj bj
A7.6.3
A7.6.4
A7.6.5
where j D 1 or 2 and the j are given by eqs (A7.6.3) and (A7.6.4). From eqs (A7.6.5)
which give only the ratio aj /bj for each j we find that
b1 D ia1
A7.6.6
b2 D ia2
A7.6.7
The normalized eigenvectors have to be determined by using the relationship

X X D I
where the unitary matrix X is given by

N1 a1
X D
N1 b1
N2 a2
N2 b2
A7.6.8
A7.6.9
380
The Raman Effect
After introducing the relationships given by eqs (A7.6.6) and (A7.6.7) we find that
N1 a1 2 D N2 a2 2 D 1/2. A choice of signs now has to
p be made and, forpreasons of
consistency with later applications, we take N1 a1 D 1/ 2 and N2 a2 D 1/ 2. Thus
p
p

1/p 2 1/ p2
X D
A7.6.10
i/ 2 i/ 2
If we now use the unitary matrix X and its conjugate transpose matrix X to perform
a similarity transformation on R we find that
i

e
0
1 0
X RX D
D
A7.6.11
0
ei
0 2
thus confirming that the transformed matrix is diagonal and that its two diagonal terms
are the characteristic roots of R. We also see that
TrfRg D TrfX RX g D 2 cos
A7.6.12
illustrating the important property that the trace of a matrix is invariant under a similarity
transformation. Finally we can confirm that RX D X 3 since
p
p

cos sin
1/p 2
i 1/p 2
De
A7.6.13
sin cos
i/ 2
i/ 2
and

p
p
cos sin
1/ p2
1/ p2
i
De
A7.6.14
sin cos
i/ 2
i/ 2
It will be shown in Chapter A8, Section A8.10, that the matrix R relates to the anticlockwise rotation of a two-dimensional cartesian vector system through an angle . The
diagonalized matrix X RX describes the same rotation using a different coordinate system
whose relationship to the cartesian coordinate system can be found using the matrix X .
Both these matrices involve complex quantities. In this new coordinate system there is
no mixing of the coordinates under the rotation transformation since eq. (A7.6.11) is
diagonal.
Similarity transformations and matrix diagonalization can play important roles in the
rotational transformations of vectors and tensors. In particular they make it possible to
find coordinate systems in terms of which rotational transformations can be expressed
in simpler forms. These matters are considered in some detail in Chapters A8, A9, A10
and A14.
381
A8
Vectors, I
The moral of that is, Take care of
the sense and the sounds will take
care of themselves.
Lewis Carroll
A8.1 INTRODUCTION: SCALARS, VECTORS AND TENSORS

The electric and magnetic phenomena which underlie light scattering will be discussed
in terms of fields of electric charges and currents. A field is a function which describes a
physical function at all points in space.
In scalar fields a physical quantity is completely specified by a single number at each
point. This number, which may be positive or negative, is called a scalar quantity. It may
vary from point to point and there may well be some mathematical relationships between
a scalar quantity and the location of the point with which it is associated. The ensemble
of points, each with a scalar quantity attached to it, forms a scalar field. Examples of
scalar quantities are temperature, density, electric potential and electric charge.
However, in a vector field a number and a direction must be specified at each point
in space. The number and direction together constitute a vector. Examples of vector
quantities are velocity, electric field strength and electric dipole. The ensemble of points
in space, each with its associated vector which may vary from point to point, constitutes
a vector field.
As shown in Chapter A10 a vector can be regarded formally as a tensor of rank 1 and a
scalar as a tensor of rank 0. For the purposes of presentation, however, we have made the
conventional distinction between vectors and tensors and developed the detailed properties
of vectors in Chapters A8 and A9, and those of tensors separately in Chapter A10.
382
The Raman Effect
In this chapter we confine our attention to vectors in three dimensional space which
are represented by cartesian coordinates. Vectors represented by other coordinate systems
in three-dimensional space and also vectors in n-dimensional space are considered in
Chapter A9.
We use bold sloping (i.e. italic) type, for instance A or e, to indicate a vector quantity.
Medium sloping (i.e. italic) type, for instance A or e, is used to indicate either a scalar
quantity or the magnitude of a vector quantity.
A8.2 BASIC DEFINITION OF A VECTOR

A vector A at a point O can be represented graphically by a directed line or arrow whose
length represents the magnitude of the vector on some chosen scale and whose direction
is the direction of the vector. Here, direction includes the sense. This is illustrated in
Fig. A8.1 where the tail of the arrow is located at O and the head of the arrow indicates
that the sense of the direction is from O towards the arrowhead. The length of the line
OP represents the magnitude A of the vector A.
Such a directed line of representative length completely defines a vector. However, the
introduction of a set of axes is necessary to be able to specify a vector in a mathematically
useful manner. If a set of right-handed cartesian axes is located at O, it is evident that
the vector A can be completely specified in relation to these axes by the triad of the
coordinates Ax , Ay and Az of the point P which we may write thus
A
D Ax , Ay , Az
A8.2.1
In such a triad representation, Ax , Ay and Az are sometimes loosely referred to as the

components of A but, as we shall see from eq. (A8.3.6), they are strictly the magnitudes
of the components of A.
z
Unit vector eA
Az
Vector A; |A| = A = OP
lAz
Q
lAy
lAx
O
Ax
x
Figure A8.1 Vectors and unit vectors.
P (Ax, Ay, Az)
Sometimes called the ordered triple of numbers.

We use
D to mean is represented by.
This is not to be read as a row matrix.
Ay
y
Vectors, I
383
The magnitude A or modulus of the vector A is given by

A D jAj D jA2x C A2y C A2z 1/2 j
A8.2.2
as is evident from geometrical considerations.
A8.3 UNIT VECTORS

A unit vector is defined as a vector of unit magnitude along a specified direction. For
example, we may choose eA to be a vector of unit magnitude in the direction of the vector
A so that
A D AeA
A8.3.1
The vector eA is defined in Fig. A8.1 by the directed line of unit length OQ. The magnitudes of its components along the x, y and z axes are simply the direction cosines, lAx ,
lAy and lAz of eA (or the line OQ) with respect to these axes; the normalization property
of these direction cosines, namely
l2Ax C l2Ay C l2Az D 1
A8.3.2
is consistent with the unit length of the line OQ. The triad of these direction cosines
forms a representation of eA as follows
eA
D lAx , lAy , lAz
A8.3.3
Using these direction cosines we can write Ax D AlAx , Ay D AlAy and Az D AlAz . Thus
an alternative triad representation of A is given by
A
D AlAx , AlAy , AlAz
A8.3.4
Of particular usefulness are the unit vectors ex , ey and ez which lie along the positive
x, y and z directions, respectively. These unit vectors can be represented by ordered triads
which give the magnitudes of their components along the x, y and z axes in that order.
That is
ex D 1, 0, 0
ey D 0, 1, 0
and
A8.3.5
ez D 0, 0, 1
Using the unit vectors ex , ey and ez , we can express the vector A as the following linear
combination of vectors:
A8.3.6
A D Ax ex C Ay ey C Az ez
From this equation we see why Ax , Ay and Az are strictly the components of the vector
A relative to the cartesian vectors ex , ey and ez .
384
The Raman Effect
A unit vector is sometimes called a direction vector because its essential function is to
define a direction. An alternative name is a basis vector since, for example, the set of three
unit vectors ex , ey and ez forms a basis for representing all vectors in three dimensional
space. The term basis vector is to be preferred as it can be applied to the general case
of n-dimensional space. Such a space, which is said to be spanned by n basis vectors,
is treated in more detail in Chapter A9. We shall use braces f g to enclose a set of basis
vectors.
The set of basis vectors fex , ey , ez g is called the standard cartesian basis for representation of a vector and its elements, ex , ey and ez are called the cartesian basis vectors. The
cartesian vector components of A, and their magnitudes, may be referred to alternatively
as the vector components of A, and their magnitudes, relative to the basis fex , ey , ez g. With
this choice of basis vectors, eq. (A8.3.6) forms a unique representation of the vector A.
However, the choice of basis vectors is not unique. Other basis vectors may be selected
and may prove more convenient in certain situations. For example, reorientation of the
original cartesian axis system may facilitate the mathematical treatment of a particular
problem. Or, the use of a different set of coordinates, as for example spherical coordinates,
can be helpful in certain cases.
The basis vectors need not be of unit length nor mutually perpendicular but they must
not be co-planar. The test for non-coplanarity of the basis vectors is as follows. Let e1 ,
e2 and e2 be any three vectors in cartesian space. If we form a 3 3 matrix G whose
columns are the cartesian components of e1 , e2 and e3 then, provided det G 6D 0, e1 , e2
and e3 form a basis set, and any vector A may be represented uniquely as
A D a1 e1 C a2 e2 C a3 e3
A8.3.7
where a1 , a2 and a3 are the magnitudes of the components of A relative to the basis set
e1 , e2 and e3 . For example, for the cartesian basis set

1 0 0
A8.3.8
GD 0 1 0
0 0 1
and det G D 1, confirming that fex , ey , ez g is a proper basis set.
Vectors and mathematical operations associated with them can be conveniently represented using matrices. For example, we may represent the cartesian basis vectors by a
row matrix,
[ ex ey ez ]
A8.3.9
and the magnitudes of the components of a vector A relative to this basis by a column
matrix,

Ax
A8.3.10
Ay
Az
Then the rules of matrix multiplication allow the vector to be represented by

Ax
A D [ex ey ez ] Ay
A8.3.11
Az
Vectors, I
385
This matrix representation is entirely equivalent to the unique representation given by

eq. (A8.3.6).
Of course it would have been possible to have chosen a row matrix to represent the
magnitudes of the vector components and a column matrix to represent the basis vectors.
With this choice eq. (A8.3.11) would be replaced by

ex
A8.3.12
A D [Ax Ay Az ] ey
ez
However, we shall adhere throughout to the convention defined in eqs. (A8.3.9) to (A8.3.11).
A8.4 VECTOR ADDITION, SUBTRACTION AND

MULTIPLICATION BY A SCALAR
These algebraic operations for vectors are straightforward. Addition of two vectors A
and B is the sum of their components; and their subtraction is the difference of their
components. Thus
A8.4.1
AB
D Ax Bx , Ay By , Az Bz
or
AB
D [Ax Bx ]ex , [Ay By ]ey , [Az Bz ]ez
A8.4.2
Multiplication by a scalar quantity results in a scaling of the magnitude of the vector

and the magnitudes of its components, but no change in direction. Thus
A D Ax , Ay , Az
A8.4.3
A D Ax ex , Ay ey , Az ez
A8.4.4
or
Equality between two vectors A and B requires the following equalities between components:
A8.4.5
Ax D Bx ; Ay D By ; Az D Bz
A8.5 MULTIPLICATION OF TWO VECTORS

The operation of multiplication of two vectors is more complicated since three types of
multiplication need to be considered.
The first type of vector multiplication is the scalar (or dot, or inner) product of the
vectors A and B. This is represented by A B and is defined by
A B D AB cos
angle
A8.5.1
between the two vectors. The resultant product is a scalar quantity

where is the
and is the ordinary product of the magnitude of A and the magnitude of the projection of
B on A.
The angle between two vectors is defined in Fig. A8.2.
386
The Raman Effect
It follows from this definition that the scalar product is commutative,

AB DB A
A8.5.2
A B C C D A B C A C
A8.5.3
and also distributive,

The associative law has no meaning here because in a product A B C the item A B
is a scalar and cannot form a scalar product with C. However, (A B ) C is meaningful
and implies the multiplication of a vector C by the scalar product (A B ).
It follows from the definition of the scalar product that the following relationships exist
between the unit vectors:
e x ex D e y ey D e z ez D 1
ex ey D e y ez D e z ex D 0
A8.5.4
By writing the vectors A and B in the component form given by eq. (A8.3.6) and using
the distributive law and the above relations we may obtain an alternative expression for
their scalar product:
A B D Ax ex C Ay ey C Az ez Bx ex C By ey C Bz ez
D Ax Bx C Ay By C Az Bz
A8.5.5
A special case is the scalar product of a vector with itself which yields the square of its
magnitude or modulus as defined in eq. (A8.2.2):
A A D A2x C A2y C A2z D jAj2
A8.5.6
The scalar product may also be defined using the matrix representation of two vectors as
Q
A B D AB

Bx
D [Ax , Ay , Az ] By
Bz
D Ax Bx C Ay By C Az Bz
A8.5.7
Q
It should be noted that in eq. (A8.5.7) it is the use of the matrix product in the form AB
and the rules of matrix multiplication that lead to the selection of the appropriate products
of pairs of components consistent with the definition of A B given in eq. (A8.5.5). In
eq. (A8.5.5) the selection of the appropriate pairs of products followed from the properties
of the scalar products of the basis vectors.
The second type of vector multiplication is called the vector product, or cross-product,
and will be represented here by A B . It is defined as follows
C DAB
D AB sin
A8.5.8
Vectors, I
387
where is the angle between the vectors A and B, and C is a vector which is the result
of forming A B . The direction of C, which is defined by the unit vector eC , is given
by the right-hand screw rule. It is the direction along which would advance a right-hand
screw whose axis, perpendicular to the plane of A and B, is rotated in the sense that
causes the first named vector A to rotate towards the second named vector B through the
smaller angle. This is illustrated in Fig. A8.2.
The vector product is not commutative, that is
A B 6D B A
A8.5.9
A B D B A
A8.5.10
However
This follows from the definition of the vector product in eq. (A8.5.7) and the fact that
sin D sin .
The vector product is distributative, however, that is
A B C D D A B C A D
A8.5.11
where D is a third vector.

From the definition of a vector product the following relationships exist involving the
unit vectors
ex ex D 0;
ey ey D 0;
ez ez D 0;
ex ey D ez ;
ey ez D ex ;
ez ex D ey ;
A8.5.12
ey ex D ez ; ez ey D ex ; ex ez D ey
If the vector product A B is written out in terms of magnitudes and components and
the relationships given in eq. (A8.5.12) are invoked, the magnitudes and components Cx ,
Cy , Cz of the vector C may be obtained. Thus
C D A B D Ax ex C Ay ey C Az ez Bx ex C By ey C Bz ez
D Ay Bz Az By ex C Az Bx Ax Bz ey
A8.5.13
C Ax By Ay Bx ez
z
C
B
q
A
x
Figure A8.2 Vector cross-product.
A vector product can also be formulated using the Levi-Cività alternating tensor. See Chapter A10, p. 427,
eqs. (A10.4.13) and (A10.4.14).
388
The Raman Effect
and hence,
Cx D Ay Bz Az By
Cy D Az Bx Ax Bz
A8.5.14
Cz D Ax By Ay Bx
Equation (A8.5.13) may be written in the form of a determinant as follows

ex

C D A B D Ax

Bx
ey
Ay
By

ez

Az

Bz
A8.5.15
The vector product has many useful applications. For example, we see from Fig. A8.3
that the area of a parallelogram whose sides are defined by the co-terminal vectors A
and B can be represented by a vector S D A B , which is perpendicular to the plane
of the parallelogram. The magnitude of S is given by AB sin which is the area of the
parallelogram.
The vector product is particularly useful when dealing with angular momentum. The
angular momentum L of a particle mass m about a point O is defined by the vector
product r p. Here r is the vector drawn from the point O to the particle and p is
its linear angular momentum vector which is given by p D mv, where v is its linear
velocity vector, or alternatively by p D mr where is its angular velocity and v D r.
Its magnitude is rp sin , where is the angle between the vectors r and p. The vector L
is perpendicular to the plane containing r and p and is directed in the direction travelled
by a right-hand screw turned in the sense of the rotation. By convention it is directed
outwards from O.
z
y
A
Figure A8.3 Vector cross-product and area of a parallelogram.
Vectors, I
389
Using eq. (A8.5.15) we may write

LDr p

ex ey ez

y
z
D x

px py pz
D ypz zpy ex C zpx xpz ey C xpy ypz ez
A8.5.16
where x, y and z refer to a right-handed cartesian axis system with its origin at O. Since
we can write the angular momentum in component form as
L D L x ex C L y ey C L z ez
A8.5.17
eq. (A8.5.16) enables the components Lx , Ly and Lz to be identified as

Lx D ypz zpy
Ly D zpx xpz
A8.5.18
Lz D xpy ypx
The case when Lx D Ly D 0 is illustrated in Fig. A8.4.
y axis
r
px
py
Particle of mass m
coordinates x and y
x axis
Figure A8.4 The angular momentum of a particle of mass m rotating about the point O
in the xy plane. Lz D r p D Lz ez with Lz D xpy ypx . Lz is directed out of the plane of
the paper from O.
The magnitude of the angular momentum in classical physics is L the length of the
vector L, and is given by
L D jLx2 C Ly2 C Lz2 1/2 j
A8.5.19
Another example (Fig. A8.5) relates to the torque produced by a force F acting with a
moment arm r about a point O. The torque T is given by the vector product r F which
may be written as r pP since F D dp/dt D p.
P Thus the detailed development just given
for angular momentum can be carried over to this case. The torque T acts perpendicular
390
The Raman Effect

z
F
q
P
y
0
Figure A8.5 Torque and vector cross-product.
to the plane defined by the vectors r and F and along the axis of rotation and in the
direction travelled by a right-hand screw turning in the sense of the force. The magnitude
of the torque is rF sin .
The third kind of multiplication of two vectors A and B is the direct product which is
represented by AB. It is defined by writing each vector in terms of its components using
eq. (A8.3.6) and forming the nine possible products of pairs of components by ordinary
multiplication. Thus
AB D ex Ax C ey Ay C ez Az ex Bx C ey By C ez Bz
D Ax Bx ex ex C Ax By ex ey C Ax Bz ex ez C Ay Bx ey ex C Ay By ey ey
C Ay Bz ey ez C Az Bx ez ex C Az By ez ey C Az Bz ez ez
A8.5.20
It should be noted that the unit vector terms are direct or ordinary products and are
not scalar or vector products, so that ex ex 6D 1, ex ey 6D 0 etc. Each of the nine terms in
eq. (A8.5.20) consists of a pair of unit vectors, sometimes called unit vector dyads, and
a pair of vector components, sometimes called vector dyads.
It is interesting to note that the scalar product of two vectors involves three product
pairs of the type Ax Bx , the vector product the six mixed product pairs of the type Ax By ,
and the direct product all nine possible product pairs of the components of A and B.
The direct product AB finds application in the representation of a second-order tensor.
Tensors are essential mathematical tools for the treatment of light-scattering phenomena
and are treated in detail in Chapter A10.
A8.6 TRIPLE PRODUCTS OF VECTORS

For three arbitrary vectors A, B and C a variety of products may be conceived. Only
four of these are of interest here. We shall merely give their properties without proof
Sometimes called the outer product or dyad.

A sum of dyad terms is sometimes called a dyadic. Thus AB is a dyad which can be expressed as a
dyadic.
Vectors, I
391
and draw attention to their significance and usefulness. The proofs are straightforward
and involve writing the vectors out in terms of their coordinates and performing the
appropriate multiplications.
A8.6.1 AB C
Since B C is the scalar BC cos the triple product AB C is simply a new vector
whose direction is the same as that of A and whose magnitude is ABC cos .
A8.6.2 A B C
This is called the scalar triple product and is clearly a scalar. The order of writing the
factors is immaterial as long as the cyclic order is preserved, and the dot and the cross
may be interchanged without altering the value. There are thus six ways of writing this
product:
A B C D B C A D C A B D B C A D C A B D A B C
A8.6.1
If the cyclic order is not maintained there will be a change of sign since B C D
C B . Thus, for example
C A B D C B A
A8.6.2
Because of the properties of the unit vectors, given in eq. (A8.5.12) the scalar triple
products of all unit vectors vanish except the following.
ex ey ez D ex ez ey D ey ez ex D ez ey ex D
ey ez ex D ey ex ez D ez ex ey D ex ez ey D
A8.6.3
ez ex ey D ez ey ex D ex ey ez D ey ex ez D 1

Note that in the positive terms the unit vectors occur in cyclic order, whereas in the
negative terms they are in non-cyclic order.
If the vectors A, B and C are written in terms of their components and the appropriate
multiplications performed it can be shown that
A B C D Ax By Cz C Bx Cy Az C Cx Ay Bz
Ax Cy Bz Bx Ay Cz Cx By Az

Ax Ay Az

D Bx By Bz

Cx Cy Cz
A8.6.4
The geometrical significance of the scalar triple product is readily seen. As the modulus
of B C is a measure of the area formed by line segments represented by B and C, it
follows that if the angle between A and B C is acute then A B C is the measure
of the volume of the parallelepiped formed by line segments A, B and C. If the angle is
obtuse then this volume is given by A B C .
392
The Raman Effect
A8.6.3 A B C )
This is called the triple vector product. The result of forming this product is a vector
which we shall denote by V, so that
V D A B C
A8.6.5
The properties of the vector product require that V is perpendicular to both A and to
B C , thus
V AD0
A8.6.6
V B C D 0
A8.6.7
V must be in the plane of B and C as it is perpendicular to the vector product of B and

C, which is itself perpendicular to B and C.
The most important property of the vector triple product is that it permits decomposition
into two scalar products, through the relationship
A B C D B A C C A B
A8.6.8
As the vector product changes sign when the order of multiplication is changed, the sign
of the triple vector product will change when the order of the factors in parentheses is
changed or when the position of the parentheses is changed. Thus,
A B C D A C B D C B A D B C A
A8.6.9
A8.6.4 ABC
This is called the direct product of three vectors ABC. It yields a linear combination of
27 terms each of the type ex ey ez Ax Bx Cx when the three vectors are written in terms of
their components and the products formed by ordinary multiplication:
ABC D ex Ax C ey Ay C ez Az ex Bx C ey By C ez Bz ex Cx C ey Cy C ez Cz
D Ax Bx Cx ex ex ex C Ax By Cy ex ey ey C Ax Bz Cz ex ez ez
C Ax Bx Cy ex ex ey C Ax By Cx ex ey ex C Ax By Cz ex ey ez
C Ax Bz Cy ex ez ey C Ax Bz Cx ex ez ex C Ax Bx Cz ex ex ez
C Ay Bx Cx ey ex ex C Ay By Cy ey ey ey C Ay Bz Cz ey ez ez
C Ay Bx Cy ey ex ey C Ay By Cx ey ey ex C Ay By Cz ey ey ez
C Ay Bz Cy ey ez ey C Ay Bz Cx ey ez ex C Ay Bx Cz ey ex ez
C Az Bx Cx ez ex ex C Az By Cy ez ey ey C Az Bz Cz ez ez ez
C Az Bx Cy ez ex ey C Az By Cx ez ey ex C Az By Cz ez ey ez
C Az Bz Cy ez ez ey C Az Bz Cx ez ez ex C Az Bx Cz ez ex ez
A8.6.10
This direct product finds application in the representation of a third order tensor. This is
treated later in Chapter A10.
Vectors, I
393
Products of more than three vectors may always be reduced to one of the preceding
types of triple products by successive application of the above rules. An example is
A B C D D A C B D A DB C
A8.6.11
A8.7 FORMAL DEFINITION OF A VECTOR IN TERMS OF ITS

TRANSFORMATION UPON ROTATION OF AXES
We may use direction cosines to express the components Ax0 , Ay 0 and Az0 of a fixed vector
A in a new coordinate system x 0 , y 0 , z0 obtained by rotation of the original coordinate
system x, y, z in which the vector A has components Ax , Ay and Az . By resolving each
of the components Ax , Ay and Az along the axes x 0 , y 0 and z0 in turn we find that
Ax0 D lx0 x Ax C lx0 y Ay C lx0 z Az
Ay 0 D ly 0 x Ax C ly 0 y Ay C ly 0 z Az
A8.7.1
Az0 D lz0 x Ax C lz0 y Ay C lz0 z Az

or, in compact matrix notation,
A0 D c DA
where

c
DD
lx0 x
ly 0 x
lz0 x
lx0 y
ly 0 y
lz0 y
A8.7.2
lx0 z
ly 0 z
lz0 z
A8.7.3
and A and A0 are column matrices of the components of A and A0 .

We may say that the components of the vector A undergo a linear transformation
upon rotation of the coordinate axes. The direction vectors ex , ey and ez also undergo a
related transformation. However, the vector A itself, having a given magnitude and being
associated with a fixed direction in space, is invariant under the rotation of the axes.
If we introduce the summation convention and use the subscripts 0 , 0 , 0 for the primed
axis system and the subscripts , , for the unprimed axis system, the transformation
can be written
A8.7.4
A0 D l0 A
and the corresponding inverse transformation as
A D l0 A0 D l0 A0
A8.7.5
where l0 D l0 .

The invariance of the magnitude of A under the rotation of the axis system follows
from the above equations and the orthonormal relationships of the direction cosines.
Equation (A8.7.4) may be used to give a definition of a vector in terms of its transformation properties. If three quantities A , A and A transform upon rotation of the axes
according to the rule given by eq. (A8.7.4) they are said to be components of a vector
Chapter A.5, Section A5.4 deals with direction cosine matrices.
394
The Raman Effect
with respect to those axes. The transformed quantities A0 , A 0 and A 0 are the components
of the same vector with respect to the primed axes.
In Section A8.2 a vector was defined as a quantity possessing magnitude and direction,
properties which are intuitively comprehended. However, the purely analytical definition
of a vector given by eq. (A8.7.4) is that which is most generally applicable. It leads
naturally to the most general definition of a tensor (see Chapter A10) and may be readily
extended to n-dimensional space to which physical pictures of direction and magnitude
cannot be applied.
It should be noted that in this book when a quantity is defined in terms of magnitude
and direction and called a vector it is to be understood that it is a proper vector, that is,
it transforms according to the cosine law of eq. (A8.7.4).
A8.8 POLAR AND AXIAL VECTORS: TIME EVEN AND TIME

ODD VECTORS
This is an appropriate point at which to introduce a further aspect of vector classification.
A vector which is changed in sign by the space inversion operator is called a polar or
true vector and in this section it will be denoted by the symbol P . A position vector r is
an example of a polar vector, as Fig. A8.6 shows. If the sign of a vector is not changed
by the space-fixed inversion operator E it is called an axial or pseudo vector and in this
section it will be denoted by the symbol A. The angular momentum vector L D r p
is an example of an axial vector. As the polar vectors r and p both change sign under
inversion their vector product, L does not. In other words L is defined relative to the sense
of rotation by a right-hand rule and the operation does not change the sense of rotation
(see Fig. A8.7).
We may illustrate these transformation laws by considering how an inversion of the
coordinate axes affects polar and axial vectors. For inversion of the axes each direction
cosine is l0 D 1 since the angle between each pair of axes is 180 . Then, applying
0
eq. (A8.7.4) to a polar vector, the new components P of a polar vector P are related to
the original components P by
P0 D l0 D P
A8.8.1
Inversion
r
Figure A8.6 Polar vector changed in sign on inversion.
So far, we have used A, B and C to represent general vectors. Henceforth it will prove convenient in
certain situations to use other symbols to represent vectors.
Vectors, I
395
L=rp
L= (r) (p) = L
Inversion
Figure A8.7 Axial vector unchanged in sign on inversion.

r = r
r
T
(a)
= dr
dt
dr =
d (t)
(b)
L = r (p) = L
L=rp
T
(c)
Figure A8.8 Effect of time reversal operator upon (a) position vector r, (b) velocity
vector v and (c) angular momentum vector L.
However, for an axial vector

A0 D 1l0 A D A
A8.8.2
A vector whose sign is not changed by the time reversal operator T is called time even,
and a vector whose sign is reversed by T is called time odd. A position vector is time
even, but the velocity and angular momentum vectors are time odd, since velocity and
angular momentum are linear functions of time. Figure A8.8 illustrates the effect of T on
various vectors.
A8.9 VECTOR DIFFERENTIATION

A8.9.1 The Operator
We first introduce an operator which is represented by the symbol (called del) and
is defined as
C ez
A8.9.1
D e x C ey
x
y
z
We now consider the quantities where is either a scalar or a vector and is either
to be omitted or replaced by a dot or a cross to give products which are meaningful. The
396
The Raman Effect
operator is not a vector in the geometrical sense as it has no scalar magnitude, but
it does transform according to the cosine law and may therefore be treated formally as
a vector. The possible meaningful products are: where is a scalar point function;
and V and V , with V defined as representing a vector field. We now treat each
of these products in turn.
A8.9.2 The Gradient

The product which is called the gradient (often abbreviated as grad ) of the scalar
point function is given by

C ey
C ez
A8.9.2
D e x
x
y
z
It is itself a vector since it is the product of a scalar and a vector ; and its components
are the rate of change of with distance along the coordinate axes x, y, z. To explain the
significance of the gradient we consider a family of surfaces x, y, z D constant. At
some point P with coordinates x, y, z on one of these surfaces is perpendicular to that
surface of the family which passes through P and its direction is that in which is increasing.
Now, if the scalar function is a continuous and differentiable function of the coordinates
and has a value at a particular point, then d , the change in the scalar function over a
distance dl in a direction defined by the unit vector el measured from that point, is
d D el dl
A8.9.3
Clearly d has its maximum value when el has the same direction as r , that is when
, the angle between them, is zero. Thus the gradient of is a vector whose magnitude
and direction give the maximum space rate of change of . Figure A8.9 illustrates this.
In terms of its components, the magnitude of is

2 2 2 1/2

j j D
C
C
A8.9.4

x
y
z

To summarize, the gradient of a scalar function is a vector with the following properties.
Its components at any point are the rates of change of the function along the directions
Q
f + df
f
el
q
90
P
f
Figure A8.9 The gradient of the scalar function .
Vectors, I
397
of the coordinate axes at that point. Its magnitude at the point is the maximum rate of
change of the function with distance. Its direction is that of the maximum rate of change
of the function, and it points toward larger values of the function.
The gradient of a scalar function finds application, for example, in electrostatics as
discussed in Chapter A11.
A8.9.3 The Divergence

The scalar product V which is called the divergence (often abbreviated as div ) of a
vector V is given by

C ez
V D ex C ey
fex Vx C ey Vy C ez Vz g
x
y
z
Vy
Vz
Vx
C
C
.
A8.9.5
D
x
y
z
and is itself a scalar point function.
One useful application of the divergence is in the calculation of outgoing flux. Suppose
for example that V represents at each point in space the direction and magnitude of flow
of some fluid such as a liquid or gas. If the density of the fluid is and it is moving
with velocity v then V D v. V is called the flux density and represents the total flow of
fluid per unit cross section in unit time. It can be shown that the total loss per unit time
(or outgoing flux) from a small parallelepiped of volume d is given by V d. Thus
V is the outgoing flux per unit volume.
The total outward flux is also equal to the surface integral of the normal outward
component of V. Thus we may write
V dS D V d
A8.9.6
S
where dS is an element of the total surface area and is the total volume enclosed by
S. This relationship is known as the divergence theorem. Note that in eq. (A8.9.6) the
left hand side involves only the values of V on the surface S whereas the right hand side
involves the values of V throughout the volume enclosed by S.
A8.9.4 The Curl, V

The product V which is called the curl of V (or the rotation of V, often abbreviated
as rot V) is given by

Vz Vy
Vx Vz
Vy
Vx

V D ex
C ey
C ez
A8.9.7
y
z
z
x
x
y

ex ey ez

D

x y z

Vx Vy Vz
and is itself a vector function. The curl finds application for example in electromagnetic
theory.
398
The Raman Effect
A8.9.5 The Divergence and the Curl Illustrated and Compared

We now discuss two examples which illustrate the nature of the functions div and curl .
First we consider the vector function V which is defined as
V D yex C xey
A8.9.8
The vector function is shown in Fig. A8.10(a) where it is represented graphically by a

series of contours of equal magnitudes of V and scaled arrows whose length and direction
denote the magnitude and direction, respectively, of V at the origin of each arrow. It can
be seen that the function V represents a circulation of some kind about the z axis, which
is perpendicular to the plane defined by ex and ey .
Secondly we consider the vector function V 0 , which is defined as
V 0 D xex C yey
A8.9.9
This function is shown in Fig. A8.10(b) where it is represented by a series of contours

and scaled arrows whose functions are the same as those described above. The function
V can be seen to represent a radial flow away from a central point.
If we now evaluate the divergences of V and V 0 we find that V D 0 but V 0 D 2.
These values reflect the natures of V and V 0 as illustrated in Figs. A8.10(a) and A8.10(b):
V does not diverge, but circulates around the centre point, whereas V 0 does diverge.
The curls of V and V 0 are similarly descriptive. For the curl of V we have

ey ez
ex

V D
A8.9.10
D 2ez
x y z

y x
0
This result indicates that V curls around ez , with the sense of the rotation defined
according to the right-hand screw rule. For the curl of V 0 we have V 0 D 0 indicating
that V 0 does not curl (it diverges as already shown).
Finally, we consider briefly a vector function A which is defined by
A D bV C cV 0
A8.9.11
where V and V 0 have the forms given by eqs (A8.9.8) and (A8.9.9), respectively, and b
and c are scalar quantities. For this composite vector both the divergence and the curl are
non-zero as is evident from Fig. A8.10(c) which illustrates the vector function A using
the same conventions as for Figs. A8.10(a) and A8.10(b).
A8.9.6 Composite Functions Involving

We give below a number of useful relationships involving
A C B D A C B
A D A C A
A8.9.12
A8.9.13
Vectors, I
399
y
(a)
(b)
(c)
Figure A8.10 (a) Vector function V with V D 0 and V D 2ez . (b) Vector function
V 0 with V 0 D 0 and V 0 D 2. (c) Vector function A D bV C cV 0 with A 6D 0
and A 6D 0.
400
The Raman Effect

U V D V U C U V C V U C U V A8.9.14
U V D V U U V
A8.9.15
U V D V U V U U V C U V
A8.9.16
In these equations A and B are either scalars or vectors depending on the choice of (*),
is a scalar, and U and V are vectors. These relationships may be verified by expanding
the vectors in terms of their components and the unit vectors ex , ey and ez .
A8.9.7 Successive Applications of

There are six possible combinations in which occurs twice. The resulting relationships
can all be proved by expansion of the vector(s) in terms of their components and the unit
vectors ex , ey and ez . The results will now be summarized.
(a)
D r 2 D
2 2 2
C 2 C 2
x 2
y
z
A8.9.17
It should be noted that the operator r2 is a scalar. It is usually called the Laplacian
and is a familiar operator in quantum mechanics.
(b)
2 V
2 V
2 V
C
C
x 2
y 2
z2
2
2
2
Vx
Vy
Vz
D ex 2 C ey
C ez 2
2
x
y
z
V D r2 V D
A8.9.18
The result of applying the scalar operator r2 to a vector is of course a vector.

2 Vx
2 Vy
2 Vz
V D e x 2 C e y
C
e
z
x
y 2
z2

2 Vy
2 Vx
2 Vx
2 Vz
2 Vz
2 Vy
C ex
C ey
C ez
A8.9.19
C
C
C
xy
xz
xy
yz
xz
yz

ey
ez
ex

D x y z D 0
A8.9.20
(d)

x y z
This is an identity. If for some vector V, V D 0, then V D where is some
scalar function. Under these conditions V is said to be irrotational.
(c)
(e)
V D0
A8.9.21
This is another identity. It follows that, if for any vector W, W D 0, then

W D V and W is said to be solenoidal.
Vectors, I
(f)
401
V D V V
D V r 2 V
A8.9.22
A8.9.8 Time Derivative of a Vector

The time derivative of a vector is equal to the vector sum of the time derivatives of its
components, that is
dVx
dVy
dVz
dV
D ex
C ey
C ez
A8.9.23
dt
dt
dt
dt
A8.9.9 Caveat
The definitions given throughout Section 8.9 have been based on the cartesian coordinate
system. The definitions for other coordinate systems are more complicated. Some are given
without discussion in Chapter A9. It is important to note that the Laplacian operator must
not be applied to the components of a vector except in cartesian coordinates. Thus, for
example, the identity given by eq. (A8.9.22) applies only in cartesian coordinates.
A8.10 CHANGE OF BASIS VECTORS AND EFFECT UPON

COORDINATES OF A FIXED VECTOR
In Section A8.7 above we considered how the components of a fixed vector in a new
coordinate system x 0 , y 0 , z0 obtained by rotation of the original coordinate system x, y, z
are related to its components in that original coordinate system.
We now consider how the basis vectors of such a fixed vector are transformed by rotation of the coordinate system and how this transformation is related to the transformation
of the coordinates of the fixed vector.
Let us consider the case of a vector V, fixed in space, with coordinates Vax , Vay , Vaz
in a cartesian axis system Sa whose basis vectors are exa , eya and eza . If we represent the
column matrix of the coordinates by V a and the row matrix of the basis vectors by B a ,
then using compact matrix notation we may write
V D B aV a
A8.10.1
We now introduce a second cartesian axis system Sb with basis vectors exb , eyb and ezb
which we represent by the row matrix B b . We define B b to be related to B a by the
transformation A so that
B b D B aA
A8.10.2
In the axis system Sb the fixed vector V has a different set of coordinates Vbx , Vby , Vbz
which we represent by the column matrix V b . We may then write
V D B bV b
A8.10.3
402
The Raman Effect
This redefinition of the original vector V in terms of new basis vectors exb , eyb and ezb
may be regarded as giving it an alternative description and it is sometimes called an alias
operation.
We now seek the relationship between V b and V a which we define by the matrix C
such that
V b D CV a
A8.10.4
If we substitute eq. (A8.10.2) into eq. (A8.10.3) we obtain
V D B a AV b
A8.10.5
Then comparing eq. (A8.10.5) with eq. (A8.10.1) we see that

V a D AV b
A8.10.6
Premultiplying both sides of eq. (A8.10.6) by A1 we obtain for the vector coordinates
V b D A1 V a
A8.10.7
C D A1
A8.10.8
from which it follows that

To make a direct comparison between eq. (A8.10.2) and eq. (A8.10.7) we transpose both
sides of the latter and obtain for the vector coordinates
Q 1
VQ b D VQ a A
A8.10.9
which has the same form as eq. (A8.10.2) for the basis vectors.
The relationship between the associated changes of basis vectors and vector coordinates
appears in general to be a complicated one. However, when A is a complex and unitary
Q 1 D A . Thus CQ D A and
matrix, A1 D A and hence A
VQ b D VQ a A
A8.10.10
Q and hence A
Q 1 D A. Thus
Further, if A is a real and unitary matrix A1 D A
VQ b D VQ a A
A8.10.11
and the coordinates of a fixed vector then change in exactly the same way as the basis
vectors.
We now illustrate these general results by considering the special case where the
vector V is a fixed vector in two-dimensional space and has coordinates Vax , Vay in the
two-dimensional cartesian axis system Sa , whose basis vectors are ex and ey . Thus the
appropriate expanded form of eq. (A8.10.1) is
a
Vx
a
a
V D [ex ey ]
A8.10.12
Vay
Vectors, I
403
We now introduce a second two-dimensional cartesian axis system Sb with basis vectors
exb and eyb . Thus the appropriate expanded form of eq. (A8.10.3) is
b
Vx
b b
V D [ex ey ]
A8.10.13
Vby
where Vbx and Vby are the coordinates of V in the axis system Sb .
We define the axis system Sb as being obtained from the system Sa by a rotation
through the angle in the direction shown in Fig. A8.11. Using this figure it is readily
shown from geometrical considerations or the use of direction cosines that
exb D cos exa C sin eya
eyb
sin exb
A8.10.14
cos eya
Hence, recalling that in eq. (A8.10.2) the basis vectors are represented by row matrices,
the appropriate form of the transformation matrix A is

cos sin
AD
A8.10.15
sin
cos
We now consider the relationship between Vax and Vay , the coordinates of the fixed vector
V in the axis system Sa , and Vbx and Vby , its coordinates in the axis system Sb . It can be
seen from Fig. A8.11 that
Vax D V cos ;
Vbx D V cos ;
Vay D V sin
Vby D V sin
A8.10.16
where V D jV j. Thus it follows that

Vbx D cos Vax C sin Vay
Vby
sin Vax
A8.10.17
cos Vay
y(of Sa)
y(of Sb)
V
x(of Sb)
eay
eby
aq
q
ebx
eax
a
x(of Sa)
Figure A8.11 Rotation of axis system Sa through angle to give Sb .
If the page of the paper is regarded as the face of an analogue clock, this rotation is anti-clockwise and
is so described in many texts. However, using the conventions described in Chapter A1 this is a clockwise
motion when looking along the positive z axis or, equivalently, from the back of the clock face. To avoid
ambiguity we have defined using Fig. A8.11.
404
The Raman Effect
and hence the matrix C in eq. (A8.10.4) where the vector coordinates are written as
column matrices is given by

cos sin
A8.10.18
C D
sin cos
Q 1 D A
Now A and C are real unitary matrices so we would expect that C D A1 and CQ D A
and this is indeed the case. Thus for this special case of a real unitary transformation the basis
vectors and the coordinates of the fixed vector transform in exactly the same way, when both
are represented as row matrices as explained above.
A8.11 THE EFFECT OF A SYMMETRY OPERATION ON

VECTORS AND BASIS VECTORS
We now consider the effect of a symmetry operation R on a vector V which has coordinates
Vax , Vay , Vaz in the cartesian axis system Sa . Under the symmetry operation R the vector V
is transformed into another vector W. The vector W has the same magnitude as V, that is
jW j D jV j but different coordinates, namely Wax , Way , Waz in the untransformed system
Sa . We can express the effect of the operation R by the matrix equation
a
a
Wx
Rxx Rxy Rxz
Vx
Way D Ryx Ryy Ryz Vay
A8.11.1
a
a
Wz
Rzx Rzy Rzz
Vz
which we may write in compact matrix notation as
W a D RV a
A8.11.2
where V a and W a are column matrices of the components of V and W, respectively, in

the axis system Sa , and R is the matrix representing the operation R. The operation R in
effect relocates the vector V and is sometimes called an alibi operation.
We now discuss the situation when the set of basis vectors (a) for the axis system
Sa are transformed under the operation R to give a new set of basis vectors (b) for the
transformed axis system Sb . We define these two sets of basis vectors by the row matrices
B a and B b . It follows that the coordinates of the vector W in the basis b are the same as
those of V in the basis a. Thus we may write
W D B bV a
A8.11.3
However using the set (a) of basis vectors the vector W is given by
W D B aW a
A8.11.4
Substituting eq. (A8.11.2) in eq. (A8.11.4) we obtain

W D B a RV a
A8.11.5
Vectors, I
405
On comparing eqs. (A8.11.3) and (A8.11.5) we see that

B b D B aR
A8.11.6
Thus, when the basis vectors are written as row matrices the matrix R representing their
transformation under the operation R is the same as that representing the transformation of
the coordinates of a vector under the same operation, with the vector coordinates written
as column matrices.
We now illustrate these concepts by considering as an example a vector V in a cartesian
axis system Sa subjected to a rotation through an angle about the z axis as defined in
Fig. A8.12 to give a transformed vector W of the same magnitude. Clearly Waz D Vaz and
jW j D jV j, but Wax and Way are different from Vax and Vay . Using Fig. A8.12 in which the
vector V makes an angle with the x axis, we can write
Wax D jV j cos C D jV j cos cos jV j sin sin
Way D jV j sin C D jV j cos sin C jV j sin cos
A8.11.7
It is evident from Fig. A8.12 that

Vax D jV j cos
Vay
and so it follows that
A8.11.8
D jV j sin
Wax D Vax cos Vay sin
A8.11.9
Way D Vax sin C Vay cos
Thus, in the matrix equation (A8.11.2), for the example under consideration,

cos sin 0
R D sin
cos 0
A8.11.10
0
0
1
Q
We note that R is a real unitary matrix so that R 1 D R.
y(of Sb)
Sa
Sb
y(of Sa) W
|W|=|V|
a a
with ex , ey
b b
with ex , ey
V
x(of Sb)
f
a
af
f
x(of Sa)
Figure A8.12 Rotation of a vector V through an angle to give W.
See footnote on page 403.
406
The Raman Effect
We now consider the situation when the basis vectors transform with the vector so that,
relative to the transformed basis, the coordinates of the transformed vector are unchanged.
For the operation in question which is a rotation through about the z axis, as defined
in Fig. A8.12 ezb D eza , but exb and eyb are different from exa and eya , respectively. Adapting
eq. (A8.10.14) by replacing by we have
exb D cos exa C sin eya
eyb D sin exa C cos eya
Thus in eq. (A8.11.6), where the basis vectors are written as row matrices,

cos sin 0
cos 0
R D sin
0
0
1
A8.11.11
A8.11.12
which, as required, is identical with eq. (A8.11.10).

Transformation matrices can take on simple forms when the basis vectors are chosen
appropriately. A case of special importance is the diagonalization of a transformation
matrix. As shown in Chapter A7, Sections A7.5 and A7.6, matrix theory provides a
procedure for finding transformations which achieve diagonalization of rotational matrices.
However, complex quantities are involved and so the discussion is deferred to Chapter A9.
407
A9
Vectors, II
There is no excellent beauty that hath
not some strangeness in the proportion.
Francis Bacon
A9.1 INTRODUCTION
In this chapter we consider two further aspects of vectors: the definition of vectors in
three-dimensional space in terms of cylindrical, polar and spherical coordinates; and the
extension of the vector concept to n-dimensional space. When it is not clear from the
context or the subscripts which coordinate system is being used, defining labels will be
added to the vector symbols. For example, c V for cartesian coordinates, s V for spherical
coordinates and so on.
A9.2 CYLINDRICAL COORDINATES AND BASIS VECTORS

The relationships between the cylindrical coordinates , and z and the cartesian coordinates of a point P are given by eqs. (A1.4.1) and (A1.4.2). At the point P, unit vectors ei
ej and ek can be defined as shown in Fig. A9.1. The unit vector ei is in the direction of
the perpendicular from the z axis extended through P; the unit vector ej is perpendicular
to the plane containing the z axis and P in the direction corresponding to increasing ;
and the unit vector ek is in the positive z direction, that is it is equal to ez . These unit
vectors do not maintain the same directions in space as the point P moves about but they
always remain mutually orthogonal.
For some purposes it can be useful to introduce the vector r defined by
r D ei C zek
A9.2.1
408
The Raman Effect

z
ek
r df
r
ej
dz
P
z
dr
df
ei
f
x
Figure A9.1 Cylindrical coordinates , and z, and unit vectors ei , ej and ek .
with the ej component zero. The vector r (not shown in Fig. A9.1) only determines the
position of the point P in the plane defined by ei and ek since eq. (A9.2.1) does not
involve the angle .
A9.3 POLAR COORDINATES AND POLAR BASIS VECTORS

The relationships between the polar coordinates r, and and the cartesian coordinates
of a point are given by eqs. (A1.5.1) and (A1.5.2). At the point P, unit vectors ei , ej
and ek can be defined as shown in Fig. A9.2. The unit vector ei is in the direction of
the radius vector r extended through P; the unit vector ej is perpendicular to the radius
vector in the plane containing the z axis and the radius vector r; and the unit vector ek
is perpendicular to that plane. It is important to appreciate that these unit vectors also do
not maintain the same directions in space as the point P moves, but they always remain
mutually orthogonal.
The polar basis unit vectors find application in formulae dealing with emission of
radiation by dipole and other sources, as for example in Chapter A20.
z
ei
ek
P(r, q, f)
r
q
y
O
f
ej
Figure A9.2 Polar coordinates r, and , and unit vectors ej , ej and ek .
Vectors, II
409
A9.4 SPHERICAL COMPONENTS AND SPHERICAL BASIS

VECTORS AND DIRECTION COSINES
Spherical coordinates s1 , s0 and s1 were introduced in Chapter A6 and defined in relation
to cartesian coordinates by eqs. (A6.6.1) to (A6.6.3). We now consider the spherical
components V1 , V0 and V1 of a vector V which are defined in terms of the cartesian
components of V as follows:
V1 D p1 Vx C iVy
2
V0 D Vz
V1 D
A9.4.1
p1 Vx
2
iVy
We shall use the subscripts , and to label spherical components of a vector where
, and can each be 1, 0 or 1. We reserve the subscripts , and
for labelling
cartesian components of a vector where , and
can each be x, y or z. For general
descriptive purposes we shall use the terms spherical system and cartesian system.
The definition of the V in terms of the V given in eq. (A9.4.1) is not unique because
a particular choice of phase factor has been made. The phase factor used here is the same
as that first introduced by Condon and Shortley (1935) and is probably now the preferred
choice. However a number of authors have used a different phase factor which results
in an alternative set of spherical components which we denote by V01 , V0 and V01 with
V01 D V1 but V00 D V0 and V01 D V1 . It is therefore vital to ascertain the phase factor
(or definitions of spherical components) used by other authors before comparing their
results with those given here.
We now introduce unit vectors e1 , e0 and e1 in the spherical system. Their relationships
to the unit vectors, ex , ey and ez in the cartesian system follow from eq. (A9.4.1) and are
given by
e1 D
1
p
e
2 x
C iey
A9.4.2
e0 D e z
e1 D
p1 ex
2
iey
Standard texts which use this phase factor include Placzek and Teller (1933), Bhagavantam (1942) and
Weber (1973).
Fano and Racah (1959) have introduced two further alternative definitions which are as follows:
Standard set
VFR
1
p1 iVx
2
Contrastandard set
0
C Vy
p1 iVx
2
p1 iVx
2
C Vy
VFR
0 D iVz
VFR
1 D
VFR
D
1
VFR
D iVz
0
0
C Vy
VFR
1 D
p1 iVx
2
C Vy
The contrastandard set is simply the complex conjugate of the standard set.
410
The Raman Effect
The inverse relationships are

ex D p12 e1 e1
ey D pi e1 C e1
2
A9.4.3
ez D e0
Spherical vector components and basis vectors are in general complex and the following
types of relationships involving their complex conjugates exist.
V D 1 V
A9.4.4
1 e
A9.4.5
e D
D e1 but V0 D V0 and e0 D e0 .

Thus V1 D V1 , e1 D e1 , V1 D V1 , e1
In terms of the unit vectors ex , ey and ez the vector V is given by
V D Vx ex C Vy ey C Vz ez
or more compactly by
V D
V e ,
D x, y, z
A9.4.6
A9.4.7
However, using the unit vectors e1 , e0 and e1 , the vector V is given by
V D V1 e1 C V0 e0 V1 e1
or more compactly by
V D
1 V e
D 1, 0, 1
A9.4.8
A9.4.9
Equation (A9.4.9) is readily confirmed by taking eq. (A9.4.6) and using eqs. (A9.4.1)
and (A9.4.2) to express the V and e in terms of the V , and e .
The nature of the relationship between the vector V and V and e has certain consequences, which we now consider. Whereas the orthonormal condition for the unit vectors
in the cartesian system is
A9.4.10
e e D
for the unit vectors e in the spherical system the orthonormal condition is
e e D 1
A9.4.11
Thus, whereas ex ex D 1 and ex ey D 0, e1 e1 D 1 and e1 e1 D 0. Also, whereas the

scalar product of two vectors V and W in the cartesian system is given by

c
V cW D
V W D x, y, z
A9.4.12
We have avoided the use of the Einstein summation convention throughout section A9.4.
Vectors, II
411
in the spherical system the scalar product is given by

s
V sW D
1 V W D
V W D 1, 0, 1

A9.4.13
Consequently the tensorial relations between vectors in the spherical system have forms
different from those in the cartesian system. The properties of tensors in the spherical
system are treated in detail in Chapter A14.
It should be noted that these results arise because spherical vectors span complex space.
In complex space the scalar product of two vectors A and B is given by

B1
A B D [ A1 A0 A1 ] B0
A9.4.14
B1
D A1 B1 C A0 B0 C A1 B1
instead of eq. (A8.5.7), where A1 is the conjugate complex of A1 etc. As A1 D A1 ,
A0 D A0 and A1 D A1 , eq. (A9.4.14) can be rewritten as
A B D A1 B1 C A0 B0 A1 B1
A9.4.15
which corresponds to eq. (A9.4.13). Similarly, the square of the length of the vector A is
given by
A A D A1 A1 C A0 A0 A1 A1
A9.4.16
The use of spherical basis vectors also merits attention. Using the convention introduced
in Chapter A8 of writing basis vectors as row matrices a vector A will be represented as
follows in complex space,

A1
A9.4.17
A D [ e1 e0 e1 ] A0
A1
This may be compared with eq. (A8.3.11). It is instructive to obtain the scalar product
A B employing matrix equations for A and B . Using eq. (A9.4.17) we obtain
A B D [A1 e1 C A0 e0 C A1 e1

] [B1 e1 C B0 e0 C B1
e1 ]
A9.4.18
and using the relationships between spherical unit vectors and spherical vector components
given above this reduces to eq. (A9.4.15).
We now consider how the components of a vector in the spherical basis transform
under rotation of the axis system from S to S0 . For each V0 0 D 1, 0, 1 we write

l0 V
D 1, 0, 1
A9.4.19
V0 D

where the l0 are direction cosines relating to pairs of axes 0 and .
The superscript label s is omitted in eqs (A9.4.14) to (A9.4.18) and the accompanying text to keep the
nomenclature uncluttered.
412
The Raman Effect
For the inverse transformation for each V D 1, 0, 1 we have

V D
l0 V0
0 D 1, 0, 1
A9.4.20
0
The direction cosines l0 may be expressed in terms of the cartesian direction cosines
l0 . The relations between the l0 and l0 correspond to those between the signed vector
dyads 1 V0 W and V0 W and so can be obtained from the matrix B defined in
Chapter A14, section A14.4. Table A9.1 gives the elements of the corresponding direction
cosine matrix c D0 .
Alternatively the direction cosines l0 may be expressed in terms of the Euler angles
by transforming the entries in Table A9.1 using the matrix Eu D0 which relates the
l0 to the Euler angles and is defined in Chapter A5, section A5.4. Table A9.2 gives the
elements of the corresponding direction cosine matrix Eu D0 . The notation used for the
direction cosine matrices is defined in Chapter A5, section A5.4.
It should be noted that the dependence of the Eu D0 on and has the general form
exp i0 C . This is of importance in connection with rotational selection rules for
a symmetric top since its rotational wave function includes the term exp iK C M.
The l0 satisfy the following normalization and orthogonality relationships:

l0 l0 D
A9.4.21
0
Table A9.1 Elements of the direction cosine matrix c D0

c
D0
1
12 flx0 x
10
1
0
2 flx x
0
ly 0 y C ilx0 y C ly 0 x g
C ly 0 y C ilx0 y ly 0 x g
p1 flz0 x
2
00
1
C ilz0 y g
1
flx0 x
2
C ly 0 y ilx0 y ly 0 x g
p12 flx0 z
12 flx0 x ly 0 y ilx0 y C ly 0 x g
p1 flx0 z
2
p12 flz0 x ilz0 y g
Table A9.2 Elements of the direction cosine matrix
Eu
C ily 0 z g
ily 0 z g
lz 0 z
D0

Eu
10
0
D0
10
00
12 cos 1ei ei

1
2 1
C cos ei ei

p1
2
sin ei
1
1
1
2
C cos ei ei
p1
2
sin ei
12 cos 1ei ei
p12 sin ei
p12 sin ei
cos
Vectors, II
413
and
l0 l0 D 0
A9.4.22
Also
l0 D 1j jjj l0
A9.4.23
A9.5 ROTATION OF VECTORS USING SPHERICAL

COORDINATES
We now consider how the spherical coordinates are related to the eigenvectors of a
rotational matrix. We consider for simplicity the case of a vector c V which has components
Vx and Vy in a two-dimensional cartesian axis system S. Under an anticlockwise rotation
through an angle the vector c V is transformed into another vector c W which has the
same magnitude as c V (that is jc W j D jc V j) but different components in the cartesian
system namely Wx and Wy (see Fig. A9.3). Using compact matrix notation we can write
c
W D cR cV
A9.5.1
Adapting eq. (A8.11.10) the expanded form of this is

Wx
Vx
cos sin
D
sin
cos
Wy
Vy
A9.5.2
In order to diagonalize the matrix c R, defined by eqs. (A9.5.1) and (A9.5.2) we define a
matrix X such that
A9.5.3
X c RX D I
If we now multiply both sides of eq. (A9.5.1) on the left by X and insert the identity
XX D I on the right-hand side we obtain
X c W D X c RX
X c V
A9.5.4
cW
Wy
cV
Vy
r
a
f
Wx
Vx
Figure A9.3 Anticlockwise rotation of a vector: jV j D jW j D r.
414
The Raman Effect
This equation represents the transformation of a vector X c V to a vector X c W (in the

same basis) by a rotational transformation matrix X c RX which is diagonal. The boxes
in eq. (A9.5.4) emphasize this interpretation.
The characteristic roots and vectors of the matrix c R were obtained in Chapter A7
using the procedures given there. It was found that

i

1 0
e
0
3D
D
A9.5.5
0 2
0
ei
and
X D
1
p
2
p1
2
i
p
2
i
p
2
A9.5.6
with the first column of X being associated with 1 D ei and the second column with
2 D ei .
We now examine the nature of the vectors X c V , and X c W . We may regard these
vectors as relating to a new basis s. If we label them as s V and s W we may write
s
s
V D X c V
W DX W
A9.5.7
A9.5.8
We note that X is complex and so generates complex components in the basis s.

Using X obtained by forming the conjugate complex of X which is defined by
eq. (A9.5.6), we find that X c V generates components of the vectors s V in the new
basis s as follows
V1 D
1
p
Vx
2
p1 Vx
2
V1 D
iVy
A9.5.9
C iVy
A9.5.10
Taking the complex conjugates we have for the components of s V

s
s
V1 D
V1 D
1
p
Vx
2
p1 Vx
2
C iVy
A9.5.11
iVy
A9.5.12
We recognize these as the spherical components of the vector V in the xy plane. The
vector W resulting from the rotation is described in the new basis as s W and will have
corresponding spherical components, as follows:
s
s
W1 D p12 Wx C iWy
W1 D
p1 Wx
2
iWy
A9.5.13
A9.5.14
Thus, using spherical coordinates, the anticlockwise rotation of the vectors s V through
an angle to give the vectors s W can be expressed as
s
W D sR sV
A9.5.15
Vectors, II
415

where
s
ei
RD
0
A9.5.16
ei
Thus
s
s
W1 D s V1 ei
s
A9.5.17
i
W1 D V1 e
A9.5.18
Using the definitions of spherical coordinates in terms of polar coordinates given by

eqs. (A6.6.7) and (A6.6.9), and setting D /2 so that the radial polar coordinate r lies
in the xy plane, we see that
s
s
W1 D
W1 D
1 iC
p
re
2
p1 reiC
2
A9.5.19
A9.5.20
This is consistent with an anti-clockwise rotation through an angle so that the angle
defining the orientation of the vector V in the xy plane changes to C , which defines
the orientation of the vector W (see Fig. A9.3).
A9.6 VECTORS IN n-DIMENSIONAL SPACE

In Chapter A8 and the preceding parts of this Chapter we have confined our attention
to vectors in three-dimensional space. This has enabled us to visualize the vectors and
construct three-dimensional models to represent them, thus facilitating our understanding
of them.
The concept of a vector can be extended to a general space of n dimensions. We can
define an n-dimensional space by choosing a set of n linearly independent basis vectors
e1 , e2 . . . en . These vectors are said to span the n-dimensional space. Any vector V in the
space is related to these basis vectors by a linear relation of the type
V D V1 e1 C V2 e2 C . . . Vn en
A9.6.1
where V1 , V2 . . . Vn are the components of the vector relative to the basis vectors e1 ,
e2 . . . en . Equation (A9.6.1) can be written in matrix notation as
V1
V2
V D [e1 e2 . . . , en ] .
A9.6.2
..
Vn
Equations (A9.6.1) and (A9.6.2) are the analogues of eqs. (A8.3.6) and (A8.3.11) in
three-dimensional space. Clearly all the matrix transformations of vector components and
basis vectors considered in Section A8.10 can be applied to n-dimensional space simply
by considering column matrices to be of order n 1, row matrices to be of order 1 n
416
The Raman Effect
and square matrices to be of order n n. Indeed all the concepts already developed
can be readily generalized to n-dimensional space.
The concept of n-dimensional space is particularly useful when considering problems
involving the 3N degrees of freedom of the N atoms of a molecule, as for example
translational, rotational and vibrational energies, symmetry properties, rotational transformations, and the choice of basis vectors that cast matrix representations in special
forms.
REFERENCES
Bhagavantam, S. (1942). Scattering of Light and the Raman Effect, Chemical Publishing Company:
New York.
Condon, E.U. and Shortley, G.H. (1935). The Theory of Atomic Spectra, Cambridge University
Press: Cambridge.
Fano, U. and Racah, G. (1959). Irreducible Tensorial Sets, Academic Press: New York.
Weber, A. (1973). High resolution Raman studies of gases, in The Raman Effect, (ed.), A. Anderson,
2, 543757. Marcel Dekker: New York.
417
A10
Tensors
You boil it in sawdust; you salt it in glue;
You condense it with locusts and tape;
Still keeping one principal object in view,
to preserve its symmetrical shape.
Lewis Carroll
A10.1 GENERAL DEFINITIONS

The direct product AB of two vectors A and B which is defined in Chapter A8 constitutes
a second-rank tensor. We shall denote this generally by T f2g where the superscript f2g
denotes the rank. Thus
A10.1.1
T f2g D AB
This tensor is said to be of rank two because each of the nine terms in the expansion of
AB in eq. (A8.5.20) involves dyads which are the products of two vector components.
We may say that T f2g is a dyad which can be expressed as a dyadic, i.e. a sum of dyads.
The tensor T f2g is a linear operator which sends a vector V into another vector according
to the following rule:
T f2g V D ABV
D AB V
A10.1.2
Alternatives to the word rank sometimes used in the literature are degree and order. Alternatives to
tensor are names which incorporate the rank as for example: diad, triad, tetrad; or bisor, trisor, tetror. These
have not found much favour.
The superscript for rank is placed in braces f g to avoid confusion with the labels for irreducible tensor
components which are written in parentheses ( ) as e.g. j
m . In many situations the tensor rank symbol can
be omitted without creating ambiguities.
418
The Raman Effect
where T f2g V or AB (V) denotes the linear operator T f2g or AB acting on the vector V.
Since B V is a scalar product, AB V is a vector.
An extension of this process is for T f2g to act as a linear operator on a vector dyad
VW and produce a scalar according to the rule
T f2g VW D ABVW
D AB : VW
D B V A W
A10.1.3
The result is a scalar because all terms in the last line are scalars. The symbol : is termed
a double dot product and the process may be described as the double dot product of a
dyad acting on a dyad.
The direct product of three vectors A, B and C, which is defined in Chapter A8, is a
third-rank tensor which we shall denote by T f3g . Thus
T f3g D ABC
A10.1.4
This tensor is said to be of rank three because each term in eq. (A8.6.10) involves triads.
The tensor T f3g is a linear operator which sends a vector V into a second-rank tensor
according to the rule
T f3g V D ABC V
D ABC V
A10.1.5
The result is clearly a second-rank tensor as the direct product AB is a second-rank tensor
and C V is a scalar.
T f3g can also act as a linear operator on a vector dyad VW and produce a vector
T f3g VW D ABC VW
D ABC : VW
D AC V B W
A10.1.6
The result is a vector, because A is a vector and C V and B W are scalars.

Finally, T f3g can act as a linear operator on a vector triad VWX and produce a scalar
T f3g VWX D ABC VWX
.
D ABC .. VWX
D C V B W A X
A10.1.7
.
The result is clearly a scalar because all terms in the last line are scalars. The symbol .. is
termed a triple dot product and the process may be described as the triple dot product of
a triad acting on a triad.
The extension of the above concepts to tensors of higher order is straightforward and
will not be treated at this point.
Tensors
419
In the above definitions we have chosen the most general definitions, namely a vector
dyad with the two vectors different and a vector triad with the three vectors different.
However, the same considerations apply to vector dyads with V D W and vector triads
with V D W 6D X or V D W D X . The formulae above are easily adapted to these cases.
The concept of tensors based on the direct products of vectors can be extended to
include the scalar products of two vectors and the multiplication of a vector by a scalar.
Thus the scalar product of two vectors A and B to give a scalar C, namely
CDAB
A10.1.8
can also be interpreted as an operator A acting on a vector B to produce a scalar according

to the rule
C D AB
A10.1.9
The operator A may be regarded as a tensor of unit rank and it can be denoted generally as
T f1g . It is of rank one as it contains vector components singly. These are of the type ex Ax ,
ey Ay and ez Az in the cartesian system. Consequently we may rewrite eq. (A10.1.9) as
C D T f1g B
A10.1.10
T f1g D A.
A10.1.11
where
Similarly, the operation of multiplying a vector by a scalar may be interpreted as
equivalent to the operation on the vector of an appropriate tensor, which in this case is a
scalar a. It has no vector components and so is formally of rank zero, and can be denoted,
generally, as T f0g . Thus
A10.1.12
C D T f0g B
where
T f0g D a.
A10.1.13
Equation (A10.1.12) is thus equivalent to

C D aB .
A10.1.14
The concept of a tensor as a linear operator provides a consistent framework for tensors
ranging from zero rank upwards and embraces the concepts of scalars and vectors.
A10.2 REPRESENTATION OR SPECIFICATION OF A TENSOR

To apply a tensor in a particular physical situation we need to proceed from the formal
representation of the linear operation implied by the symbol T fRg , where R is the rank,
to a representation, or specification, of the tensor. This involves the choice of a basis
system and the specification of a set of components for the tensor in that system. In this
chapter we consider only the cartesian basis system, but in Chapter A14 which relates to
420
The Raman Effect
a specific tensor, the polarizability tensor, we deal with three bases, the cartesian basis,
the spherical basis and the irreducible basis. The components of a tensor in the cartesian
basis, which are necessarily real, may be found by using the definition of a tensor of rank
R as the direct product of R vectors in the cartesian basis. Thus, the components of a
first-rank tensor can be obtained from eq. (A10.1.11), the components of a second-rank
tensor from eq. (A10.1.1) and the components of a third-rank tensor from eq. (A10.1.4).
It can be seen that a first-rank tensor may be specified by a set of three components, each
associated with one coordinate axis, which is consistent with a vector; a second-rank tensor
may be represented by a set of nine components, each associated with two coordinate
axes; a third-rank tensor may be specified by 27 components, each associated with three
axes, and so on. In general, for three-dimensional space, the number of components is 3R
where R is the rank of the tensor. Each component of a tensor of rank R has R subscripts
and is associated with R coordinate axes.
The components of a tensor clearly will depend on the axis system chosen; a change of
axis system, for example by rotation, will lead to another set of components for the tensor,
except in the case of a tensor of zero rank, that is a scalar. However, the tensor itself is
a geometric invariant. It is a linear operator which represents a particular operation and
the result of performing this operation cannot depend on the coordinate system used to
represent the tensor. Tensors can be said to be blissfully unaware of the way we choose
to represent them. The representations are our concern and are our devices for calculating
the effect of a tensor operation. The representation is usually chosen to be apposite to the
problem being treated.
We have already seen that the representation of vectors (or first-rank tensors) by row or
column matrices is useful particularly in relation to the mathematical operations associated
with them. We now consider the application of matrix representation to tensors of rank
higher than one.
We may represent a second-rank tensor T f2g by the following 3 3 matrix of its nine
dyad components:
Txx Txy Txz

A10.2.1
T f2g
D Tyx Tyy Tyz
Tzx Tzy Tzz
where Txx D Ax Bx , Txy D Ax By etc. in accordance with eq. (A8.5.20). This representation
has as its basis the nine pairs of cartesian basis vectors, or dyads.
We can see that the linear operation of T f2g on a vector V defined in eq. (A10.1.2) may
now be expressed in the form of a matrix equation:
0
Vx
Txx Txy Txz
Vx
V0y D Tyx Tyy Tyz Vy
A10.2.2
Tzx Tzy Tzz
Vz
V0z
where Vx , Vy and Vz are the components of the original vector V and V0x , V0y and V0z are
the components of the transformed vector V 0 . Both sets of components are related to the
same cartesian axis system, namely x, y, z.
Tensors
421
It is often convenient to use the compact matrix form for eq. (A10.2.2) and write
V 0 D T f2g V
A10.2.3
Equation (A10.2.2) is equivalent to the following set of three linear equations:

V0x D Txx Vx C Txy Vy C Txz Vz
V0y D Tyx Vx C Tyy Vy C Tyz Vz
A10.2.4
V0z D Tzx Vx C Tzy Vy C Tzz Vz

In the definition of T f2g given by eq. (A10.1.2) the selection of the appropriate products of
pairs of components of B and V is made by following the rules of scalar product formation.
In the matrix representation the corresponding process is performed by following the
rules of matrix multiplication. These rules involve matching the second subscripts of a
component of T having first subscripts x, with the same subscript of a component of V
and then summing over all possible values x, y, z of these second subscripts of T f2g to
give V0x and so on. The selection process is readily seen if the set of eqs. (A10.2.4) is
written in the summation nomenclature,
V0 D T V
D x, y, z
D x, y, z
A10.2.5
Equation (A10.2.5) is yet another method of representing eq. (A10.2.2).

Likewise, we may represent a third-rank tensor T f3g by a 3 9 matrix of its 27 components
Txxx Txyy Txzz Txxy Txyx Txyz Txzy Txzx Txxz

T f3g D Tyxx Tyyy Tyzz Tyxy Tyyx Tyyz Tyzy Tyzx Tyxz
A10.2.6
Tzxx Tzyy Tzzz Tzxy Tzyx Tzyz Tzzy Tzzx Tzxz
This representation has as its basis the 27 triads of cartesian basis vectors etc. The linear
operation of T f3g on a vector dyad, defined in eq. (A10.1.6) may then be expressed in the
form of a matrix equation
Vx Wx
Vy Wy
Vz Wz
0
V W
Vx
Txxx Txyy Txzz Txxy Txyx Txyz Txzy Txzx Txxz x y
V0y D Tyxx Tyyy Tyzz Tyxy Tyyx Tyyz Tyzy Tyzx Tyxz Vy Wx
V W
0
Tzxx Tzyy Tzzz Tzxy Tzyx Tzyz Tzzy Tzzx Tzxz y z
Vz
Vz Wy
Vz Wx
Vx Wz
A10.2.7
where we have used a column matrix of nine rows to represent the components of the
vector dyad. The set of three linear equations for V0x , V0y and V0z follow readily from
422
The Raman Effect
eq. (A10.2.7). The compact form of eq. (A10.2.7) is

V 0 D T f3g VW
A10.2.8
V 0 D ABC : VW
A10.2.9
or equivalently
Comparison of eq. (A10.2.9) with eq. (A8.6.10) which gives the 27 cartesian-based
components of the direct product of three vectors ABC shows how the subscripts on
the 27 terms Txxx , Txyy ,. . . etc arise.
Equation (A10.2.7) may be rewritten using the summation convention
D x, y, z
V00
D T V W
D x, y, z
A10.2.10
D x, y, z
A10.3 TRANSFORMATION OF TENSORS UPON ROTATION OF

AXES
We may use direction cosines to express the components of a second-rank tensor in a
new cartesian axis system x 0 , y 0 , z0 in terms of its components in the system x, y, z. The
new primed axis system x 0 , y 0 , z0 , which is obtained by rotation of the original system x,
y, z, and the direction cosines relating the two axis systems are defined and described in
Chapter A4.
We consider as an example of a second rank-tensor the polarizability tensor a which
relates the two vectors p0 , the electric dipole amplitude, and E0 , the amplitude of the
electric field strength. Using matrix representation and the unprimed axis system, we may
write, following eq. (A10.2.2),
p x0
xx xy xz
Ex 0
py0 D yx yy yz Ey0
A10.3.1
pz0
zx zy zz
Ez0
Using the summation convention this may be written in the form
p0 D E0
A10.3.2
When the vectors and tensor are defined in the primed axis system the corresponding
form of eq. (A10.3.2) is
p00 D 0 0 E00
A10.3.3
We recall from Chapter A8, Section A8.7 that for a vector V its coordinates V0 in
the primed axis system, and V in the unprimed axis system, are related, using direction
cosines, as follows :
A10.3.4
V0 D l0 V
See the footnote on page 354 in Chapter A3.
Tensors
423
and inversely as
V D l0 V0 D l0 V0
A10.3.5
since the direction cosines l0 and l0 are equal .

Using eqs. (A10.3.4) and (A10.3.5) with eq. (A10.3.2) we may write
p00 D l0 p0
D l0 E0
D
A10.3.6
l0 l0 0 E00
Comparison of eqs. (A10.3.2) and (A10.3.3) yields

0 0 D l0 l 0
A10.3.7
which is the desired relationship between the 0 0 and the . We should note that this
condensed notation represents nine equations (one for each of the new tensor components
0 0 ), each consisting of nine terms (one for each of the old components ).
We can proceed in a similar manner for a third-rank tensor. We consider as an example
the hyperpolarizability tensor b which relates components of the second-order electric
dipole vector p02 and terms which are quadratic in the amplitudes of the electric field
strength vectors. In component form, and using the summation convention in the unprimed
axis system x, y, z, we have
A10.3.8
p2
0 D E0 E0
and in the primed axis system x 0 , y 0 , z0 we have
0
0
0 0 0
p2
0 D E0 E0
A10.3.9
By successive application of the vector transformation, eqs. (A10.3.4) and (A10.3.5),

we find that
A10.3.10
0 0 0 D l0 l 0 l 0
which is the desired relationship between the 0 0 0 and the .
If we recall that a vector is a first-rank tensor then eq. (A10.3.4) is also the transformation equation for the components of a first-rank tensor under rotation of the cartesian
axes. Bearing in mind too, that a scalar, which is a tensor of zero rank, is invariant under
rotation of the cartesian axes, we see that a tensor may be defined as a quantity which
transforms according to the equation
T0 0 0 ... D l0 l 0 l 0 . . . T...
A10.3.11
The rank of a tensor is thus determined by the number of direction cosines involved
in the transformation equation (A10.3.11), or equivalently by the number of subscripts
on the components of T. It must be emphasized that the sole criterion as to whether
a set of numbers consitutes a tensor is whether or not the transformation rule for the
See the footnote on page 351 in Chapter A3.

424
The Raman Effect
corresponding set of numbers in any other coordinate system is of the form corresponding
to eq. (A10.3.11).
In Chapter A8 we made the distinction between polar and axial vectors. We now
introduce this distinction for tensors. The generalization of the tensor transformation law
eq. (A10.3.11) to a polar tensor P of any rank is
P0 0 ... D l0 l 0 l 0 . . . P...
A10.3.12
and to an axial tensor A, of any rank is

A0 0 0 ... D l0 l 0 l 0 . . . A...
A10.3.13
The negative sign in eq. (A10.3.13) is taken for transformations such as reflections and
inversions which change the sign of the axis (improper rotations), and the positive sign
for transformations which do not change the sign of the axis (proper rotations). For an
inversion the corresponding direction cosine is l0 D 1. Hence it follows that axial
tensors of even rank change sign under inversion but axial tensors of odd rank do not.
It is clear from these definitions that the scalar product P P0 of two polar vectors or
A A0 of two axial vectors is a number which does not change sign upon inversion, that is
a true scalar. However the scalar product P A of a polar and an axial vector is a number
that does change sign under inversion and this is termed a pseudo-scalar.
It follows from eq. (A8.5.12) that the vector product of two polar vectors P and P 0 is
given by
P P 0 D Py Pz0 Pz Py0 ex C Pz Px0 Px Pz0 ey
C
Px Py0
Py Px0 ez
A10.3.14
A
Thus P P 0 is an axial vector A whose components are equivalent to the components of

the second-rank antisymmetric polar tensor
P D P P0 P P0 D P
The explicit components are
Pxx Pxy Pxz
Pyx Pyy Pyz

Pzx Pzy Pzz
0
Px Py0 Py Px0
D Px Py0 Py Px0

0
0
0
0
Px Pz Pz Px Py Pz Pz Py0
0
Az Ay
D Az
0
Ax
Ay Ax
0
A10.3.15
Px Pz0 Pz Px0
Py Pz0 Pz Py0 A10.3.16

0
Tensors
425
In general, an axial tensor can be represented by an antisymmetric polar tensor of higher

rank, which usually provides a more fundamental description of the corresponding physical entity.
A10.4 SOME PROPERTIES OF TENSORS

A10.4.1 General
We now consider some properties of tensors. We begin by stating a general property,
applicable to all tensors: the sum or difference of two or more tensors of the same rank
and type is a tensor of the same rank and type. We now consider some specific properties
of tensors of ranks two, three and four.
A10.4.2 Tensors of rank two

If T f2g is a second-rank tensor with nine real components T then if
T D T
A10.4.1
for all and it is said to be symmetric. If

T D T
A10.4.2
for all and then it is said to be anti-symmetric. The condition (A10.4.2.) requires all
the diagonal components of a second-rank antisymmetric tensor to be zero. When neither
of these relationships holds, a given second-rank tensor can always be represented as the
sum of a symmetric tensor T s and an antisymmetric tensor T a . It is easily verified that
T D Ts C Ta
A10.4.3
Ts D 12 T C T
A10.4.4
Ta D 12 T T
A10.4.5
where
and
The property of being symmetric or antisymmetric is unaltered when a tensor is transformed from one axis system to another.
A second-rank tensor in which the three diagonal elements are unity and the six offdiagonal elements zero is called a unit tensor, for which we shall use the symbol I.
The set of quantities (Kronecker delta) defined in eq. (A3.2.8) is a symmetric unit
polar tensor of rank two, which we shall call the delta tensor.
The summation convention implied in writing means that
D xx C yy C zz D 3
A10.4.6
The case of complex components is considered in Chapter A14, Section A14.3.2.

Also called skew symmetric, particularly in describing matrices. See for example Chapter A7,
Section A7.3.
426
The Raman Effect
It should be carefully noted that in this convention xx D yy D zz D 1 but D D 3.

The product is equivalent to because
D xx xx C xy xy C xz xz
C yx yx C yy yy C yz yz
C zx zx C zy zy C zz zz D 3
A10.4.7
It follows from eq. (A10.4.6) that

D 9
A10.4.8
A10.4.3 Tensors of rank three: the alternating or LeviCività tensor

A useful tensor of rank three is the alternating tensor with components , sometimes
called the LeviCività tensor. This is a completely antisymmetric, unit, axial tensor. The
only non-vanishing components are those with three different subscripts. We set
xyz D 1 and the other non-vanishing components are either C1 or 1, depending on
whether the sequence can be brought into the order x y z by a cyclic or a non-cyclic
permutation. Thus
xyz D zxy D yzx D 1
xzy D yxz D zyx D 1
A10.4.9
and all other components are zero. This definition applies whether x, y, z refers to the axes
of a right- or left-handed coordinate system because components of axial tensors of odd
rank do not change sign under inversion of the coordinate axis. Further, the components
of the unit tensor are identical in all coordinate systems. It is another example of an
isotropic tensor.
The invariance under rotation may be shown as follows. Application of the direction
cosine transformation gives
A10.4.10
0 0 0 D l0 l 0 l 0
Expanding and remembering that the only non-zero terms are those which have ,
and all different, we retain the following direction cosine components:
l0 l 0 l 0 C l 0 l 0 l0 C l 0 l0 l 0
l 0 l0 l 0 l 0 l 0 l0 l0 l 0 l 0
A10.4.11
If D all components cancel; similarly, if D or D . If , and are all different

then we can write the retained direction cosine components in the form

l0 l0 l0

l 0 l 0 l 0
A10.4.12

0

l l 0 l 0
and this determinant is equal to 1 if , , are in even order and 1 if they are in odd
order. Thus transforms into 0 0 0 under the rule for third-rank tensors.
Tensors
427
It follows from the definition of that the tensor formulation of the vector product
defined by eqs (A8.5.13) and (A8.5.14) is
A D P P0
A10.4.13
For example, taking D x, summing over pairs of repeated subscripts and using
eq. (A10.4.3), we have
Ax D xxx Px Px0 C xxy Px Py0 C xxz Px Pz0
C xyx Py Px0 C xyy Py Py0 C xyz Py Pz0
A10.4.14
C xzx Pz Px0 C xzy Pz Py0 C xzz Pz Pz0

D Py Pz0 Pz Py0
Two useful relationships involving the delta and alternating tensors are
D D 0
A10.4.15
D
A10.4.16
and
The latter may be established as follows. If D or D both sides of eq. (A10.4.16)
vanish independently. Without loss of generality we may now choose D x and D y.
The left-hand side of eq. (A10.4.16) then becomes
xyx x C xyy y C xyz z D z
and the right-hand side becomes x y x y D , say.
As 6D there are just the following possibilities: D z, in which case D 0 for all
; D z, in which case D 0 for all ; D x and D y, giving D 1; and D y and
D x, giving D 1. Hence D z and the identity is proved.
A10.4.4 Isotropic tensors of various ranks

There is no isotropic tensor of rank one. The only isotropic tensors for tensors of ranks
two and three are scalar multiples of and , respectively. For tensors of rank four
there are three independent ones, namely

C

A10.4.17
A10.4.5 Tensor contraction

The sum of the diagonal components of a second-order tensor, Txx , Tyy and Tzz , is
invariant under rotation because adapting eq. (A10.3.7) and setting 0 D 0 we have
T0 0 D l0 l0 T D T D T
A10.4.18
428
The Raman Effect
The sum
T D Txx C Tyy C Tzz
A10.4.19
is a scalar. This is an example of the operation of putting two subscripts equal in a tensor
and then summing. This is called tensor contraction and results in the order of the tensor
being reduced by two.
We may use eq. (A10.4.16) to provide other examples. Contraction of eq. (A10.4.16)
by setting D gives
D D 3 D 2
A10.4.20
and further contraction by setting D yields

D D 9 3 D 6
A10.4.21
A10.5 IRREDUCIBLE TENSORIAL SETS

We begin by defining a tensorial set. A tensorial set is the set of all the components of
a tensor, and we shall represent these as a column matrix. For a first-rank tensor T f1g ,
this column matrix has three entries; for a second-rank tensor T f2g it has nine entries;
and for a tensor of rank R it has 3R entries. For the special case of a first-rank tensor, the
tensorial set of three components corresponds to the set of three components of a vector.
Under a rotational transformation (rotation of coordinate axes), a tensorial set a with
components ai transforms to a new set b with components bk . The bk are related to the
ai as follows:

bk D
Dki ai
A10.5.1
i
where the Dki are transformation coefficients. In general, each bk is a linear combination
of all the ai . The relationship between the tensorial sets b and a may also be expressed
in compact matrix form as
b D Da
A10.5.2
where D is a 3R 3R matrix of the transformation coefficients Dki , and R is the rank of
the tensor. D is required to have an inverse, that is det D 6D 0.
However, it is often possible to replace the tensorial set a by a new set a 0 , such that a
rotational transformation transforms certain subsets of a 0 separately . The components of
the new set a 0 are related to the components of the old set a by a unitary transformation,

ar 0 D
Ari ai
A10.5.3
i
where the Ari are coefficients. The corresponding compact matrix equation is
a 0 D Aa
A10.5.4
It is implied here that if a subset transforms separately the rest of the set also transforms separately. This
is because we consider sets whose rotational transformations are unitary.
Tensors
429
Figure A10.1 Diagrammatic representation of a transformation matrix in reduced form.

All matrices outside the shaded areas vanish.
where A is a 3R 3R matrix of the coefficients Ari . It then follows that the rotational
transformation of the new tensorial set a 0 is given, in compact matrix notation, by
b 0 D D 0a 0
A10.5.5
D 0 D ADA1
A10.5.6
where
The transformation matrix D 0 has its non-vanishing elements concentrated in square blocks
or sub-matrices along the diagonal (see Fig. A10.1). Each of the subsets constitutes a
separate tensorial set whose transformation matrix is one of the submatrices along the
diagonal of D 0 .
This process of reducing a tensorial set into separate subsets has an ultimate limit. The
ultimate subsets are called irreducible tensorial sets and their transformations are called
irreducible transformations. The process of finding irreducible tensorial sets is not usually
a trivial matter, and in general we shall quote the results and then show that they have
the desired properties.
Before we treat specific cases of irreducible tensors it will be helpful to consider further
the nature of the reduction process. When the ai are regarded as the components of a
vector in a representative space and their transformations are considered as changes of
coordinates in this space, the reduction procedure may be regarded as follows. A rotation
of coordinate axes of the physical space produces a change of axes of the representative
space. This change intermixes all axes corresponding to the vector components ai , but
the axes corresponding to each subset of the new components ar0 intermix only among
themselves and not with the axes corresponding to other subsets. In other words, the axes
corresponding to each irreducible subset ar0 span a subspace of the representative space
which remains invariant under rotational transformations. Each irreducible subset of ar0
may be regarded as the set of components of the projection of the initial vector into one
invariant subspace.
The sets of components of tensors are in general reducible, except for the special
cases of tensors of rank zero (scalars) and rank one (vectors). Sets of unit basis tensors
See Chapter A8, Sections A8.10 and A8.11.
430
The Raman Effect
are similarly reducible. Their reduction is made by a transformation which is said to

be contragredient to that which reduces the tensor components. The irreducible sets
obtained by reducing sets of tensor components and sets of unit basis tensors may be
used to construct new tensors in various ways.
Consider first the simultaneous reduction of the set of components of an ordinary tensor
T and of the corresponding set of unit basis tensors. This operation splits the tensor into
a sum of tensors, each of which consists of an irreducible set of components and of
an irreducible set of basis tensors. Each of these tensors may accordingly be called an
irreducible tensor. For example, a second-rank tensor represented by
T D ex ex Txx C ex ey Txy C C ez ez Tzz
A10.5.7
resolves into three irreducible parts,

T D T 0 C T 1 C T 2
A10.5.8
where
T 0 D 13 ex ex C ey ey C ez ez Txx C Tyy C Tzz
T 1 D 12 [ey ez ez ey Tyz Tzy C ez ex ex ez Tzx Txz
T 2
A10.5.9
A10.5.10
C ex ey ey ex Txy Tyx ]

D 12 13 2ez ez ex ex ey ey 2Tzz Txx Tyy C ex ex ey ey Txx Tyy
Cey ez C ez ey Tyz C Tzy C ez ex C ex ez Tzx C Txz C ex ey C ey ex Txy C Tyx

A10.5.11
and the superscripts in parentheses label irreducible representations and not rank (see
footnote on p. 417).
The five equations above need careful study. Equations (A10.5.9) to (A10.5.11) emphasize the structures of T 0 , T 1 and T 2 in terms of irreducible basis sets. However, T 0 ,
T 1 and T 2 may also be written out in the form given by eq. (A10.5.7) giving essentially vector representations. In this sense they may each be regarded as tensors of second
rank with non-independent components.
Consider first T 0 . We see from eq. (A10.5.9) that it is the product of two quantities:
1/2
ex ex C ey ey C ez ez which involves dyads of basis vectors; and 31/2 Txx C Tyy C
3
Tzz which involves components of the basis vector dyads. Each of these is invariant
under a unitary rotation transformation and so is a scalar. Thus T 0 is a scalar or a zero
rank tensor. However, if we introduce a quantity a given by
a D 13 Txx C Tyy C Tzz
A10.5.12
then (A10.5.9) can also be written as

T 0 D ex ex a C ey ey a C ez ez a C ex ey 0 C . . .
A10.5.13
If x 0 D Px and y 0 D P 1 y, x and y are called contragredient variables and are said to undergo a contragredient transformation.
Tensors
431
using the form defined by eq. (A10.5.7) We see that this corresponds to a second-rank
tensor of a particular type, namely an isotropic tensor. It has only three non-zero components which are associated with ex ex , ey ey and ez ez and each component is equal to a.
The matrix representation of the components is

a 0 0
0
T D 0 a 0
A10.5.14
0 0 a
We consider now eq. (A10.5.10) for T 1 . We see that T1 is a vector with
components 21/2 Txy Tyx , 21/2 Tzx Txz and 21/2 Tyz Tzy and three basis
vectors 21/2 ex ey ey ex , 21/2 ez ex ex ez and 21/2 ey ez ez ey . It is also a tensor
of unit rank.
Equation (A10.5.10) can also be written as
T 1 D ex ex 0 C ey ey 0 C ez ez 0
C ex ey 21 Txy Tyx C ey ex 21 Tyx Txy
A10.5.15
C ey ez 21 Tyz Tzy C ez ey 21 Tzy Tyz

C ez ex 21 Tzx Txz C ex ez 21 Txz Tzx
using the form defined by (A10.5.7). This corresponds to an antisymmetric second-rank
tensor. The matrix representation of the components is
1
1
0
2 Txy Tyx
2 Txz Tzx
1
0
A10.5.16
T 1
D 12 Tyx Txy
2 Tyz Tzy
1
2 Tzx
Txz
1
2 Tzy
Tyz
Finally we consider eq. (A10.5.11) for T 2 which involves five components and five basis
vectors. For example 61/2 2Tzz Txx Tyy is the component associated with the basis
vector 61/2 2ez ez ex ex ey ey and so on. The five basis vectors 61/2 2ez ez ex ex
ey ey , 21/2 ex ex ey ey . . . constitute the simplest set of irreducible unit basis vectors of
order five. We may also represent T 2 using the form defined by eq. (A10.5.7):
T 2 D ex ex Txx a C ey ey Tyy a C ez ez Tzz a
C ex ey 21 Txy C Tyx C ey ex 21 Tyx C Txy
C ey ez 21 Tyz C Tzy C ez ey 21 Tzy C Tyz
C ez ex 21 Tzx C Txz C ex ez 21 Txz C Tzx
A10.5.17
This constitutes yet another type of second-rank tensor, a symmetric traceless tensor. Its
matrix representation is
1
1
Txx a
2 Txy C Tyx
2 Txz C Tzx
1
Tyy a
A10.5.18
T 2
D 12 Tyx C Txy
2 Tyz C Tzy
1
2 Tzx
C Txz
1
2 Tzy
C Tyz
Tzz a
432
The Raman Effect
Thus, to summarize, a second-rank tensor which has as the basis for its representation
the nine vector dyads ex ex . . . ex ey . . . as defined in eq. (A10.5.7) can be reduced to give
three irreducible tensors. One is of rank zero (a scalar) whose representation involves
one new basis vector formed by an appropriate linear combination of the original basis
vectors; one is of rank one (a vector) whose representation involves three new basis
vectors; and one is of rank two whose representation involves five new basis vectors. In
total, nine new basis vectors have been formed by appropriate linear combinations of the
nine original basis vectors. Furthermore these three irreducible tensors can be represented
by three second order tensors with non-independent components.
433
A11
Electrostatics
Electricity is of two kinds, positive and negative.
The difference is, I presume, that one comes a
little more expensive, but is more durable . . .
Stephen Leacock
A11.1 INTRODUCTION
For many purposes molecular systems may be considered to be simply collections of
positive and negative electric charges. The interactions between these charges and the
effects upon them of external fields can then be treated according to the laws of classical
physics. We now consider briefly the aspects of those laws which we use in the treatment
of light-scattering phenomena. In this Chapter A11, entitled Electrostatics, we treat cases
where the electric charges are at rest. In the next Chapter A12, entitled Magnetostatics,
we treat cases where the charges are in motion.
A11.2 FORCE BETWEEN CHARGES

If two electrostatic point charges Qa and Qb are at rest in vacuum and separated by a
distance r, Qa exerts a force on Qb which acts along the line joining the two charges. Let
er be a unit vector along this line and pointing towards the charge Qb on which the force
is being calculated, that is from Qa towards Qb . Then the force on Qb , a vector quantity
which we denote by Fb is given, in the rationalized form used in the SI, by
Qa Qb
A11.2.1
Fb D 40 1 2 er
r
In eq. (A11.2.1), which is a form of Coulombs law, 0 is the electric permittivity of
vacuum and has the value 8.85 1012 F m1 . Its presence is a consequence of the use
434
The Raman Effect
of SI units for the other quantities, namely Q (SI unit: C), r (SI unit: m) and F (SI
unit: N). The factor 4 in eq. (A11.2.1) arises as part of a process of rationalizing the
equations used in electromagnetic theory. It is useful to note that the value of 40 1
is 8.987 109 F1 m.
An alternative form of eq. (A11.2.1) is
Qa Qb
Fb D 40 1 3 r
A11.2.2
r
where
A11.2.3
r D rer
This form emphasizes the spherical symmetry of the force at a distance rb from Qa .
When Qa and Qb have the same signs Qa repels Qb . This is consistent with eqs.
(A11.2.1) and (A11.2.2) since when the product Qa Qb is positive the vector Fb has the
same direction as the vector er . (See Fig. A11.1) Similarly when Qa and Qb have opposite
signs Qa attracts Qb ; that is when the product Qa Qb is negative the vector Fb has the
direction er .
If there are a number of point charges Qa , each at a point ai defined by a position
vector rai which points from ai towards b, then the total force Fb acting on Qb is the
vectorial sum of the individual forces produced by the Qa , and is given by
Qa Qb
i
Fb D 40 1
erai
A11.2.4
2
r
ai
i
where
rai D rai erai
A11.2.5
Equation (A11.2.4) may be rewritten as

Fb D 40 1 Qb
Qa
ra2i
erai
A11.2.6
Fb
is a product of the charge Qb and a vector that depends on

and we see that the force
the distribution of the other charges.
If the charge is not concentrated at individual points but distributed continuously with
a total charge density (SI unit: C m3 ) then we can treat the charge dvi in the small
Fb
Qb
rb
er
rb = rb eb
Qa
Figure A11.1 Charges Qa and Qb separated by a distance rb . The force exerted on Qb by

Qa is Fb . Here Fb has the same direction as the unit vector er since Qa and Qb are taken to
have the same signs so that Qa repels Qb .
Electrostatics
435
volume element dvi whose position is defined by the vector ri as a point charge. The
force Fb is then given by the analogue of eq. (A11.2.6), namely

dvi
1
A11.2.7
Fb D 40 Qb 2 eri
v ri
where
A11.2.8
ri D ri eri
If the charges are not in a vacuum, additional factors must be taken into account. In
the special case of a uniform dielectric, the formulae for Fb and Fb may be obtained
simply by replacing 0 by , the electric permittivity of the dielectric, in the formulae
given above.
The magnitudes of Coulombic forces are very considerable relative to gravitational
forces. For two point charges each of charge 1 C, separated by a distance of 1 m, the
force exerted by one upon the other is 9 109 N.
A11.3 ELECTRIC FIELD STRENGTH

The force which a charge Qb experiences when in the vicinity of another charge Qa may
be attributed to the presence of the electric field produced by the charge Qa . We define
the electric field strength E (SI unit: V m1 ) as the ratio of the force on a point charge
to the point charge itself. Thus the electric field Ea due to the point charge Qa at the
location of a test charge Qb is given by
Fb
Ea D
A11.3.1
Qb
and introducing eq. (A11.2.1) into eq. (A11.3.1) we obtain
Qa
Ea D 40 1 2 er
A11.3.2
rb
from which see that Ea is a vector quantity which does not depend on Qb . Here the unit
vector er points in the direction from Qa towards the point for which E is being calculated.
When Ea is positive it has the same direction as er , a situation which is achieved when
Qa is positive. When Ea is negative it has the direction erb , which arises when Qa is
negative.
If the electric field is produced by more than one charge, each charge produces its own
field and the resultant electric field E is simply the vector sum of all the individual
electric field strengths. Thus using eq. (A11.3.1) and eq. (A11.2.6) we have
Qa
i
E D 40 1
eri
A11.3.3
r
2
a
i
i
If the electric field is produced by charges distributed continuously, then using

eq. (A11.3.1) and eq. (A11.2.7) we have

dvi
1
E D 40
e
A11.3.4
2 ri
v ri
436
The Raman Effect
Whereas the electric field due to a point charge Qa is the same whether or not the point
test charge Qb is in the field, even if Qb is large in magnitude compared to Qa , the electric
field produced by a charge distribution might well be disturbed by the introduction of
a finite test charge Q0 because the field due to the test charge may disturb the positions
of the other charges. Then E is defined as the limiting force per unit charge as the test
charge Q0 ! 0, that is
F
A11.3.5
E D lim
Q0 !0 Q0
Equation (A11.3.3) may be used to calculate the electric field due to particular arrays
of point charges. Of particular interest to us are the electric fields due to electric dipoles,
and electric quadrupoles. These are considered in Chapter A13.
It follows from the foregoing discussion that the force F on a charge Q placed in an
electric field defined by the positive vector E is given by
F D EQ
A11.3.6
F and E act along the same line. A positive charge experiences a force in the positive
direction of the electric field; and a negative charge experiences a force in the opposite
direction.
The idea of lines of force can prove useful conceptually, and can be introduced conveniently at this point. Lines of force are lines that originate only on positive electric
charges and terminate only on negative electric charges. The number of lines originating
and terminating on charges is proportional to the magnitude of the charges. The direction
of these lines is everywhere the same as the direction of the electric field. The density of
lines is the number of lines per unit area passing through a surface perpendicular to the
direction of the lines at a given point. A unit density of lines of force is defined as one
line per square metre and corresponds to an electric field magnitude of 1 V m1 .
A11.4 ELECTROSTATIC POTENTIAL

The potential difference between the points A and B in a field, is the external work that has
to be done to move a unit positive charge from A to B. It is independent of the path taken
between A and B. The potential of a given point P is the external work necessary to bring
a unit charge from infinity to P, the potential at infinity being arbitrarily chosen to be zero.
We illustrate these concepts by considering first the special case of the potential due
to a point charge Qa at a distance r. It follows from eq. (A11.3.6) that the magnitude
of the force due to Qa on a unit positive charge is the magnitude of the electric field,
Ea D jE j. The work done by this electric field in moving the unit positive charge an
infinitesimal distance dr is Ea dr, since Ea acts along r. Thus the work done against the
field is Ea dr and therefore
r
D
Ea dr
A11.4.1
1
Using eq. (A11.3.2) this becomes

1
D 4o Qa
1
dr
r2
Electrostatics
437
which upon integration gives
Qa
A11.4.2
r
It is clear from eq. (A11.4.2) that the potential (SI unit: V) is a scalar quantity and its
sign is that of Qa .
Since the potential is a scalar quantity the potential due to a distribution of point charges
Qai is simply the scalar sum of the potentials due to individual charges, that is
Qa
i
A11.4.3
D 4o 1
rai
i
D 4o 1
where rai is defined in eq. (A11.2.5). Similarly, if the charge distribution is continuous then

dvi
1
D 4o
A11.4.4
ri
The relationship between the potential and the electric field follows from differentiating
eq. (A11.4.2) and using eq. (A11.3.2). We see that
d
Qa
A11.4.5
D 4o 1 2 D E

dr
r
where E is defined through the relation
E D Eer
A11.4.6
In eq. (A11.4.5) dr has been taken to be in the direction of the field and the magnitude
of the gradient of the potential is a maximum in this direction. A more general relation
between the potential and field is

ex
C ey
C ez
DE
A11.4.7
x
y
z
or more compactly
D E
A11.4.8
The negative sign in these equations indicates that, of itself, a positive charge will move
from a higher to a lower potential, whereas work must be done to move it in the opposite
direction.
Since the potential is a scalar quantity the potential at any point is simply the algebraic
sum of the potentials due to each separate charge. However, E is a vector quantity and the
resultant field is the vector sum of the individual fields, as already emphasized. Hence,
it is usually simpler to work in terms of potentials rather than fields. When the potential
distribution has been found, the field at any point can be calculated using eq. (A11.4.8).
A11.5 GAUSSS LAW

Gausss law is concerned with the relation of the flux of the electric field strength E over
a closed surface to the total charge enclosed within the surface. It states that the total flux
of E out of any closed surface is equal to the total charge enclosed by it, divided by o .
438
The Raman Effect
If the charge enclosed by a surface S is distributed continuously then the total enclosed
charge Q is given by

A11.5.1
Q D d

where is the charge density and d is an element of the volume enclosed by the
surface S. The flux of E through a vector element of area dS in S is E dS and thus we
may express Gausss law in the form

E dS D o 1 d
A11.5.2
S
theorem
If we now introduce the divergence

which states that for a vector quantity A

A dS D A d
A11.5.3
S
then eq. (A11.5.2) becomes

E d D o
1
d
A11.5.4
This equation is valid for any closed surface in the field and thus the integrands must be
equal. Hence, at every point in the field
E D o 1
A11.5.5
A11.6 THE EQUATIONS OF POISSON AND LAPLACE

If we introduce into eq. (A11.5.5) the electrostatic potential defined as in eq. (A11.4.8)
we obtain
A11.6.1
r2 D o 1
which is known as Poissons equation. If, in a region of the field, the charge density is
zero, then eq. (A11.6.1) becomes
A11.6.2
r2 D 0
which is known as Laplaces equation.
The significance of Laplaces equation can be better appreciated if we express
eq. (A11.6.2) in cartesian coordinates, when we obtain
2 2 2
C 2 C 2 D0
A11.6.3
x 2
y
z
From this we see that at every point in the field, the second partial derivatives of the
potential in each of the three orthogonal directions must add to zero. Thus there can be
no maximum or minimum of potential in a charge free region. At a maximum or minimum
the second partial derivatives are all different from zero, and all must necessarily have
the same sign and so their sum cannot add to zero.
Flux and vector area are discussed in Chapter A12, see pages 441 and 440 respectively.
See Chapter A8, Section A8.9.3.
439
A12
Magnetostatics
. . . he lived to learn
By no endeavour
Can magnet ever
Attract a Silver Churn!
W. S. Gilbert
A12.1 INTRODUCTION
The previous chapter, A11, on electrostatics was concerned with electric charges at rest.
It considered the forces between electric charges and the effects upon them of external
electric fields which themselves arise from arrays of charges. In this chapter we consider
electric charges in motion. The pattern of development will be similar to that used in
Chapter A11 but the treatment is necessarily somewhat more complicated. After establishing the basic concepts we shall generally just quote results and indicate analogies with
the corresponding formulae in electrostatics.
Charges in motion constitute an electric current and we shall be almost entirely
concerned with charges in uniform motion, that is with steady currents.
It will be useful to establish some definitions relating to electric current. Consider a
stream of electrons in a vacuum. If there are N electrons per unit volume, each of charge
e, then the charge density in the vacuum is given by
D Ne
A12.1.1
If the electrons are all moving with the same velocity v (magnitude v) then the charge
dQ out of an area dS in time dt, where dS is normal to v, is given by
dQ D Ne dSv dt
A12.1.2
440
The Raman Effect

dS
dS
, J
, J
dS
dt
dS cos q
(a)
(b)
Figure A12.1 Electron stream in a vacuum: (a) v and J normal to dS; (b) v and J
making an angle with the normal to dS.
as is evident from Fig. A12.1(a). Therefore the charge flowing out per second is
dQ
A12.1.3
D Nev dS
dt
We now introduce a current density vector J (magnitude J), defined as the positive
charge flowing per unit area per second at a point. For the special case illustrated in Fig
A12.1(a), the area dS is also normal to J. It follows that
J D Nev D v
A12.1.4
and hence
dQ
D J dS
A12.1.5
dt
In the more general case illustrated in Fig. A12.1(b), when v and J are not normal to
dS it is useful to introduce the vector element of area dS which is defined by
dS D en dS
A12.1.6
where en is a unit vector normal to the area dS. Thus dS is to be regarded as a vector of
magnitude dS directed along en .
Equation (A12.1.3) must now be replaced by
dQ
A12.1.7
D Nev dS
dt
and eq. (A12.1.5) by
dQ
D J dS
A12.1.8
dt
It follows from eq. (A12.1.6) that
v dS D v dS cos
A12.1.9
and
J dS D J dS cos
A12.1.10
P
where is the angle between dS and v (or J ). The vector form of eq. (A12.1.4) is,
J D Nev D v
A12.1.11
An electric current I is measured for a particular area S and is therefore given by

dQ
A12.1.12
I D J dS D
dt
S
Magnetostatics
441
Thus the electric current is the flux of the current density. The SI unit for electric current
is the ampere (symbol A) and the SI unit for current density is A m2 . We note that
A D C s1 where C is the symbol for coulomb, the SI unit of electric charge or quantity
of electricity.
A12.2 MAGNETIC FORCES

Circuits carrying electric currents exert forces upon each other. In the simple case of two
parallel wires separated by a distance r and carrying steady currents Ia and Ib the force
is found to be proportional to Ia Ib r 1 . This force is attractive if the currents are in the
same direction and repulsive if they are in opposite directions.
In the more general case of a pair of currents in the situation illustrated in Fig. A12.2
the force Fab which current Ib exerts on current Ia in vacuum is found, experimentally,
to be

0
dla dlb r1
Fab D
A12.2.1
Ia Ib
4
r2
a b
where the line integrals are evaluated over the two wires. The vectors dla and dlb point in
the directions of the positive current flow; r1 is a unit vector pointing from dlb to dla ; and
r is the distance between the two vector length elements dla and dlb . In eq. (A12.2.1) the
unit of force is the newton (symbol N), the unit of electric current is the ampere, and the
unit of length is the metre. The constant 0 is called the permeability of vacuum and is
arbitrarily chosen to be exactly 4 107 N A2 . Alternatively, if we introduce the unit
of inductance, the henry (symbol H), then 0 D 4 101 H m1 because N D H m1 A2 .
The double integral in eq. (A12.2.1) can be interpreted as follows. First choose a fixed
element dla on circuit a and add the vectors dla dlb r1 /r 2 corresponding to each
Ib
dlb
r1
dla
Ia
Figure A12.2 The interaction of two currents Ia and Ib .
The flux d of a vector A through an infinitesimal surface of magnitude dS is given by d D A dS

where the vector element of area dS is defined by eq. (A12.1.6). Thus the flux d is the component of the
vector A normal to the surface multiplied by the area of the surface.
For a finite surface the total flux is

found by integrating A dS over the whole surface, that is D S A dS .
442
The Raman Effect
element dlb of circuit b. We then repeat this operation for all other elements dla of circuit
a and finally calculate the overall sum. In general this integration cannot be performed
analytically and numerical integration is then necessary.
Equation (A12.2.1), which gives the force between charges moving with uniform velocities, is the analogue of Coulombs law which relates to the force between two electrostatic point charges at rest in vacuum (see Chapter A11, Section A11.2). Coulombs
law involves 0 the electric permittivity of vacuum, and unlike 0 its value of 8.85
1012 F m1 is determined experimentally.
A12.3 THE MAGNETIC INDUCTION

We can use eq. (A12.2.1) to express Fab as the result of the interaction of current Ia with
the field of current Ib . We rewrite eq. (A12.2.1) in the form

0
dlb r1
Fab D Ia dla Ib
A12.3.1
4
r2
a
Then introducing the quantity Bb , defined by

0
dlb r1
Bb D
Ib
4
r2
A12.3.2
we have
dla Bb
Fab D Ia
A12.3.3
The vector B is called the magnetic induction and Bb can be regarded as the magnetic
induction of circuit b at the position of the element dla of circuit a.
The SI unit for magnetic induction is the tesla (symbol T). We note that T D Wb m2
where Wb is the symbol for the weber, the unit of magnetic flux (discussed later in this
section) In superconducting solenoids, magnetic inductions of the order of 10 T or more
are achievable. This may be compared with values of 1 T in an average electromagnet,
104 T at the earths surface and 1014 T generated by the functioning of the human heart.
An alternative acceptable name for B is magnetic flux density, as is evident from the unit
Wb m2 , but we use magnetic induction throughout.
Equation (A12.3.2) is called the BiotSavart law, and here again the integration can
be performed analytically only for the simplest geometrical forms. It follows from this
law that the element of force dF on an element of wire dl carrying a current I in a region
where the magnetic induction is B, is given by
dF D Idl B
A12.3.4
B is sometimes incorrectly called the magnetic field. However, the magnetic field strength is denoted by
the symbol H and has SI units Am1 . In a vacuum B D 0 H so that the only distinctions between B and H
are in their magnitudes and units. This is sometimes invoked to justify loose terminology. The relation of the
magnetic induction and the magnetic field is considered further in Chapters A13 (Section A13.10) and A16.
Magnetostatics
443
If the current I is distributed in space with a current density J then I is given by JdS
and must be included under the integral sign in eq. (A12.3.3). Since J is parallel to dl
(JdS dl can be written as J d where d is an element of volume. Thus in the general
case the magnetic induction B at a point in space is given by

J r1
0
d
A12.3.5
BD
4 r 2
where the integration is to be carried out over any volume which includes all the
currents.
We see that the magnetic induction B is the magnetostatic analogue of the electric
field E in electrostatics. Lines of magnetic induction which are everywhere tangential to
the direction of B can be useful conceptually in describing a magnetic field and are the
analogues of lines of force used for electric fields (see Chapter A11, Section A11.3). It
is also useful to introduce the concept of the flux (symbol ) of the magnetic induction
B through a surface. This is given by

A12.3.6
D B dS
S
where dS represents a vector element of area as defined in eq. (A12.1.6). The unit of
magnetic flux the weber (symbol Wb) is that uniform flux per second which produces
one volt Wb D Vs.
We may illustrate the above equations by considering two simple cases. In an infinitely
long straight wire carrying a current I as in Fig. A12.3, an element I dl of the current
will produce a magnetic induction whose magnitude dB at the point P is given by
dB D
0 dl sin
I
4
r2
A12.3.7
I
r
dB
r
l
f
r1
dl
Figure A12.3 The magnetic induction dB produced by a vector length element dl of an

infinitely long straight wire carrying a current I. The vector dB lies in a plane which is
perpendicular to the wire and which passes through P.
444
The Raman Effect
Expressing dl, sin

and r in terms of the angle defined in Fig. A12.3 and integrating
from D /2 to D C/2, we find that

0 I
0 I C/2
cos d D
A12.3.8
BD
4 /2
2
Thus the magnitude of B falls off inversely as the first power of the distance from an
infinitely long wire and its direction is perpendicular to a plane carrying the wire. The
lines of magnetic induction are circles lying in a plane perpendicular to the wire and
centered upon it.
We can use this result to find the force between two infinitely long parallel wires
carrying currents Ia and Ib and separated by a distance as in Fig. A12.4. The current Ib
produces a magnetic induction Bb (magnitude Bb ) at the position of the current Ia . Using
eq. (A12.3.4) the force dF acting on a vector length element dla of the wire with current
Ia is given by
A12.3.9
dF D Ia dla Bb
Then using eq. (A12.3.8) we have
dF D
0 Ia dla Ib
2
A12.3.10
and the force per unit length is given by

0 Ia Ib
dF
D
dla
2
A12.3.11
Equation (A12.3.11) provides a definition of the ampere: two long thin parallel wires
separated by a distance of 1 m exert on each other a force of 2 107 N m1 when the
current in each one is 1 A.
Ib
Ia
dla
B
dF
Figure A12.4 The element of force dF acting on the element dla with current Ia as a result
of the magnetic induction Bb produced by the current Ib in the adjacent parallel wire b.
Magnetostatics
445
A12.4 THE LORENTZ FORCE ON A POINT CHARGE MOVING

IN A MAGNETIC FIELD
In Section A12.3 dF , the element of force on a current element Idl was shown to be
dF D Idl B
A12.4.1
Introducing the current density J and the cross-sectional area dS of the conductor we have
Idl D J dS dl D J d
A12.4.2
where d is an element of volume with cross-sectional area dS and length dl. Then
dF D J B d
A12.4.3
or
dF
DJ B
A12.4.4
d
For N charge carriers per unit volume each with charge Q, following eq. (A12.1.11)
we can set J D NQv. Introducing this into eq. (A12.4.4) and rearranging we have
1 dF
A12.4.5
D Qv B
N d
As Nd is the number of charge carriers in the volume element d , dF /N d is the force
per unit charge carrier which we shall denote by f.
The magnetic force on an individual charge Q is therefore
f D Qv B
A12.4.6
This force is known as the Lorentz force. It is perpendicular both to v and B .

If we add to this the Coulomb force which will arise if there is also present an electric
field E, the total force on a charge Q moving with velocity v in both an electric and
magnetic field is
A12.4.7
f D Q[E C v B ]
This result is perhaps not immediately obvious as the Coulomb force describes the interaction of stationary charges. However, the properties of electric charges are invariant at
all speeds, even relativistic ones, so that eq. (A12.4.7) is always true.
An alternative formulation is in terms of the Lorentz force density acting on a region
of charge and current density in an electromagnetic field. The analogue of eq. (A12.4.7)
is then
f D E C J B
A12.4.8
where is the charge density and J the current density.
Equation (A12.4.6) can be used to define the tesla (symbol T), the unit of magnetic
induction. One tesla is that magnetic induction which produces a force of 1 N on a charge
of 1 C moving with a velocity of 1 m s1 normal to the field.
446
The Raman Effect
A12.5 THE DIVERGENCE OF THE MAGNETIC INDUCTION B

The BiotSavart law as formulated in eq. (A12.3.2) may be used to show that
B D0
A12.5.1
This result is one of the four Maxwell equations.

The net flux of magnetic induction through any closed surface is always equal to zero
since

B dS D B d D 0
A12.5.2
S
This result follows from the definition of the magnetic induction which itself is based on
the empirical law describing the forces between current elements. There are no sources of
magnetic induction at which the divergence of B would be different from zero and there
are no free magnetic charges corresponding to the free electric charges in an electrostatic
field. Equation (A12.5.1) may be compared with eq. (A11.5.5) which is a statement of
Gausss law in electrostatics.
A12.6 THE VECTOR POTENTIAL A

We saw in Chapter A11, Section A11.4, that the electrostatic field strength E can be
derived from the potential
using the relation E D
. Similarly for the magnetic
induction B there exists the following relationship involving a new quantity A, which by
analogy is called the vector potential:
B DA
The vector potential A is defined by
0 I
AD
4
dl
r
A12.6.1
A12.6.2
and has as its unit Wb m1 . If the current is distributed with a current density J, then
instead of eq. (A12.6.2) we have

0
J d
AD
A12.6.3
4 r
Equation (A12.6.1) is a completely general result and is true under all circumstances
including time-dependent currents. Equation (A12.5.1) follows immediately from
eq. (A12.6.1) because the divergence of the curl of a vector is always zero.
The vector potential A is not uniquely defined by eqs. (A12.6.2) or (A12.6.3) as we
can add to their right-hand sides any term whose curl is zero.
Chapter A16 deals with the Maxwell equations.

Chapter A8 deals with the properties of vectors.
Magnetostatics
447
The magnetic flux through a surface S can also be expressed in terms of the vector
potential. Introducing eq. (A12.6.1) into eq. (A12.3.6) we obtain

A12.6.4
D A dS
S
We now invoke the theorem due to Stokes. This relates the line integral to the surface
integral over any surface of which the line integral is a boundary, and is given by

A dl D r A dS
A12.6.5
S
Using this relation eq. (A12.6.4) becomes

D
A dl
A12.6.6
Thus the flux through S is given by the line integral of A dl around the boundary of S.
The vector potential A is related to the scalar electrostatic potential
. In the special
case when
is constant with time we have
AD0
This is a special case of what is called the Lorentz condition.

In free space it is found that
B D 0 J
A12.6.7
A12.6.8
for quasi-stationary currents. This leads to the Poisson equation (for cartesian coordinates
only) for the vector potential:
r2 A D 0 J
A12.6.9
again for quasi-stationary currents only. This may be compared with eq. (A11.6.1),
Poissons equation in electrostatics.
Chapter A11, Section A11.4 deals with the scalar electrostatic potential.
449
A13
The Interaction of a System of
Electric Charges with Electric
and Magnetic Fields
Ready to break forth into new perturbations. . .
Francis Bacon
A13.1 INTRODUCTION
In this section we consider some properties of a system of many electrical charges.
The system of particular interest to us is a molecular system, which has, not only a
static charge distribution, but also a current distribution arising from the motion of the
charges.
The properties of such a system are determined by the distributions of the charges and
the currents. The distribution of the charges can be expressed in terms of electric multipole
moments and the distribution of currents in terms of magnetic multipole moments. The
energy of the charges and of the currents in electric and magnetic fields, both static and
dynamic, can then be related to the electric and magnetic multipoles. These multipoles
must include not only any permanent moments but also those induced in the system
by the external fields. The problem will be treated classically in the first instance. This
leads to interaction hamiltonians which can be cast readily in operator form for quantum
mechanical purposes.
A completely general approach to the problem is rather onerous and leads to quite
complicated formulae which are needed for only a small number of cases. We therefore
treat the problem in two stages.
450
The Raman Effect
In the first stage, which we shall term the basic treatment, we introduce some physically
realistic approximations. We assume that the charge distribution can be defined in terms
of the electric dipole alone, and neglect all higher electric multipoles and all magnetic
multipoles. We also assume that the electric field is uniform over the system of charges
so that the electric field gradient is zero. These assumptions substantially simplify the
treatment, but impose limited restrictions on the applicability of the results. In the second
stage we present a more general treatment.
In both stages we shall treat first the case of static fields and then consider what changes,
if any, are required for dynamic fields, particularly those associated with electromagnetic
radiation.
A13.2 POINT CHARGES IN MOLECULAR SYSTEMS

We begin with the definition of the point charges involved in molecular systems. The
charge on a proton is Ce and has the value 1.602 1019 C; the charge on the electron
is e. The net charge Q of a collection of such point charges ei , is given by

Q D
ei
A13.2.1
i
where ei is Ce when the particle is a proton and e when the particle is an electron. The
net charge can be considered to be the electric monopole moment or zeroth moment of
the set of charges and is, of course, a scalar quantity. In neutral molecules Q D 0.
A13.3 ELECTRIC DIPOLE IN A MOLECULAR SYSTEM

The first moment of a set of charges is called the electric dipole moment. It is a vector
quantity which is defined by the relationship

pD
ei r i
A13.3.1
i
where p is the electric dipole moment vector and ri is the position vector of the ith charge
ei . The electric dipole moment vector is independent of the choice of origin only if the
net charge Q is zero. For charged species it is convenient to choose the centre of mass
as the origin. With this convention the hydroxide ion OH , for example, will possess
a dipole moment but ions such as NO2 C and SO4 2 will not. Of course, if Q is not
zero, there exists a unique point, called the centre of charge, relative to which p D 0. It
is useful to note that if the charge distribution is symmetric about say the z axis, then px
and py , the x and y components of p, respectively, are zero, and only one component of
p, namely pz is needed to describe the first moment of the charges.
It follows from eq. (A13.3.1) that for the simple case of just two point charges CQ
and Q separated by a distance r, the magnitude of the resulting electric dipole is Qr
and this acts along the line joining the two charges. The convention as to the direction of
this dipole is to regard it as directed from the negative charge to the positive charge as
shown in Fig. A13.1.
The Interaction of a System of Electric Charges with Electric and Magnetic Fields
451
+Q
r
Figure A13.1 The electric dipole. By convention the direction of an electric dipole is taken
to be from the negatively charged end to the positively charged end, as shown by the bold
arrow. For the electric dipole shown above the magnitude, p is given by p D Qr and the
electric dipole vector p is given by p D Qr.
The SI unit for the electric dipole moment is C m. If Q has the magnitude of the
electronic charge, that is jQj D jej, and the separation distance r is 1010 m the magnitude
of the corresponding electric dipole moment is 1.602 1029 C m.
The classical unit for the electric dipole moment, based on the esu system, is the
debye (symbol D). This esu unit corresponds to 3.336 1030 C m, and thus 1.602
1029 C m is approximately 4.8 D. Molecular dipoles are usually of the order of a few
debye. An alternative unit is electric dipole length, l D p/jej. Thus an electric dipole
moment of 3.336 1030 C m corresponds to a dipole length of 2.083 1011 m or
20.83 pm.
A13.4 BASIC TREATMENT OF THE ENERGY OF A

DISTRIBUTION OF POINT CHARGES AT REST IN A
UNIFORM ELECTRIC FIELD
We now consider the energy of a distribution of charges ei located at points defined by
vectors ri (corresponding to scalars ri ) from the origin O to ei which are placed in an
external static electric field which is produced by distant charges. If the scalar potential
of this external field at the point defined by ri is i , the energy of interaction U of the
distribution of charges with the external electric field is

ei i
A13.4.1
UD
i
We may express i in terms of the scalar potential 0 and its derivatives at the origin O.
Then, to the first order, eq. (A13.4.1) becomes

ei 0 C
xi C
yi C
zi
A13.4.2
UD
x
y
z
0
0
0
i
where the subscript 0 indicates that the values of the potential and its derivatives are those
at the origin O. It is convenient to introduce at this stage a shorthand notation which will
For an esu system defined in terms of four base units and four independent dimensions 1 D D 1018 Fr cm.
The symbol Fr stands for the franklin, one definition of which is that two charges of 1 Fr, 1 cm apart in a
vacuum, repel each other with a force of 1 dyne.
452
The Raman Effect
prove useful in subsequent developments. We rewrite eq. (A13.4.2) as

ei 0 C
ri
UD
r 0
i
A13.4.3
where Greek subscripts denote cartesian components, so that ri can be xi , yi , or zi and
repeated Greek subscripts imply a summation over all components. Thus, for example,
ri ri D xi2 C yi2 C zi2 D ri2
If we now introduce the definitions of Q and p given by eqs (A13.2.1) and (A13.3.1)
respectively and recognize from Chapter A11 that

D E 0
A13.4.4
r 0
where E 0 is the component of the electric field at the origin O, then eq. (A13.4.3)
becomes
U D Q 0 p E 0
A13.4.5
It is clear from eq. (A13.4.4), that restricting the expansion of the potential to the
first order in eq. (A13.4.2) corresponds to our assumption, at this stage, that the electric
field is uniform over the system of charges.
For an uncharged molecule, for which Q D 0 eq. (A13.4.5) reduces to
U D p E 0
A13.4.6
which may be written in vector form as

U D p E 0
A13.4.7
U D pE0 cos
A13.4.8
or alternatively as
where is the angle between the electric dipole vector and the electric field vector. These
results are illustrated in Fig. A13.2 which shows how the energy of a simple electric
dipole in a uniform electric field is determined by consideration of the potential gradient.
It should be noted that when D 0 and the electric dipole vector and electric field vector
are parallel, U D pE corresponding to the minimum interaction energy. The maximum
interaction energy is U D pE, which arises when D 180 and the electric dipole vector
and the electric field vector are antiparallel. In a uniform electric field there is no resultant
force on a dipole, but it is of course subject to a torque which is given by T D p E . This
has a maximum value when D 90 and is zero when D 0 . Figure A13.3 illustrates
how the torque on a simple electric dipole in a uniform electric field is determined.
The classical relationships given by eqs. (A13.4.6) to (A13.4.8) for the interaction of a
molecule with an external static electric field are readily adapted to give the corresponding
O Thus, using eq. (A13.4.7) we have
interaction hamiltonian H.
O D p E 0
H
A13.4.9
y
+Q
r
2
q
x
r
2
Q
f
= E
x
f r
f+, potential at +Q = fo +
cos q
x o 2
The potential at origin O = fo;
f, potential at Q = fo
f
x
r
cos q
2
Energy of dipole = Q (f+ f)

= Q E r cos q
= p E
Figure A13.2 Energy of an electric dipole in a uniform electric field.
+Q
F = QE
r
2
r
2 sin q
F = QE
r
2 sin q
r
2
Q
Figure A13.3 Torque on an electric dipole in a uniform electric field.

T D QEr sin
T DpE
453
454
The Raman Effect
Introducing eq. (A13.3.1) we have the alternative form

O D
ei ri E 0
H
A13.4.10
As a molecular dipole arises from the displacement of the electrons, ei D e and hence
O can also be written as
H

O De
H
ri E 0
A13.4.11
i
O unless care is taken to ascertain the precise

Confusion can arise about the sign of H
O given in the literature.
definition of the symbol used for charge in expressions for H
A13.5 BASIC TREATMENT OF PERMANENT AND INDUCED

MOLECULAR DIPOLES IN A UNIFORM STATIC
ELECTRIC FIELD
The electric dipole moment which appears in the expressions developed in the preceding
section for the interaction energy and hamiltonian of a system of charges in a uniform
static electric field will, in general, be the sum of two contributions: any permanent electric
dipole; and the electric dipole induced by the electric field, which at this stage, we take
to be static. For neutral molecules the induced electric dipole is the only contribution.
Provided the interaction between the molecule and the electric field is weak, this situation
can be analyzed by expanding the interaction energy U in a Taylor series about the energy
in the absence of the field. Thus

U
U[E 0 ] D U0 C E 0
E 0 0

1
2 U
C E 0 E 0
2
E 0 E 0
0

3
1
U
C E 0 E 0 E 0
6
E 0 E 0 E 0
0

4
1
U
C E 0 E 0 E 0 E
0
C ...
24
E 0 E 0 E 0 E
0
0
A13.5.1
The electric field itself, E 0 is taken at the molecular origin and U0 , U/E 0 . . .
etc. represent the energy, the derivative of the energy with respect to the E component
of the field, etc., evaluated for zero field strength at the molecular origin.
Using eq. (A13.4.5) and setting Q 0 D U0 we also have
U D U0 p E 0
A13.5.2
455
Thus we may readily deduce that

p D
U
E 0
A13.5.3
If we now differentiate eq. (A13.5.1) with respect to E we obtain for the component
of the total dipole moment, both permanent and induced, of the molecule in the presence
of a static electric field the following result:
1
p D pper
C E 0 C 2 E 0 E 0
C 16
E 0 E 0 E
0 C . . .
where
pper

U
D
E 0 0

2 U
D
E 0 E 0
0

3
U
D
E 0 E 0 E 0

4 U
D
E E 0 E 0 E
0
A13.5.4
A13.5.5
A13.5.6
A13.5.7
A13.5.8
It follows that we can rewrite eq. (A13.5.1) in the form

1
1
UE0 0 D U0 pper
E 0 E 0 E 0 E 0 E 0 E 0
2
6
1

E 0 E 0 E 0 E
0 . . .
A13.5.9
24
per
where p is the component of the permanent electric dipole vector p per and is
a coefficient which relates the component of E0 to the component of p. In general,
nine such coefficients are required to relate the three components of the vector E0 to the
three components of the vector p. The nine coefficients are the components of the
electric polarizability tensor a which is a second-rank tensor.
Similarly, is a coefficient which relates the and components of E0 to the
component of p. In general 27 such components are needed to relate the nine dyads of the
electric field components to the three electric dipole components. The 27 coefficients are
the components of the electric hyperpolarizability tensor b, which is a third-rank tensor.
Likewise
is one of the 81 components of the fourth-rank tensor g which relates
the 27 triads of the electric field components to the three electric dipole components.
Using these definitions eqs. (A13.5.4) and (A13.5.9) may be written in tensor form as
p D p C a E C 1 b : EE C 1 g ... EEE C . . .
A13.5.10
0
456
The Raman Effect
and
.
U D U0 p E 12 a : EE 16 b ... EEE C g ... EEEE
A13.5.11
.
The symbols , :, ... and ... represent single, double, triple and quadruple dot products,
respectively, and are explained in Chapter A10 which deals with the general properties
of tensors. The terms in eq. (A13.5.11) assume simple forms in certain special cases.
For example, when a has spherical symmetry so that xx D yy D zz D then the term
1
1
2
2 a : EE reduces to 2 E , where is the one component required to describe the polarizability.
The specific properties of the tensors a, b and g are of fundamental importance in the
treatment of light scattering phenomena. The polarizability tensor a is treated in detail in
Chapter A.14. We shall not consider the tensors b and g in any detail because they are not
involved in Rayleigh and Raman scattering. The hyperpolarizability tensor b is involved
in hyper Rayleigh and hyper Raman scattering; and the second hyperpolarizability tensor
g in second hyper Rayleigh and hyper Raman scattering. These types of scattering are
described briefly in Chapter 2.
Since, in the SI system the units of p are C m and of E are V m1 , it follows that
the SI units of the components of the several tensors are given by the general formula
C Vn mnC1 where n D 1 for the components of , n D 2 for the components of b and so
on. Thus the units for components of a are C V1 m2 ; for components of b are C V2 m3
and for components of are C V3 m4 . Typical orders of magnitude are: 1040 C V1 m2
for components of a, 1050 C V2 m3 for components of b, and 1061 C V3 m4 for
components of g.
It is convenient to separate and label the various types of contributions to the total
electric dipole p that arise in eq. (A13.5.10). We first separate off p per , the permanent
electric dipole, by writing eq. (A13.5.10) in the form
p D p per C p ind
A13.5.12
where p ind , the total induced electric dipole, is given by

p ind D a E C 12 b : EE C 16 g ... EEE C . . .
A13.5.13
The induced electric dipole plays a determining role in light scattering.

It should be noted that some authors choose to define the induced electric dipole through
the relationship
p ind D 40 fa E C 12 b : EE C 16 g ... EEE C . . .g
A13.5.14
Since, in the SI system 0 has the units C V1 m1 , it follows that with this definition of
p ind the units of components of a are m3 , the units of b are m4 V1 and the units of g are
m5 V2 . On this basis the units for components of a, at least, are agreeably simple and
Since J D V C an alternative form of this general formula is (C m)nC1 Jn . Then the units of components
of a are C2 m2 J1 , of components of b are C3 m3 J2 , and of components of are C4 m4 J3 .
457
also conceptually useful, since polarizability may be correlated with molecular volume.
Historically the polarizability itself has often been quoted as having units of volume,
3 ), but the correct unit is 40 m3 D 1.113 1010 C V1 m2 , (or
namely m3 (or cm3 or A
3 D 1.113 1040 C V1 m2 ). However
40 cm3 D 1.113 1016 C V1 m2 or 40 A
we shall adhere strictly to SI units as defined earlier.
The contributions to p ind may be further separated into p L , the contribution which is
linear in E, and p NL the sum of the contributions which are non-linear in E. Thus we
may write
A13.5.15
p ind D p L C p NL
where
and
pL D a E
A13.5.16
p NL D 12 b EE C 16 g ... EEE C . . .
A13.5.17
Alternatively, we may label the contributions to p ind according to the power of E involved
and write
p ind D p 1 C p 2 C p 3 C . . .
A13.5.18
where
p 1 D a E D p L
A13.5.19
1
2b
1
6g
A13.5.20
2
3
: EE
... EEE
A13.5.21
We note that the induced dipole p ind , which involves the nth power of the electric field
strength is associated with a tensor of rank R D n C 1.
A13.6 BASIC TREATMENT OF MACROSCOPIC POLARIZATION

AND ELECTRIC SUSCEPTIBILITIES
So far our approach has been a microscopic one in which we have been concerned
with the electric dipoles, permanent or induced, of individual molecules. We now take a
macroscopic approach and consider the electric polarization of an assembly of molecules
and how this is determined by the electrical properties of individual molecules.
Throughout this book our treatment is largely confined to an assembly of identical,
non-interacting, freely rotating molecules, that is, to an ideal gas at low pressure. Under
these conditions the local electric field acting on each molecule in the assembly is the
same as the macroscopic electric field; the electrical properties of a molecule are not
modified by molecular interactions; and orientation effects can be dealt with by classical
averaging. These conditions will be assumed to be operative in the treatment that follows.
458
The Raman Effect
We begin by introducing the electric polarization P . For the special conditions given
above we define P through the relation
N
A13.6.1
P D hp ind i
V
where hp ind i is the isotropically averaged induced electric dipole moment per molecule
and N is the number of molecules in the volume V. The quotient N/V is thus a number
density or number of molecules per unit volume. It follows from this definition that P
has units of C m2 and is isotropic.
If we now separate the induced electric dipole moment into its linear and non-linear
contributions, using eq. (A13.5.15), then eq. (A13.6.1) becomes
N
A13.6.2
P D fhp L i C hp NL ig
V
Alternatively, using eq. (A13.5.18), with eq. (A13.6.1.) we have
N
P D fhp 1 i C hp 2 i C hp 3 i C . . .g
A13.6.3
V
Introducing eqs. (A13.5.19) to (A13.5.21) into eq. (A13.6.3) we obtain
N
P D fhai E C 12 hbi : EE C 16 hgi ... EEE C . . .g
A13.6.4
V
We recall that we are considering a static electric field only at this stage.
We can therefore consider the total electric polarization to be the sum of electric
polarizations of the first order, second order etc. We can thus write
P D P 1 C P 2 C P 3 C P N
A13.6.5
and set
P 1 D 0 e1 E
P
2
3
2
1
2 0 e
3
1
6 0 e
A13.6.6
: EE
... EEE
A13.6.7
A13.6.8
where the numerical factors are n!1 .

It follows from a comparison of eqs. (A13.6.6) to (A13.6.8) with eq. (A13.6.4) that
N hai
e1 D
A13.6.9
V 0
N hbi
e2 D
A13.6.10
V 0
N h gi
A13.6.11
e3 D
V 0
Sometimes called the dielectric polarization or dipole moment per unit volume.
In general the macroscopic electric polarization is not isotropic. A more general definition of P is given
in Chapter A14.
459
The quantity e1 is termed the first-order electric susceptibility and is dimensionless;
2 is the second-order electric susceptibility with units V1 m; 3 is the third-order
electric susceptibility with units V2 m2 . The nth order electric susceptibility en would
have units V1 mn1 . As J D VC, an equivalent form is (CJ1 m)n1 . The units for 2
are then CJ1 m, and for 3 are C2 J2 m2 . We may define n as follows:
en D
N hgn i
V 0
A13.6.12
if we regard a, b and g as the historic symbols for g1 , g2 and g3 .
The electric susceptibilities can be seen to characterize a dielectric medium in the same
way as the polarizability and hyperpolarizabilities characterize a molecule.
It is important to appreciate that, unfortunately, the above definitions of P and n are
by no means universally used although they are in accord with IUPAC recommendations.
For example, the numerical factor n!1 is often not included in definitions of P n . Its
inclusion is however desirable because then the definitions of p and P have the same form
and the factor n!1 does not appear in the relationships between n and gn . Further,
the permittivity of a vacuum, 0 , traditionally, has been omitted from the definition of the
non-linear electric polarizations. Its inclusion is logical and puts all the P n on the same
basis. In addition, the units for the electric susceptibilities are simplified and are given
by the same general formula throughout, namely V1 mn1 . The hybrid quantity, the
n
, defined as n /N is also used. The reader is therefore
microscopic susceptibility, micro
warned to pay special attention to the definitions used by other authors when consulting
the literature, especially as some authors add to the possibility of confusion by folding
0 into their definitions of n .
We now consider the isotropic averages which occur in the definitions of 1 , 2
and 3 given by eqs (A13.6.9) to (A13.6.11). As we have assumed the ideal gas model
we are dealing, as already emphasized, with an isotropic system. For an isotropic system
hai is given by the isotropic part of the polarizability, the mean polarizability a of a
molecule which is always non-zero; hence 1 is always non-zero. However, for an
isotropic system, hbi is zero irrespective of the values of the molecular hyperpolarizability
components and hence 2 is zero; thus there is no second-order electric polarization. The
third-order electric polarization is always non-zero since hgi is non-zero. In general, for
an isotropic system P n is zero for n even and non-zero for n odd. This is a consequence
of the properties of hcosn i which determines the isotropic averages of gn ; hcosn i is
zero for n even and non-zero for n odd (see Chapter A4).
In accord with IUPAP and IUPAC recommendations the subscript e is added to to give e as the
symbol for electric susceptibility. The magnetic susceptibility is, however, denoted by , the symbol m
being reserved for molar magnetic susceptibility.
This matter is further considered when dealing with the electric displacement D (see the next
Section, A13.7).
Folding appears to be a euphemism for writing n and meaning 0 n .
460
The Raman Effect
A13.7 BASIC TREATMENT OF THE ELECTRIC DISPLACEMENT

FOR A UNIFORM STATIC ELECTRIC FIELD
We now introduce the electric displacement D (unit: C m2 ) which is defined as
D D 0 E C P
A13.7.1
where P is the electric polarization of the medium. Any further development of

eq. (A13.7.1) requires a knowledge of the constitution of the medium and its response
to an electric field. For this reason eq. (A13.7.1) is usually referred to as a constitutive
relation.
In vacuum P D 0 and thus we have the very simple constitutive relationship.
D D 0 E
A13.7.2
We now consider several cases of media with specific properties. We assume throughout
that the media are homogeneous and isotropic; and, as we are considering a basic treatment, we ignore the magnetic dipole and electric quadrupole contributions to P at this
stage.
In a material medium which is linear in its response to an electric field we have, as
discussed in Section A13.6,
A13.7.3
P L D P 1 D 0 e1 E
Introducing eq. (A13.7.3) into eq. (A13.7.1) we obtain
D D 0 1 C e1 E
A13.7.4
D 0 1 C e1
A13.7.5
If we now introduce , defined by
the constitutive relationship eq. (A13.7.4) can be written in the very simple form
D D E
A13.7.6
in which is the permittivity of the medium (unit: F m1 or C V1 m1 ). e1 , the linear
electric susceptibility is dimensionless. We now introduce, the dimensionless quantity r ,
the relative permittivity, defined as
r D
A13.7.7
It then follows from eq. (A13.7.5) that

e1 D r 1
Formerly called the dielectric constant.
A13.7.8
461
Equation (A13.7.4) can be expressed in an alternative form using r , namely

D D r 0 E
A13.7.9
In a material medium which is non-linear in its response to an electric field, the electric
polarization term P in eq. (A13.7.1) is given by
P D P L C P NL
A13.7.10
Thus the general constitutive relation for a non-linear medium is

D D E C P NL
A13.7.11
where the linear polarization is subsumed in and P NL involves the higher order electric
susceptibilities 2 , 3 , . . . n which were discussed in Section A13.6.
A13.8 THE IMPLICATIONS OF USING DYNAMIC ELECTRIC

FIELDS
The discussions in Sections A13.4 to A13.7 have all been in terms of a static electric
field. To complete the basic treatment we now discuss what changes, if any, are needed
in the case of a dynamic electric field associated with electromagnetic radiation.
We consider first the interaction energy and interaction hamiltonian for a system of
point charges at rest in a dynamic electric field. The detailed treatment of this problem is
relatively complicated and the arguments leading to its solution will not be rehearsed here.
A good summary with references to detailed accounts of special points has been given by
Barron (1982). The eventual result is, however, reassuringly simple. The dynamic interaction energy and hamiltonian can be cast in exactly the same forms as we have established
for the static case. Thus the dynamic interaction energy is given by eqs. (A13.4.5) to
(A13.4.8) and the dynamic interaction hamiltonian by eqs. (A13.4.9) to (A13.4.11). The
O in terms of the electric dipole, for the dynamic field case, is not
formulation of U and H
the only one possible, but it is certainly the most appealing, because its physical significance is readily grasped and it is easy to apply. The dynamic interaction hamiltonian is of
considerable importance. It can be used to calculate molecular wave functions perturbed
by the dynamic electric field of electromagnetic radiation and hence expectation values
of the transition electric dipole moment. The dynamic tensors a, b and g can then be
identified in the resulting power series in the electric field strength.
We now consider whether equations of the type (A13.5.4), (A13.5.10), (A13.6.4) and
(A13.6.6) to (A13.6.8) also hold for the dynamic fields of electromagnetic radiation. This
question requires careful analysis and necessarily involves knowledge of the properties of
electromagnetic waves. Again we refer the reader to Barron (1982) for an account of the
arguments which show that the results we have obtained for static fields are also valid
for the special case of monochromatic plane electromagnetic waves. Happily, in most
experimental situations, such electromagnetic waves are closely approximated to, and so
formulae derived on this basis are widely applicable.
462
The Raman Effect
A13.9 MORE GENERAL TREATMENT OF ENERGY OF

INTERACTION OF POINT CHARGES AT REST IN A
STATIC ELECTRIC FIELD
As indicated in Section A13.1 we now present a more general treatment of the interaction
of a system of electric charges with a static electric field. In this treatment the electric
field is not assumed to be constant over the system of charges, that is the electric field
gradient is non-zero. As a consequence the energy of a system of charges in the electric
field will now involve not only electric dipoles but also higher order electric multipoles
namely electric quadrupoles, electric octopoles and so on. We shall only consider electric
quadrupoles as the contributions from electric octopoles are insignificant in the light
scattering phenomena we consider.
The electric quadrupole is the second moment of a set of charges. One definition is in
terms of the tensor Q, whose component Q is defined, in cartesian tensor notation, as

ei ri ri
A13.9.1
Q D
i
where ri and ri are the and cartesian coordinates, respectively, of the ith charge.
An alternative definition involves the tensor , whose component , is defined
in cartesian tensor notation by
1
ei 3ri ri ri2 D 12 3Q Q
A13.9.2
D
2 i
where is the Kronecker delta. The corresponding tensor form is
1
D
ei 3ri ri ri2 I
2 i
A13.9.3
where ri2 is the scalar product ri ri D xi2 C yi2 C zi2 and I is the symmetric unit tensor
ex ex C ey ey C ez ez .
It follows from these definitions that is a symmetric second-rank tensor, that is
D
A13.9.4
D xx C yy C zz D 0
A13.9.5
and has zero trace, so that
Thus has five independent components as shown below:
xx
xy
xz
.
yy
yz
D yx D xy
zx D xz zy D yz zz D xx yy
A13.9.6
It is always possible to find three mutually perpendicular axes such that D 0 if
6D . Then, there are three principal quadrupole moments xx , yy and zz , of which
only two are independent because the trace is zero as indicated in eq. (A13.9.5).
463
The electric quadrupole moment is only independent of the choice of origin if both
the monopole moment, or net charge, and the dipole moment are zero. If the net charge
is zero but the dipole moment is not, a centre of dipole exists relative to which
D 0.
The traceless definition of eq. (A13.9.6) is preferable for the situations treated here.
With this definition if the charge distribution is symmetric about the z axis, then
xx D yy D 12 zz and only one quadrupole is needed to specify the interaction of
an axially symmetric charge system with an arbitrary electric field. For a spherical charge
distribution the electric quadrupole moment vanishes.
An electric quadrupole is more difficult to visualize than an electric dipole. It is perhaps
helpful to appreciate that the quadrupole moment of a system with zero dipole moment
can be thought of as arising from a separation of equal and opposite dipoles, the magnitude of the quadrupole being proportioned to the product of the dipole moment and the
separation.
The SI unit for components of Q and is C m2 , and a typical order of magnitude is
40
C m2 .
10
The higher-order electric multipoles play a role in determining the energy of interaction
U of a distribution of charges with an external electric field when second-and higher-order
terms are included in the Taylor series expansion of the potential produced by the external
electric field. Instead of eq. (A13.4.3) we now write

1

2
ei 0 C
ri C
ri ri
UD
r 0 2 r r
i
0

1
3
C
ri ri ri C
A13.9.7
6 r r r
0
We then develop this equation as follows. We make the same replacements for the first
two terms as we made in proceeding from eq. (A13.4.3) to eq. (A13.4.5); we neglect the
third derivative; we use eq. (A13.9.1) to introduce Q ; and we recognize that

2
E
0
D E 0 D
A13.9.8
r r
r 0
0
where E0 0 is the component of the electric field gradient at the origin O. Then
eq. (A13.9.7) becomes
U D Q 0 p E 0 12 Q E0 0
A13.9.9
By introducing the traceless tensor , eq. (A13.9.9) can be written in the alternative form
U D Q 0 p E 0 13 E0 0
A13.9.10
The neglected third derivative in eq. (A13.9.7) would give rise to a term involving the
octopole moment.
For a charge distribution which is symmetric about the z axis, the interaction energy
has the simple form
A13.9.11
U D Q 0 pz Ez 0 12 zz E0zz 0
464
The Raman Effect
It can be seen from eqs. (A13.9.9) to (A13.9.11) that the energy of a charge system in an
electrostatic field is the sum of the energies of a point charge in a potential, a dipole in a
uniform field and a quadrupole in a field gradient. If the third derivative in eq. (A13.9.7)
had been kept, there would have been an additional contribution from the energy of an
octopole in the gradient of a field gradient.
In pursuit of a more general treatment we must now include the electric field gradient
in the Taylor series for U[E 0 ]. Thus, instead of eq. (A13.5.1) we write

U
2 U
1
C E 0 E 0
U[E 0 ] D U0 C E 0
E 0 0 2
E 0 E 0
0

3
1
U
C E 0 E 0 E 0
6
E 0 E 0 E 0
0

4
1
U
C E 0 E 0 E 0 E
0
24
E E 0 E 0 E
0
0

U
C E 0
E 0 0

1
2 U
C E 0 E
0
C
A13.9.12
2
E 0 E
0
0
Here the electric field itself, E 0 , is taken at the molecular origin and U0 and the
derivatives

U
U
,
. . . etc.
E 0 0 E 0
represent respectively the energy and derivatives of the energy with respect to the electric field, electric field gradient etc., evaluated for zero field strength at the molecular
origin. We note that this expansion consists of two types of derivatives: those involving
only the electric field components; and those involving only the electric field gradient
components. Mixed derivatives involving electric field and electric field gradient components have not been included. The implications of this are discussed at the end of this
section.
It follows readily from eq. (A13.9.10) that
p D
U
E 0
and
D 3
U
E 0
A13.9.13
A13.9.14
If we now differentiate eq. (A13.9.12) with respect to E , ignoring the last two terms
involving the electric field gradient we obtain, of course the result given by eq. A13.5.4.
465
Similarly, differentiation of eq. (A13.9.12) with respect to the field gradient E , gives
, the component of the total quadrupole moment, permanent and induced, of the
molecule:
per
D
C C,
E
0
A13.9.15
where
per

C,

U
D 3
E 0 0

2 U
D 3
E 0 E
0
A13.9.16
A13.9.17
0
per
is the component of the permanent quadrupole moment; C,

is a component
of the field gradient quadrupole polarizability C and describes the electric quadrupole
induced by the field gradient. The tensor C is symmetric in , and in ,
; and C,
is unaffected by exchanging the pairs and

. The SI unit for components of C is
CV1 m4 or JV2 m4 .
It follows from the above considerations that we can rewrite eq. (A13.9.12) in the
following form
1
1
U[E 0 ] D U0 pper
E 0 E 0 E 0 E 0 E 0 E 0
2
6
1

E 0 E 0 E 0 E
0
24
1 per
1
E 0 C,
E 0 E
0 . . .
A13.9.18
3
6
As indicated earlier, the above treatment has neglected the following types of derivatives.

2 U
3 U
and
E 0 E 0
E 0 E 0 E
0
0
The former relates to the electric dipole moment induced by the electric field gradient
or the electric quadrupole moment induced by a uniform electric field; and the latter to
the electric dipole moment induced by the electric field and the electric field gradient
or the electric quadrupole moment induced by the second power of the electric field.
When these are included, eq. (A13.9.18) would contain additional terms of the type
13 A, E E and 16 B,
E E E
; where A, and B,
are components of new
tensors A and B respectively, with corresponding additional terms in eq. (A13.5.4) for
p and eq. (A13.9.15) for .
466
The Raman Effect
A13.10 INTERACTION OF CHARGES IN MOTION WITH A

STATIC MAGNETIC FIELD
We have so far treated charges at rest. We now consider charges in uniform motion, that
is steady currents, and their interaction with a static magnetic field. These interactions are,
at best, of second order relative to those of static charges with static electric fields and are
significant only in a limited number of scattering phenomena, as for example scattering
by a chiral system. In consequence, the treatment given here will be less detailed than that
given for static charges and usually the relevant formulae will be quoted without proof.
The procedure we shall use parallels that employed for static charges, and we shall find
that most of the formulae have the same general form as in the case of static charges.
The distribution of circulating currents can be described in terms of magnetic multipole
moments. Magnetic monopoles have not been observed. The first moment of a circulating
current distribution is the magnetic dipole moment. Magnetic quadrupoles can also exist
but they will not be considered here as their contributions to the phenomena under discussion are never significant. Thus we can limit the treatment to a uniform magnetic field.
Also, contributions to the induced magnetic dipole which are non-linear in the magnetic
field need only be considered in outline as, generally, they are unimportant.
In what follows we consider only electrons and ignore nuclear contributions. On this
basis there are, in general, two contributions to the magnetic moment of a molecule: the
orbital angular momentum and the spin of an electron. In the absence of an external
magnetic field the magnetic dipole moment m arising from the electrons is given by
e
li C ge si
A13.10.1
mD
2me i
where li is the orbital angular momentum of the ith electron which is given by
li D ri pi
A13.10.2
Here pi is the linear momentum of the ith electron whose position relative to the origin
is defined by ri ; si is the spin angular momentum of the ith electron; and e, me and ge
are the charge, mass and Lande factor of an electron. For a free electron ge D 2.0023.
B
li C ge si
A13.10.3
mD
h
i
where B , is the Bohr magneton defined by
B D
eh
2me
A13.10.4
The value of the Bohr magneton is 9.274 1024 J T1 . The magnetic moment m is
independent of the choice of origin only in the absence of a time-dependent electric dipole
moment. The units of m are A m2 and typical values are of the order of 1023 A m2 .
See section A13.2, p. 450.
467
The potential energy U of a system of currents in a uniform static magnetic field, with
magnetic induction B is given by
d
U D m B 12
B B
A13.10.5
d is the component

where m is the component of the magnetic dipole moment,
of the diamagnetic susceptibility tensor and summation over repeated Greek subscripts is
d is defined as follows:
implied.
d

D
e2
ri ri ri2
4m
e
i
A13.10.6
d B
This can be considered as generating a magnetic field induced magnetic moment 12

which opposes the inducing field. Equation (A13.10.5) for the magnetic potential energy
is given here without proof. It has some correspondence with eq. (A13.9.10) for the
electric potential energy. However, there are important differences. There is no magnetic
monopole term; it contains only dipole interactions because its derivation is based upon a
uniform magnetic field; and there is a diamagnetic contribution to the interaction energy
which has no electrical analogue. It follows straightforwardly from eq. (A13.10.5) that
O m for a static magnetic field is given by
the interaction hamiltonian H
O m D m B 1 d B B
H
2
A13.10.7
The potential energy U of a molecule in a uniform static magnetic field can also be
expressed as a Taylor series expansion in the magnetic induction as follows:

U
U[B 0 ] D U0 C B 0
B 0 0

1
2 U
C B 0 B 0
C ...
A13.10.8
2
B 0 B 0
where B 0 , is the component of B 0 , the magnetic induction at the molecular origin;
and U0 , and the derivatives [U/B 0 ]0 etc. denote the potential energy and its derivatives with respect to the field evaluated for zero field strength at the molecular origin.
Equation (A13.10.8) parallels eq. (A13.5.1) for the potential energy of a molecule in a
uniform static electric field.
If we differentiate eq. (A13.10.5) with respect to B we obtain

U
d

D m C
B 0 D m0
A13.10.9
B 0
where m0 is the component of the magnetic dipole moment including the diamagnetic
contribution. Then using eqs. (A13.10.9) and (A13.10.8) we can write the molecular
magnetic dipole moment in the presence of a static uniform magnetic field as
m0 D m0 C B 0
A13.10.10
468
The Raman Effect

where
m0
U
D
B 0

and
D
p

d

A13.10.11
0
2 U
D
B 0 B 0
A13.10.12
Here m0 is the component of the permanent magnetic dipole moment; is a
p
component of the magnetic susceptibility tensor 4 and is the sum of the temperature
d
independent paramagnetic susceptibility and the diamagnetic susceptibility compop
d has
nents. is the magnetic analogue of the electric polarizability component ;
no electrical analogue as we have already indicated.
It is useful to rewrite eq. (A13.10.10) in tensor notation as
m 0 D m0 C 4 B C . . .
A13.10.13
The second term in this equation is the linear induced contribution to m 0 for which it is
convenient to introduce the symbol m L . Thus
mL D 4 B
A13.10.14
As already indicated, non-linear induced contributions to m are also possible but they are
of much less significance than the non-linear induced contributions to p.
The specific properties of the tensor 4 are of importance in the treatment of some lightscattering phenomena, as for example in magnetic Rayleigh and Raman scattering. Such
phenomena are considered later. At this point we consider just the units and magnitudes
of components of 4.
In the SI system the units of m are J T1 or A m2 and of B are T, Wbm2 or
V m2 s. Thus the units of d as defined in eqs. (A13.10.5) or (A13.10.6) and defined
in eq. (A13.10.12) are V1 A s1 m4 which may be recast in the form J T2 or H1 m4
since H D V A1 s.
So far we have taken a microscopic approach and considered individual molecules.
We now take a macroscopic approach and consider the magnetization of an assembly of
molecules. To this end we introduce the magnetization M. Following the treatment given
in Section A13.6 for the electric polarization of an ideal gas at low pressure we define
M as
N
A13.10.15
M D hmi
V
where N is the number of the molecules in the volume V and hmi is the isotropically
averaged induced magnetic moment per molecule. It follows that M is the magnetic
dipole per unit volume and has units of A m1 . Then combining eq. (A13.10.14) with
eq. (A13.10.15) we obtain
N
M L D h4i B
A13.10.16
V
469
where M L is the linear magnetization. For isotropic materials h4i is a scalar quantity and
hence so is the linear magnetization
We now introduce the magnetic field strength H of a medium which is defined as
H D
B
M
0
A13.10.17
where 0 is the permeability of vacuum and M is the magnetization of the medium

defined in eq. (A13.10.15). The units of H are A m1 . Equation (A13.10.17) is another
constitutive relationship as it involves a knowledge of the constitution of the medium and
its response to a magnetic flux.
In vacuum M D 0 and the constitutive relationship is then simply
H D
B0
0
A13.10.18
For a medium which is homogeneous, isotropic and linear in its response to magnetic
flux we can define the relationship between M L the linear magnetization, and H as
M L D H
A13.10.19
where is called the magnetic susceptibility . This is a scalar quantity; it is dimensionless.

For non-isotropic materials it is necessary to introduce a magnetic susceptibility tensor of
second rank; and for non-linear materials it is necessary to consider higher rank magnetic
susceptibility tensors associated with H n where n D 2, 3, . . . .
Introducing eq. (A13.10.19) into eq. (A13.10.17) and rearranging we have
B D 0 1 C H
A13.10.20
We now introduce , the permeability, defined by

D 0 1 C
A13.10.21
and then eq. (A13.10.20) can be rewritten as

B D H
A13.10.22
Thus the constitutive relationship also has a simple form for a linear medium, the linear
magnetization being subsumed in .
If we now introduce the relative permeability r , defined as

r D
A13.10.23
0
it follows from eq. (A13.10.21) that
D r 1
A13.10.24
Many texts use m for the magnetic susceptibility but this symbol should now be reserved for the molar
magnetic susceptibility (unit: m3 mol1 ). Thus m D Vm where Vm is the molar volume.
470
The Raman Effect
Equation (A13.10.22) can be expressed in an alternative form using r , namely

B D r 0 H
A13.10.25
If we now introduce eq. (A13.10.22) into eq. (A13.10.14) and compare the result with
eq. (A13.10.19) we see that for isotropic and linear materials
N
D h4i
V
When non-linear polarization has to be taken into account we must write
M D M L C M NL
A13.10.26
A13.10.27
where M NL is the non-linear macroscopic polarization whose dependence on H is left

unspecified and M L is as defined in eq. (A13.10.16). It follows from eq. (A13.10.17) that
B D 0 H C 0 M L C 0 M NL
A13.10.28
B D 0 1 C H C 0 M NL
A13.10.29
or
This can be rearranged to give a general constitutive relationship
H D
M NL
B

r
A13.10.30
Thus far we have considered a static magnetic field. For the case of the dynamic
magnetic field associated with electromagnetic radiation, the situation is the same as
that which obtains for static and dynamic electric fields. In a dynamic magnetic field,
the interaction energy and the interaction hamiltonian of a system of charges in motion
can be cast in the same forms as we have established for a static magnetic field. Thus
the dynamic interaction energy is given by eq. (A13.10.5) and the dynamic interaction
hamiltonian by eq. (A13.10.7). The reader is again referred to Barron (1982) for an
account of the arguments justifying this. In addition it can be shown that equations of
the type (A13.10.7) also hold for the dynamic magnetic field of electromagnetic radiation
provided monochromatic plane electromagnetic waves are being considered.
REFERENCE
Press: Cambridge.
471
A14
The Polarizability Tensor
One by one, or all at once.
Lewis Carroll
A14.1 INTRODUCTION
We now consider those properties of the polarizability tensor which are important for
Rayleigh and Raman scattering. Since the treatment of particular aspects of these lightscattering phenomena can often be simplified by a judicious choice of the bases for the
tensors, we shall consider three bases for these tensors: the cartesian basis, the spherical
basis and the irreducible spherical basis.
It follows from eq. (A13.5.19) that the relation between p1
0 , the amplitude of a firstorder time-dependent induced electric dipole, and the amplitude of the electric field
strength of the electromagnetic radiation which produces it has the linear from
p1
0 D a E0
A14.1.1
where a is a second-rank tensor to which we shall give the general name polarizability
tensor. In the several theoretical treatments of first-order light scattering considered in
the main text, p1
0 may be either a classical electric dipole amplitude or a transition
electric dipole amplitude and a the appropriate corresponding polarizability tensor, as for
example an equilibrium polarizability tensor, a derived polarizability tensor, a transition
polarizability tensor and so on.
We therefore present the treatment in terms of a general polarizability tensor a and
its components; and further, to maintain complete generality, we shall include the cases
where p1
0 , a and E0 are complex. It is a straightforward matter to adapt the general
472
The Raman Effect
formulae to specific cases. It is merely necessary to change the nomenclature and take
account of any specific symmetry. For example, if a is replaced by afi the formulae
become those for a general transition polarizability tensor associated with a transition
electric dipole amplitude p1
0 fi , and if then the anti-symmetric terms are put equal to
zero the formulae reduce to those for a symmetric transition polarizability tensor, and
so on.
A14.2 THE POLARIZABILITY TENSOR IN THE CARTESIAN

BASIS
As discussed in Chapter A10, a second-rank tensor can be represented by a 3 3 matrix
of its nine dyad components of the form given by eq. (A10.2.1); and the linear operation
of the tensor on a vector, by a matrix equation of the form given by eq. (A10.2.2). In the
cartesian basis the dyad components and the components of the matrix have a one-to-one
correspondence, but this is not always the case for other bases.
We shall represent the polarizability tensor in the cartesian basis by c a and the corresponding matrix by [c a] where the label c indicates that the basis is the cartesian one.
When it is useful to simplify the nomenclature and no ambiguity can arise, we shall usually
omit the basis label. The components of this tensor and of the corresponding matrix will
be denoted in general by , and specifically by xx , xy . . . etc. These subscripts are
unique to the cartesian basis and no further label is therefore required where components
are concerned.
It is useful to introduce at this stage two vectors V and W, the dyads of whose components in the cartesian basis are the components of the tensor c a in accordance with
eq. (A10.2.1). Using the basis and component labels given above, we have
c
a D cV cW
A14.2.1
where VW is a direct vector product formed with the vector components in the cartesian
basis. Thus
D V W
A14.2.2
with
xx D Vx Wx ;
xy D Vx Wy
etc.
A14.2.3
The component form of eq. (A14.1.1) in the cartesian basis is

p1
0 D E0
A14.2.4
with defined by eq. (A14.2.2)

At this point the physical significance of the vectors V and W need not concern us.
However, we note that the quantum mechanical treatment of light scattering will show
that they are related to transition electric dipole moments of the scattering molecule.
473
A14.2.2 Reduction of the tensor

The reduction of a general second-rank tensor to a set of three irreducible tensors has
been treated in Chapter A10. For a, a general second-rank tensor in the cartesian basis,
we can write
c
a D c a0 C c a1 C c a2
A14.2.5
where c a0 is a diagonal tensor, c a1 an anti-symmetric traceless tensor and c a2 a
symmetric traceless tensor. The procedures for obtaining the irreducible tensors is dealt
with later, (see Section A14.5) and here we shall merely state without proof the general
definitions of the irreducible tensors c a0 , c a1 and c a2 and then consider some special
cases which arise when c a has some symmetry.
General definitions
ca
xx
yx
zx
xy
yy
zy
xz
yz
zz
A14.2.6
with
Trfc ag D xx C yy C zz
A14.2.7
It is convenient to introduce at this point the mean polarizability a, defined by

a D 13 xx C yy C zz D 13 Trfc ag
A14.2.8
As the trace of a matrix is invariant under rotation of the axis system, the mean polarizability is an invariant of the tensor. In general, a second-rank tensor has two other
invariants making a total of three in all. The invariants of the polarizability tensor are
treated in detail in Section A14.7.
c a0 : a 0 0
0 a 0
A14.2.9
0 0 a
with
Trfc a0 g D 3a

c a1
0

xy yx

2

xz zx

2
xy yx
2
A14.2.10
0
yz zy
2

xz zx
2

yz zy
2
A14.2.11
with
Trfc a1 g D 0
A14.2.12
474

c a2
xx a

yx C xy
2

zx C xz
2
xy C yx
2
yy a
zy C yz
2
The Raman Effect

xz C zx
2

yz C zy
2
A14.2.13
zz a
with
Trfc a2 g D 0
A14.2.14
The components of c a0 , c a1 and c a2 can be given general definitions as follows
1
0
D 3 D a
A14.2.15
1
1
D 2
A14.2.16
1
2
A14.2.17
2
C a
It should be noted that because of the summation convention for repeated Greek subscripts.
D xx C yy C zz D 3a
A14.2.18
When c a has some symmetry, these general definitions reduce to simpler forms. Some
special cases are considered below.
(a) c a symmetric, i.e. xy D yx etc.
c 0
A14.2.19
: as A14.2.9
Trfc a0 g D 3a
c 1
: vanishes
c 2
: xx a
xy
xy
yy a
xz
yz
c 2
Trf a g D 0
a
a
xz
yz
zz a
(b) c a diagonal, i.e. xy D 0 etc.

c 0
A14.2.20
: as A14.2.9
Trfc a0 g D 3a
c 1
: vanishes
c 2
: xx a
0
0
yy a
0
0
c 2
Trf a g D 0
a
a
0
0
zz a
475
(c) c a symmetric and a D 0
A14.2.21
c 0
: vanishes
c 1
: vanishes
c 2
: xx xy xz
xy yy yz
xz yz zz
Trfc 2 g D 0
a
a
a
(d) c a diagonal xy D 0 etc. and xx D yy D zz D a

c 0
A14.2.22
: as A14.2.25
Trfc a0 g D 3xx
c 1
: vanishes
c 2
: vanishes
a
a
A14.2.3 The polarizability ellipsoid

The polarizability tensor may be represented graphically by using a polarizability ellipsoid.
For simplicity we shall confine our attention to the symmetric real polarizability tensor
with the following array of components:
xx
xy
xz
xy
yy
yz
zx
zy
zz
A14.2.23
The equation
xx x 2 C yy y 2 C zz z2 C 2xy xy C 2yz yz C 2zx zx D 1
A14.2.24
is the equation of a polarizability ellipsoid which has its centre at the origin of the
coordinate system but whose axes are not coincident with the axes of the coordinate
system. Such a polarizability ellipsoid is depicted in Fig. 14.1(a). It can be shown that
a line drawn from the origin to a point on the surface of the ellipsoid with coordinates
1/2
x, y, z has a length equal to E , where E is the polarizability in that direction. This
means that, if an electric field E is applied along this direction then pE , the component
in this direction of the induced electric dipole p, is given by
pE D E jE j
A14.2.25
Along the principal axes a, b and c of the ellipsoid, and only along these directions, the
direction of p is the same as the direction of E. Thus, for these directions, we may write
p a D a Ea
p b D b Eb
p c D c Ec
where a , b , c are the polarizabilities along the principal axes a, b, and c.
A14.2.26
476
The Raman Effect

Z
z
y
x
(a)
(b)
Figure A14.1 The polarizability ellipsoid: (a) principal axes not coincident with system
axes; (b) principal axes coincident with system axes.
Consequently, if the coordinate system is rotated in such a way that the new cartesian
axes X, Y, Z coincide with the principal axes of the polarizability ellipsoid, then the
polarizability tensor defined in relation to the X, Y, Z axes takes on the much simpler
diagonal form in which XY D YZ D ZX D 0.
In this new coordinate system, the equation (A14.2.26) becomes
pX D XX EX
pY D YY EY
A14.2.27
pZ D ZZ EZ
and the array of polarizability tensor components is now
0
0
XX
0
A14.2.28
0
YY
0
0
ZZ
where XX , YY and ZZ are called the principal values of the polarizability. The equation
for the polarizability ellipsoid becomes
XX X2 C YY Y2 C ZZ Z2 D 1
1/2
1/2
XX , YY
A14.2.29
1/2
ZZ .
and
It should be noted that
and its semi-minor axes have the values
this means that the largest axis of the ellipsoid corresponds to the smallest principal value
of the polarizability, and vice versa. Two special cases should be noted. If two of the
principal values of the polarizability are equal, say XX D YY and XX < ZZ , then the
polarizability ellipsoid is a solid of revolution generated by rotating about the Z axis the
ellipse defined by
A14.2.30
XX X2 C 2ZZ Y D 1
Such an ellipsoid is shown in Fig. A14.1(b). The quantity D jzz xx j can be regarded
as a measure of the anisotropy of the polarizability. The concept of anisotropy can be
477
extended to include any polarizability tensor but then a more general definition is needed.
This will be considered later in Section A14.7 where it will also be shown that is another
invariant of the polarizability tensor. If the three principal values of the polarizability are
1/2
1/2
1/2
equal, the polarizability ellipsoid becomes a sphere of radius XX D YY D ZZ . The
anisotropy is then zero as might be expected.
A14.2.4 Transformation of c a under rotation of axes

In Chapter A3 we explained how the relative orientation of two sets of cartesian axis
systems x, y, z and x 0 , y 0 , z0 with a common origin may be specified by using direction cosines of the type lx0 z where cos1 lx0 z is the angle between the x 0 and z axes. In
Chapter A10 we showed how such direction cosines may be used to relate the components of a second-rank tensor in the primed axis system x 0 , y 0 , z0 to its components in the
unprimed axis system x, y, z. Restating eq. (A10.3.7) we have, in the condensed notation
of the Einstein summation convention,
0 0 D l0 l 0
where 0 , 0 D x 0 , y 0 , z0 and , D x, y, z or, alternatively,

lx0 x ly 0 y xy
x 0 y 0 D
A14.2.31
A14.2.32
x,y,z
We recall that these summations involve nine equations (one for each of the new tensor
components 0 0 ), each consisting in general of nine terms (one for each of the old
components ). Thus the representation of the transformation of c a under rotation can
be rather cumbersome because it can mix all the to give one 0 0 . We shall see
in Section A14.6 that the transformation of a polarizability tensor under rotation can be
expressed more simply when irreducible bases are used.
A14.3 THE POLARIZABILITY TENSOR IN THE SPHERICAL

BASIS
A14.3.1 General definitions
We consider first the representation of the polarizability tensor by dyads of the spherical
components of the two vectors V and W which we introduced in Section A14.2.1. From
the treatment of vectors in the spherical coordinate system given in Chapter A9 it follows
that for the vector s V a general spherical component is denoted by V with D 1, 0, 1;
and similarly for the vector s W . A general dyad of spherical components is thus V W .
The 3 3 array of these nine dyads constitutes a representation of the polarizability tensor
in the spherical system which we shall denote by s a. The symbol s a is used to distinguish
this tensor from another tensor s a which will be introduced shortly. Thus
s
a D sV sW
A14.3.1
478
The Raman Effect
where s V s W is a direct vector product formed with the components of the vectors V and
W expressed in the spherical system.
We now consider the tensor s a as a linear operator acting on the electric field strength
vector, defined as s E0 in the spherical system, to give the electric dipole moment vector,
defined as sp0 in the spherical system. Using the definitions of a second-rank tensor as a
linear operator given in eq. (A10.1.2) we may write
s
p0 D s a s E0 D s V s W s E
A14.3.2
Equation (A9.4.13) enables us to express s W s E in component form as follows:

s
W s E D 1 W E0 D 1 W E0
A14.3.3
p0 D 1 V W E0 D 1 V W E0
A14.3.4
Thus
Identifying the dyad V W (or V W ) as a component of the polarizability tensor
(or ) eq. (A14.3.4) gives
p0 D 1 E0
A14.3.5
p0 D 1 E0
A14.3.6
or
It is convenient to have the components of sp00 and s E0 in the same order. When
and both follow the same ordered pattern of values, 1, 0, 1, this is achieved using
eq. (A14.3.6) and we shall use this form henceforth when relating p0 and E0 .
It is useful to write eq. (A14.3.2) in matrix form using eq. (A14.3.6). We then obtain
p10
11
E10
1 11 10
p00 D
01 00
01 E00
A14.3.7
p10
11 10 11
E10
We now introduce a new tensor c a whose components are related to the as
follows:
D 1
A14.3.8
It follows from the identification of with the dyad V W that
D 1 V W
A14.3.9
This definition of the as signed vector dyads is a necessary consequence of the
definitions of the spherical components of a vector given in eq. (A9.4.1) which use the
Condon and Shortley phase factor.
D V W
A14.3.10
This results from eq. (A9.4.4) which relates the complex spherical components and their
complex conjugates.
479
It follows that the analogue of the

p10
11
p00 D 01
p10
11
matrix Equation (A14.3.7) is
10
11
E10
00
01 E00
10 11
E10
A14.3.11
The corresponding component form is

p0 D E0
A14.3.12
We shall use eq. (A14.3.11) and eq. (A14.3.12) with the components defined as in
eq. (A14.3.9) in all treatments of the polarizability tensor involving the spherical system.
Especial care is necessary to ensure that the consequences of the phase-related factor
1 are carried through in all subsequent developments. Also, those encountering the
spherical coordinate basis for the first time will need to accustom themselves to the special
pattern of the subscripts on the polarizability components. Section A14.3.2 considers
such matters in some detail. It should be noted that in the literature other definitions are
sometimes used for the polarizability tensor in the spherical coordinate basis.
Because the are in general complex, eq. (A9.4.4) leads to the following general
relationship between tensor components and their conjugate complexes
D 1jjjj
A14.3.13
The specific relationships between the and the are
11 D 11
and
11 D 11 etc.
00 D 00
A14.3.14
01 D 01 and
01 D 01
10 D 10 and
10 D 10
11 D 11 and 11 D 11
Thus if D 0 and jj 6D 0, or D 0 and jj 6D 0 there is a change of sign when forming
the conjugate complex. For all other combinations of and including D 0 and D 0
there is no change of sign.
A14.3.2 Reduction of the tensor s a

For the reduction of s a to a set of three irreducible tensors (see Chapter A10) we can
write
s
a D s a0 C s a1 C s a2
A14.3.15
These irreducible tensors have as their basis vectors the nine dyads of the spherical unit
vectors e1 , e0 and e1 , just as the irreducible tensors associated with c a have as their
basis vectors the nine dyads of the cartesian unit vectors ex , ey and ez .
We now give without proof the definitions of these tensors and some of their properties
sa
11
01
11
10
00
10
11
01
11
A14.3.16
480
The Raman Effect
with
Trfs ag D 11 C 00 C 11 D 3a
A14.3.17
where a is the mean polarizability defined by

a D 13 11 C 00 C 11
A14.3.18
A general diagonal term of s a is (i.e. D ) and a general off-diagonal term

(i.e. 6D )
a0 :
a
0
0
0 0
a 0
0 a
A14.3.19
with
Trfs a0 g D 3a
s a1
1
2 11
1
2 01
1
2 10
11
1
2 10
C 01
1
2 01
C 10
A14.3.20
C 01
1
2 11
A14.3.21
C 10
11
with
Trfs a1 g D 0
s a2
1
2 11 C 11
1
2 01 10
11
1
2 10
01
11
1
2 01
00 a
1
2 10
A14.3.22
01
1
2 11
10
A14.3.23
C 11 a
with
Trfa2 g D 0
A14.3.24
The tensor s a1 is anti-hermitian and the tensor s a2 hermitian as may be established
using the relations given in eq. (A14.3.14).
If the tensor s a is symmetric then the anti-symmetric tensor s a1 is zero, and it follows
from (A14.3.21) that
11 D 11
10 D 01
10 D 01
A14.3.25
Using these relations the tensor s a2 assumes the following simpler form which is of
course hermitian
s a2
11 a
01
11
10
00 a
10
11
01
11 a
A14.3.26
481
A14.3.3 Transformation of s a under rotation of axes

By analogy with the treatment given in Section A14.2.4 for c a, the relationship between
a component of a second rank tensor s a in the rotated primed axis system 0 , 0 , and its
components in the original unprimed system , , is given, using the Einstein summation
convention, by
A14.3.27
0 0 D l0 l0
or equivalently by
0 0 D
l0 l0
A14.3.28
,
The l0 , l0 are direction cosines in the spherical system. Their definitions and properties are considered in Chapter A9. Just as for the tensor c a, the transformation under
rotation, in general, mixes all the components of in forming a given 0 0 .
A14.4 THE RELATION BETWEEN THE alm AND THE ars

We now consider how the are related to the . Utilizing eq. (A9.4.1) we can
express a dyad V W in terms of the dyads V W . Then using eq. (A14.3.9) and recognizing the V W as polarizability tensor components in accord with eq. (A14.2.2),
the relationships of the to the follow. For example, for 11 we have
11 D 11 V1 W1
D 12 fVx C iVy Wx iWy g
D 12 fxx C yy ixy yx g
A14.4.1
and similarly for the other .

It is convenient to express the relation between the and the in a form in which
the and the are regarded as components of tensorial sets which are represented
by column vectors, each of nine rows. We can then write

D
B,
A14.4.2
, some of which are complex, are components of a 9 9 matrix we

where the B,
have chosen to define as B . This transformation matrix will be found to be unitary and
so the inverse transformation is given by

D
B,
A14.4.3

where the B, are components of the matrix B which is the inverse of B . The
components of the transformation matrix B are given in Table A14.1 together with the
procedure for obtaining the components of B .
482
The Raman Effect
The matrix B is given below. Its components

B
A14.5 IRREDUCIBLE POLARIZABILITY TENSORS AND THEIR

COMPONENTS
The general properties of irreducible tensors and irreducible tensorial sets have been
treated in Chapter A10. We now consider how irreducible tensorial sets and irreducible
tensors can be obtained for the specific case of the polarizability tensor.
We start with the spherical tensor components which transform under rotations as
the product of two angular momentum states j1i and j1 i. Thus the procedures
described in Chapter A19 for finding combined angular momentum states with simple
transformation properties under rotation may also be used to find tensor components with
corresponding transformation properties. These tensor components are irreducible tensor
483
j
components and will be denoted by the symbol m . The corresponding tensor will be
denoted by IR a.
It follows from the angular momentum combination rules that the permitted values of
j are as follows: j D 2 (with five possible values of m), j D 1 (with three possible values
of m) and j D 0 (with one possible value of m), making a total of nine tensor components
j
m formed from combinations of the nine .
j
We may express the relationships between the and the m in a form in which
j
the and the m are regarded as components of tensorial sets which are represented
by column vectors each of nine rows. We can write

D
C,jm j
A14.5.1
m
j,m
where the C,jm , all of which will be found to be real, are components of a 9 9
matrix which we define as C. This matrix is orthogonal and so the inverse transformation
is given by CQ . Thus

Cjm,
A14.5.2
j
m D

The coefficients C,jm and Cjm, are found from the appropriate ClebschGordan or
Wigner 3-j symbols as described in Chapter A19. The matrix C is given in Table A14.2
together with the procedure for obtaining C 1 .
j
We now consider the transformations which relate the and the m . Treating them
as components of tensorial sets as already explained, we express the transformations in
the forms

D
A,jm j
A14.5.3
m
jm
and
j
m D
Ajm,
A14.5.4
The A,jm , some of which are complex, are components of a 9 9 matrix which we define
as A. This matrix will be found to be unitary so that its inverse is A with components
Ajm, .
It follows from eqs (A14.4.3) and (A14.5.1) that

A,jm D
B, C,jm
A14.5.5

Using compact matrix notation we can write

A D BC
A14.5.6
484
The Raman Effect
The matrix C is given below. Its components

C
and hence
A1 D A D C 1 B 1 D CQ B
A14.5.7
In component form we have

Ajm, D
Cjm, B,
A14.5.8

The matrix A is given in Table A14.3 together with the procedure for obtaining
j
A . The inter-relations of the transformations between the , and m are
presented diagrammatically in Fig. A14.2. The reader may verify these relationships using
Tables A14.1 to A14.3.
485
The matrix A is given below. Its components

A
x
ars
=B
=A
C=
(j)
am
al u
=C
Figure A14.2 The inter-relationships of , and j

m .
486
The Raman Effect
A14.6 TRANSFORMATION PROPERTIES OF THE ajm UNDER

ROTATIONS
j
We now consider the transformation properties of the m under rotations. We define the
j
irreducible spherical components referred to space-fixed axes by m and those referred
j
to molecule-fixed axes by m0 . The molecule-fixed axis system is defined relative to the
space-fixed axis system by the Euler angles ,
, . The two sets of components are
related as follows:
j j
m 0 Dm 0 m
A14.6.1
j
m D
m0
j
j
where Dm0 m is a shorthand notation for Dm0 m ,

, , an element of the angular momentum
transformation matrices for an axis rotation defined by the Euler angles ,
, and . It
can be seen from eq. (A14.6.1) that a rotation of axes mixes only irreducible spherical
components with the same j. It is this transformation property that makes the irreducible
spherical basis so valuable. The transformation properties under rotation of the cartesianbased and the spherical coordinate-based are much less simple, as we have seen.
The D transformations satisfy the following orthonormal relationships
j j
Dm0 m Dm0 m d
1

A14.6.2
D
jj mm m0 m0
d
2j C 1
where

d D sin
d
d d
A14.6.3
d D 8 2
A14.6.4
and the limits of the integrations are

, 0 ! 2 ;
, 0 ! ; , 0 ! 2
A14.6.5
j
m
The property of the

expressed by eq. (A14.6.1) enables us to calculate the isotropic
averages of quadratic products of polarizability tensor components in any basis. These
isotropic averages are of fundamental importance for the calculation of the intensity
of light scattering from a system of freely rotating, non-interacting molecules, and are
considered in detail in Section A14.7, next.
A14.7 ISOTROPIC AVERAGES AND TENSOR INVARIANTS

In a light-scattering experiment we measure intensities of scattering relative to space-fixed
axes (often referred to as laboratory axes). These intensities are determined, inter alia by
the quadratic products of polarizability tensor components relative to these axes; and as
The term quadratic product is used for complete generality so that, for example in the cartesian basis,
terms such as jxx yy j are included as well as pure square terms of the types jxx j2 and jxy j2 .
487
we are considering freely rotating non-interacting molecules we must average over all
orientations of the molecule relative to the space-fixed axes. In this section we evaluate
the isotropic averages of quadratic products of polarizability tensor components in terms
of rotational invariants of the polarizability tensor.
The most elegant and effective method of obtaining these isotropic averages is to
j
work initially with the irreducible tensor components m and then make the appropriate
transformations to give the isotropic averages associated with the and . This
j
procedure enables us to take advantage of the simple transformation properties of the m .
A14.7.2 Isotropic averages and rotational invariants G j in terms of

the amj
We begin by considering a general quadratic product involving irreducible tensor components, namely
j
j
m m
A14.7.1
where j and m are, in general, different from j and m. Using eq. (A14.6.1) we may write
j
j
m m

m0
j
j
j
j
m m0 , Dmm0 Dm m0
A14.7.2
m0
This expression relates to a single scattering molecule with an orientation relative to the
space-fixed axes defined by the Euler angles ,
, . We now average eq. (A14.7.2)
over all orientations of all the molecules. By using the orthonormal relationships given
by eq. (A14.6.2) we obtain
j0
hj
m m i D jj mm
1 j 2
j 0 j
2j C 1 m0 m
A14.7.3
where h i indicates an isotropic average.

The Kronecker delta restrictions enable us to rewrite eq. (A14.7.3) in the form
2
jj
m j D
1 j 2
j 0 j
2j C 1 m0 m
A14.7.4
for each m. This result is independent of m, and hence when we sum over the 2j C 1
values of m we are summing 2j C 1 equal terms. Thus

j
2
jj
j
D
jm0 j2
A14.7.5
m
m
m0
j
Clearly for each of the three values of j, m m is a true rotational invariant; that is,
it is independent of the orientation of the axis system. We label these three invariants
488
The Raman Effect
G j j D 0, 1, 2. They are defined as follows:

0 2
2
G 0 D
j0
m j D j0 j
A14.7.6
G 1 D
2
1 2
1 2
1 2
j1
m j D j1 j C j0 j C j1 j
A14.7.7
2 2
2 2
2
2 2
2 2
2
j2
m j D j2 j C j1 j C j0 j C j1 j C j2 j
A14.7.8
G 2 D

m
It is useful to introduce a further Placzek type invariant, G , defined as

G D G 0 C G 1 C G 2
A14.7.9
It is worth emphasizing that, because of the Kronecker delta restrictions, isotropic averages of quadratic products involving two polarizability tensor components with different
j values (that is from different irreducible tensors) are zero.
Invariants of this type were first introduced by Placzek (1934) who used a slighlty
different notation writing G0 for G 0 , Ga for G 1 and GS for G 2
However another set of invariants, a2 , 2 and 2 which have simple relations to the
Placzek invariants G j are more commonly used. These are introduced and defined in
section A14.7.5. Some other forms of the three invariants associated with the polarizability
tensor are also given there.

the ars
We can use the unitary transformation A defined in eq. (A14.5.3) to obtain the isotropic
averages of quadratic products of components of the cartesian based polarizability tensor
j
in terms of the G j and hence in terms of the m . It follows from eq. (A14.5.3) that,
for the most general form of the isotropic average, we have

j
hj 0 0 ji D
A,jm A0 0 jm j
A14.7.10
m m
j,m
It is convenient to write eq. (A14.7.10) in the form

j
hj 0 0 ji D
A ,0 0 G j
A14.7.11
where
2
jj
m j
A14.7.12
1
jA,jm A0 0 ,jm j
2j C 1 m
A14.7.13
1
jA,jm j2
2j C 1 m
A14.7.14
G j D
and
j
A ,0 0 D
When D 0 and D 0
A j
, D
489
j
Values of A ,0 0 are readily calculated from the A,jm . Values of three non-zero terms
j
j
j
of the types A xx,xx , A xy,xy and A xx,yy are given in Table A14.4 for j D 0, 1 and 2.
Using Table A14.4 with eqs. (A14.7.10) to (A14.7.14) the isotropic averages
hj 0 0 ji can be expressed in terms of the Placzek invariants G j . The results are
summarized in Table A14.5. Note that G 1 contributes only to isotropic averages of
the form hjxy j2 i. If the polarizability tensor is symmetric there is no contribution from
G 1 .
Additional isotropic averages can arise if the tensor is anti-symmetric and pure imaginary but such a tensor will not be considered here.
Table A14.5 The isotropic averages hj 0 0 ji in terms

of the Placzek invariants G j .
1 0
G C
3
1
hjxy j2 i D hjyz j2 i D hjzx j2 i
= G 1 C
6
1 0
hjxx yy ji D hjyy zz ji D hjzz xx ji = G
3
hjxx j2 i D hjyy j2 i D hjzz j2 i
2 2
G
15
1 2
G
10
1 2
G
15
All other isotropic averages are zero.
A useful rule is that all isotropic averages which involve a subscript once, such as jxx xy j, are zero.
490
The Raman Effect

the alm
We may obtain the isotropic averages of quadratic products of components of the
polarizability tensor in the spherical coordinate basis in a similar way. It follows from
eq. (A14.5.1) that for the most general form of isotropic average we have

j
hj 0 0 ji D
C,jm C0 0 ,jm j
A14.7.15
m m
We note that in this basis also, we have to consider quadratic products with 6D 0 and
6D 0 .
We now write eq. (A14.7.15) in the form
j
C ,0 0 G j
A14.7.16
hj , 0 0 ji D
j
where
C
j
,0 0
1
jC,jm C0 0 ,jm j
2j C 1 m
and G j is given by eq. (A14.7.12).

When D 0 and D 0 ,
C
j
,
A14.7.17
1
jC,jm j2
2j C 1 m
A14.7.18
j
Values of the C ,0 0 are readily calculated from the C,jm . A number of combinaj
tions of , 0 0 give rise to non-zero isotropic averages. Values of C ,0 0 are
given in Table A14.6 for j D 0, 1 and 2 for those combinations which arise in this book.
The corresponding isotropic averages h 0 0 i expressed in terms of the Placzek
invariants G j are given in Table A14.7. It can be seen that only G 2 contributes to all
isotropic averages; G 0 contributes only to those isotropic averages which involve diagonal entries in the s a tensor; and G 1 contributes only if the s a tensor is not symmetric.
A14.7.5 Isotropic averages and the rotational invariants G j , a, d and

g for the cartesian basis
j
The Placzek invariants G j which have been defined in terms of the m in eqs. (A14.7.6)
to (A14.7.8) can be expressed in terms of the by using eq. (A14.5.4). The results are
as follows
G 0 D 13 fjxx C yy C zz j2 g
G
G
D 12 fjxy yx j2 C jxz zx j2 C jyz zy j2 g

2
D 12 fjxy C yx j2 C jzx C xz j2 C jyz C zy j2 g
C 13 fjxx yy j2 C jxx zz j2 C jyy zz j2 g
1
These definitions have an appealing symmetry.
A14.7.19
A14.7.20
A14.7.21
491
of
the form
with
}
}
}
}
Table A14.7 The coefficients of the Placzek invariants
G j contributinga to isotropic averages of the general form
hj j2 i.
G 0
G 1
hj11 j2 i D hj11 j2 i
hj j2 i
hj00 j2 i
1
3
1
3
1
6
1
6
1
6
0
G 2
1
5
1
10
1
10
1
30
2
15
1
For example hj11 j2 i D 13 G 0 C 16 G 1 C 30
G 2 . These relations
2
may be recast in terms of the invariants a , 2 and 2 using
eqs. (A14.7.22) to (A14.7.24).
a
492
The Raman Effect
We now introduce a slightly different set of invariants of the polarizability tensor.

This set which is in common usage consists of the mean polarizability a, the antisymmetric anisotropy and the anisotropy . Their relationships to the Placzek G j are very
simple:
a2 D 13 G 0
2
D
2 D
A14.7.22
3 1
2G
3 2
2G
A14.7.23
A14.7.24
and hence
G D 3a2 C 23 2 C 23 2
A14.7.25
As already indicated on page 488 other definitions of the polarizability tensor invariants
have been used. These are summarized in Table A14.8.
A summary of alternative expressions for isotropic averages in the cartesian basis in
terms of tensor invariants is given in Table A14.9 for a real tensor.
We now consider the definitions of a2 , 2 and 2 in terms of the for both general
and special cases.
(a) The mean polarizability a
General definition
a D 13 xx C yy C zz
A14.7.26
Cylindrical symmetry
about z axis xx D yy
a D 13 2xx C zz D 13 2yy C zz
A14.7.27
Spherical symmetry
xx D yy D zz
a D xx D yy D zz
A14.7.28
Table A14.8 The relations of a2 , 2 and 2 to other polarizability tensor invariants. The entries
in each row are all equal to each other.
3a2
2 2
3
2 2

3
Placzek
(1934)
Sushchinskii
(1972)
Nestor and
Spiro
(1974)

Go G 0
32

Ga G 1

Gs G 2
2
2 2

3 as
2 2

3 s
Mortensen and
Hassing
(1980)
2fsc
0
4 2
2fsc
1
4 2
2fsc
2
4 2
The notation used in this work is given in brackets. It differs slightly from that used originally by Placzek
32 D 13 xx C yy C zz 2
2
fsc is the fine structure constant, which is dimensionless and has the value 7.297 103 . See Chapter 5,
Section 5.10. It arises because Mortensen and Hassing use the reduced transition tensor fi . (See Chapter 4,
Section 4.13).
493
Table A14.9 Summary of alternative expressions for isotropic averages in the cartesian
basis in terms of tensor invariants. The tensor is taken to be real.
Isotropic average
h2xx i
h2xy ia
hxx yy i
h2xy ianti
h2xx i C h2xy i
h2xx i h2xy i
h2xx i 3h2xy i
h2xx i C 3h2xy i
a2 , 2 , 2
,
G 0 , G 1 , G 2
45a2 C 4 2
45
2
15
2
45a 2 2
45
2
9
2
45a C 7 2
45
45a2 C 2
45
45a2 5 2
45
45a2 C 13 2
45
1
C 2
15
1
3
30
1
2
15
1
anti
6
1
7 C
30
1
3 C
30
5

30
1
13
30
1

9
3
anti
2
1
3
3
2
1 0
G C G 2
3
15
1 2
G
10
1
1 0
G G 2
3
15
1 1
G
6
7
1 0
G C G 2
3
30
1
1 0
G C G 2
3
30
1 0 1 2
G G
3
6
1 0 13 2
G C G
3
30
1 0
G
3
3 1
G
2
3 2
G
2
a2
2
2
a
For a symmetric tensor. The antisymmetric tensor contribution is treated separately and labelled
h2xy ianti . It involves the tensor invariant 2 .
(b) The antisymmetric anisotropy

General definition
2 D 34 fjxy yx j2 C jyz zy j2 C jzx xz j2 g
A14.7.29
Symmetric tensor
D 0
A14.7.30
(c) The anisotropy

General definition
2 D 12 fjxx yy j2 C jyy zz j2 C jzz xx j2 g

C 34 fjxy C yx j2 C jxz C zx j2 C jyz C zy j2 g
A14.7.31
494
The Raman Effect

2 D 12 fjxx yy j2 C jyy zz j2 C jzz xx j2 g
Symmetric tensor
C 3fjxy j2 C jyz j2 C jzx j2 g
A14.7.32
Symmetric tensor and

cylindrical symmetry
about z axis xx D yy ,
2 D jzz xx j2
xy D yz D zx D 0
A14.7.33
Spherical symmetry
A14.7.34
D0
The invariants a, and may also be defined using the summation convention. For
the most general case of a complex tensor we then have
a2 D 19
A14.7.35
2 D 32 0
A14.7.36
2 D 12 3
A14.7.37
and
where is the component of the complex transition polarizability tensor and

is the component of its conjugate complex. The components 0 relate to the antisymmetric tensor c 1 . The reader may further test his knowledge of the summation
convention by verifying that these condensed formulae yield the expanded forms given
earlier. Expressions for isotropic averages in terms of these definitions of the invariants
are included in Table A14.9.
A14.7.6 Isotropic averages and rotational invariants, G j , a, d and g

for the spherical basis
j
The Placzek invariants G j which were defined in terms of the m in eqs (A14.7.6)
to (A14.7.8) can be expressed in terms of the by using eq. (A14.5.2) The results
are as follows:
G 0 D 13 j11 C 11 C 00 j2
A14.7.38
G 1 D 12 fj10 C 01 j2 C j11 11 j2 C j10 C 01 j2 g

G
2
D j11 j C j11 j C
1
2 j10
01 j C
C 16 j 11 C 200 11 j2
1
2 j10
A14.7.39
2
01 j
A14.7.40
The alternative invariants a, and may be expressed in terms of the using their
relationships to the G j given by eqs. (14.7.22) to (A14.7.24). The results are given
below for the general case and for selected special cases.
495
(a) The mean polarizability a

General case
a D 13 00 C 11 C 11
A14.7.41
a D 13 00 C 211 D 13 00 C 211
A14.7.42

about z axis
11 D 11
(b) The antisymmetric anisotropy

3
General case 2 D fj10 C 01 j2 C j11 11 j2 C j10 C 01 j2 g
4
A14.7.43
Symmetric tensor
2 D 0
A14.7.44
(c) The anisotropy

General case
3
1
2 D fj11 j2 C j11 j2 C j10 01 j2
2
2
1
3
2
C j10 01 j C j00 aj2 g
2
2
A14.7.45
Symmetric tensor
D 3j11 j C 6j10 j C j11 00 j
A14.7.46
2 D j11 00 j2
A14.7.47

about z axis
11 D 10 D 0
REFERENCES
Mortensen, O. S. and Hassing, S. (1980). Advances in Infrared and Raman Spectroscopy, (eds.),
R. J. H. Clark and R. E. Hester, 6, 1. Heyden: London.
Nestor, J. and Spiro, T. G. (1974). J. Raman Spectrosc. 1, 539.
Placzek, G. (1934). Rayleigh-Streuung und Raman-Effekt, in Handbuch der Radiologie, (ed.),
E. Marx, 6, 205374. Academische Verlag: Leipzig.
Sushchinskii, M. M. (1972). Raman Spectra of Molecules and Crystals, Israel Program for Scientific Translations, Keter: New York.
497
A15
The Optical Activity Tensors, G,
, A and
What immortal hand or eye
Could frame thy fearful symmetry?
William Blake
A15.1 INTRODUCTION
In this appendix we consider the isotropic averages of the various types of products of
the tensor components which arise in the treatment of Rayleigh and Raman scattering
by chiral systems given in Chapter 10. We are concerned principally with those isotropic
averages which we represent symbolically as being of the forms haG 0 i and haAi. For
real wave functions a denotes a real electric dipole transition polarizability tensor, A
a real electric dipole-electric quadrupole optical activity transition tensor, and G 0 is the
imaginary part of G the electric dipolemagnetic dipole optical activity transition tensor.
The isotropic averages that arise for optically active Rayleigh scattering, and for optically
active Raman scattering when certain special assumptions are made, are treated in some
detail. Subsequently, the isotropic averages that occur in a more general treatment of
optically active Raman scattering are considered briefly.
A15.2 ISOTROPIC AVERAGES OF THE TYPE haG 0 i

The tensors a and G 0 are both second-rank tensors but, whereas a is a polar tensor, G 0 is an
axial tensor. We now introduce transition-defining subscripts and note that for optically
Chapter 10, Section 10.2 treats these tensors in detail.

Chapter A10, Section A10.3 deals with axial and polar second-rank tensors.
498
The Raman Effect
active Rayleigh scattering, symmetry conditions require that the same components of
aii and G 0 ii span the totally symmetric representation; for optically active Raman
scattering the same components of avf vi and G 0 vf vi must simultaneously span the
irreducible representation to which the associated normal coordinate of vibration belongs.
Thus we are concerned only with products of the types xx ii G0xx ii , xy ii G0xy ii ,
xx vf vi G0xx vf vi , and xy vf vi G0xy vf vi . We recall that it is only in the pure rotation
groups Cn , Dn , T, O, I (the so-called chiral point groups) that polar and axial tensors
have the same transformation properties.
For convenience in the developments that follow we normally omit the transitiondefining labels on the transition tensors and their components and write simply a, G 0 ,
xy , G0xy and so on.
As G 0 transforms in the same way as a the treatment of isotropic averages of the type
haG 0 i follows that for isotropic averages of the type haai.
We first introduce G0 the isotropic invariant of the second-rank tensor G 0 which is
defined as follows:
G0 D 13 G0xx C G0yy C G0zz
A15.2.1
This is analogous to a the isotropic invariant of the tensor a which is given by
a D 13 xx C yy C zz
A15.2.2
We next introduce G2 0 , the invariant of the fourth rank tensor aG 0 . When the tensor a
is symmetric this is defined as follows:
G2 0 D 1/2f[xx yy G0xx G0yy C xx zz G0xx G0zz
C yy zz G0yy G0zz ] C 3[xy G0xy C G0yx
C xz G0xz C G0zx C yz G0yz C G0zy ]g
A15.2.3
When the tensor G 0 is symmetric, G0xy D G0yx , G0xz D G0zx and G0yz D G0zy and G2 0 is
then given by
G2 0 D 12 [xx yy G0xx G0yy C yy zz G0yy G0zz
C zz xx G0zz G0xx C 6xy G0xy C yz G0yz C xz G0xz ] A15.2.4
This is analogous to 2 , the invariant of the fourth-rank tensor aa which, when a is
symmetric, is given by

2 D 12 xx yy 2 C yy zz 2 C zz xx 2

C62xy C 2yz C 2xz
A15.2.5
Chapter 10, Section 10.5 deals with such symmetry considerations.
The Optical Activity Tensors, G, G , A and A

A
499
Using these additional tensor invariants the isotropic invariants of the types hxx G0xx i,
hxy G0xy i and hxx G0yy i can be shown to be
45aG0 C 4G2 0
45
2
0
hxy G0xy i D G
15
0 2 2
45aG
G0
hxx G0yy i D
45
hxx G0xx i D
A15.2.6
A15.2.7
A15.2.8
Certain linear combinations of the above isotropic averages occur in the intensity
formulae for various polarizations of the incident and scattered radiation and these are
listed below.
45aG0 C G2 0
45
45aG0 C 7G2 0
hxx G0xx i C hxy ihG0xy i D
45
0 C 13 2
45aG
G0
hxx G0xx i C 3hxy ihG0xy i D
45
hxx G0xx i hxy ihG0xy i D
A15.2.9
A15.2.10
A15.2.11
A15.3 ISOTROPIC AVERAGES OF THE TYPE haAi

The tensor A is a polar third-rank tensor and does not transform in the same way as
the second-rank axial tensor G 0 . However, components of A always occur together with
components of the second-rank tensor a in triple products of the form A, where
is a component of the third-rank unit alternating axial tensor e. As in writing A,
contraction with respect to two pairs of subscripts is specified, the contracted tensor which
results transforms as a second-rank axial tensor just like the electric dipolemagnetic
dipole tensor G 0 .
Using the properties of the unit alternating tensor the components of the contracted
tensor are readily found to be as follows:
Diagonal terms
Off-diagonal terms

A,

A,
xx
yy
zz
Ay,zx Az,yx
Ax,zy C Az,xy
Ax,yz Ay,xz
xy
yx
xz
zx
yz
zy
Ay,yz Az,yy
Ax,xz C Az,xx
Ay,zz Az,zy
Ax,xy Ay,xx
Ax,zz C Az,zx
Ax,yy Ay,yx
500
The Raman Effect
Because of the properties of the electric quadrupole operator defined in Chapter 10,
eq. (10.1.8), A, D A, and so the sum of the diagonal terms given above is zero.
Thus the contracted matrix is traceless and there is no invariant analogous to G0 given by
eq. (A15.2.1) or a given by eq. (A15.2.2).
However, there exists an invariant which we label A2 . This is analogous to G2 0 given by
eq. (A15.2.3) and 2 given by eq. (A15.2.5) and is an invariant of the fourth-rank tensor
formed from the dual product of the tensor a and the contracted tensor derived from the
tensor A as described above. When the tensor a is symmetric A2 is defined as follows:
A2 D 1/2fyy xx Az,xy C xx zz Ay,zx C zz yy Ax,yz
C xy Ay,yz Az,yy C Az,xx Ax,xz
C xz Ay,zz Az,zy C Ax,xy Ay,xx
C yz Az,zx Ax,zz C Ax,yy Ay,yx g
A15.3.1
Note that the frequency term is included in this definition.

The non-zero isotropic averages of the type haAi which arise in the intensity formulae
developed in Section 10.3 of Chapter 10 are given below in terms of A2 :
A2
A15.3.2
15
2
A15.3.3
hzy Az,zx i D A
15
2
hyx Ay,yz i D A
A15.3.4
15
All other isotropic averages of the type haAi that occur in section 10.3 are zero.
hzx Az,zy i D
A15.4 OTHER TENSOR INVARIANTS

The discussion so far has been confined to those five tensor invariants arising in the
treatment of Rayleigh and Raman scattering by chiral systems which forms the greater
part of Chapter 10. We recall that in this treatment the wave functions are taken to be real
and the transition tensors have no antisymmetric part. In addition, for Raman scattering
the assumptions underlying the Placzek polarizability theory are assumed to be applicable.
We now consider briefly the most general case when the wavefunctions are complex, the
transition tensors have symmetric and antisymmetric parts and the Placzek polarizability
theory is not invoked. There are now 26 tensor invariants to be considered. They are of
the following 13 types.
2A
a2 2 2 aG G2 2G A2
aG G 2 G 2 A 2 A
An alternative definition of A2 is used in some of the literature in which the numerical factor of 1/2 is
replaced by 3/2. The use of the definition given by eq. (A15.3.1) is to be preferred as it avoids some awkward
numerical factors in subsequent formulae.
The Optical Activity Tensors, G, G , A and A

A
501
One set of 13 such invariants is associated with the real parts, and a second set of
13 invariants with the imaginary parts, of the corresponding products of the transition
polarizability tensor and the optical activity transition tensor. The invariants a and 2
2
relate to the symmetric part of the tensor a. The invariants aG , G 2 and A
relate to the
symmetric parts of the tensors G and A
A and have definitions corresponding to those for
aG, G2 and A2 which relate to the tensors G and A. The invariant 2 is the invariant
for the antisymmetric part of the tensor a. The invariants 2G , G 2 , 2A and 2A are the
corresponding invariants for the antisymmetric parts of the tensors G, G , A and A
A.
We shall not give definitions of these invariants here. They are treated in detail by
Hecht and Nafie (1991).
REFERENCE
Hecht, L. and Nafie, L. A. (1991). Molecular Physics 72, 441.
Chapter A14, Section A14.7.5 defines 2 in the cartesian basis.
503
A16
Maxwell Equations in Vacuum
and in Media
I have succeeded in magnetizing and
electrifying a ray of light, and in illuminating
a magnetic line.
Michael Faraday
A16.1 THE FUNDAMENTAL EQUATIONS

The Maxwell equations are the fundamental equations of electromagnetism. They are
completely general and apply to all known macroscopic electromagnetic phenomena in
media which are at rest with respect to the coordinate system used. Their validity extends
to non-isotropic, non-homogeneous and non-linear media.
Because of the interrelation of the various quantities involved, there is no unique set
of Maxwell equations. As the starting point for the developments in this chapter we have
chosen the set given below. This is based on SI units.
D D
A16.1.1
B D0
A16.1.2
B
A16.1.3
t
D
CJ
A16.1.4
H D
t
These are partial differential equations involving the space derivatives of the four electromagnetic field vectors, E the electric field strength, B the magnetic induction, D the
E D
504
The Raman Effect
electric displacement and H the magnetic field strength, together with the time derivatives of B and D, and the two sources, the free charge density and J the electric current
density. The quantities E, B, D and H have been defined and discussed in Chapters A11,
A12 and A13. We note that is the free or unbound charge density (SI unit: C m3 ) and
excludes charges forming part of the structure of atoms and molecules which are internally compensated; and that J is the analogous current density (SI unit: A m2 ) which
includes both conduction and convection currents. In a conducting medium which obeys
Ohms law for points outside the sources, we have
J D E
A16.1.5
where is the conductivity of the medium (SI unit: S m1 or 1 m1 )
The Maxwell equations embrace earlier discoveries of electromagnetic phenomena.
Equation (A16.1.1) is a differential form of Gausss theorem applied to electrostatics
and eq. (A16.1.2) is the corresponding result for magnetostatics since there are no
magnetic monopoles. Equation (A16.1.3) is the Faraday law of magnetic induction and
eq. (A16.1.4) is a modified form of Ampères law where the displacement current density
D/t (SI unit: A m2 ), which arises when the electric displacement D varies with time, is
added to the conduction current J, the current flow due to the movement of free electric
charges. The need for inclusion of the displacement current density term was one of
Maxwells important contributions to the development of electromagnetic theory.
Two of the electric field vectors, E and B, can exist even in source-free space as is
evident from eqs (A16.1.2) and (A16.1.3). E and B are termed the fundamental fields and
form a representation of an electromagnetic field. The two other electric field vectors, D
and H, are termed the auxiliary fields and are introduced to describe the response of matter
to the electromagnetic fields E and B, respectively. It should be noted that, historically,
E and H were usually regarded as the fundamental fields and D and B as the auxiliary
fields. However, the choice of E and B as the fundamental fields is now mandatory for
reasons which are related to the electrodynamics of moving media, but need not concern
us here.
As the Maxwell equations are linear, each charge or current distribution produces its
own field independently; for example, the resultant electric field strength vector at a given
point is the vector sum of the E vectors produced at that point by the various sources.
This is an example of the principle of superposition.
The relationships between E and D, and between B and H depend on the constitution
of matter and are called the constitutive relationships. These relationships are discussed
in detail in Chapter A13 but for convenience are reproduced below in their general forms
with appropriate cross-references:
D D 0 E C P
see eq. (A13.7.1), and
A16.1.6
B
M
A16.1.7
0
see eq. (A13.10.17), where both P and M may in principle include non-linear as well as
linear terms.
H D
Maxwell Equations in Vacuum and in Media
505
Using eqs. (A16.1.6) and (A16.1.7) D and H can be eliminated from the Maxwell
equations which then take the following forms
1
P
0
B D0
B
D0
E C
t

P
E
D 0 J C
CM
B 0 0
t
t
E D
A16.1.8
A16.1.9
A16.1.10
A16.1.11
In these forms the Maxwell equations are still completely general, but they are expressed
in such a way as to stress the contribution of the medium. The presence of matter has the
effect of adding the following terms: P , the bound or polarization charge density;
P /t, the polarization current density; and M , the equivalent current density. The
significance of the definition of and J in eqs (A16.1.1) and (A16.1.4) can now be better
appreciated. The total charge density T is given by
T D P
A16.1.12
and the total current density JT by

JT D J C
P
CM
t
A16.1.13
In a vacuum, infinitely remote from matter, P D 0 and M D 0. D and H then have the
same directions as E and B, respectively, and their magnitudes are simply related through
the scalars 0 and 0 . However, in the presence of matter, the situation can be much
more complicated. For example the electric polarization P is a function of the electric
field strength, and this function may be linear or non-linear depending on the electric
field strength. Further, the medium may be anisotropic so that the direction of P may
not be the same as that of E, and the relationship between the vectors P and E becomes
tensorial. Also, the medium may be non-homogeneous so that its properties are a function
of position in the medium. In the case of oscillatory electric fields the tensor may also
be a function of the frequencies of the fields and the polarization. Similar considerations
apply to H and M.
It is evident that any further development of the Maxwell equations must involve
specific assumptions regarding the behaviour of matter. We shall now consider, in turn,
a number of special cases which are relevant for light scattering. Table A16.1 lists these
and summarizes the assumptions involved in each case. Throughout, we assume that the
medium is homogeneous, and isotropic. These are characteristic properties of the ideal
gas model upon which the theoretical treatment in this book is based.
In this chapter the development is concerned only with obtaining the appropriate
second-order differential equations for the field vectors. The solutions of some of these
differential equations are dealt with subsequently in Chapter A17.
yes
Equations
(A16.1.1) to
(A16.1.4)
IV
Conducting
yes
Equations
Non-conducting (A16.2.1) to
(A16.2.4)
III
yes
yes
Equations
(A16.2.1)
(A16.2.4)
Equations
Non-conducting (A16.2.1) to
(A16.2.4)
Vacuum
yes
yes
yes
yes
Homo- Isogeneous tropic
II
Medium
Case
Form of
Maxwell
equations
Linear
Non-linear
Linear
Linear
Linear
Linear
Linear
Linear
Response Response
to E
to B
Medium
6D 0 6D 0
Table A16.1 Maxwell equations in various media
H D
H D
D D E
H D
D D E C P NL H D
D D E
D D 0 E
B
J D E

B

B

B
0
Constitutive equations
A16.5
A16.4
A16.3
A16.2
Derivation
of wave
equation
A17.4
A17.2
Solution
of wave
equation
Chapter and section for
506
The Raman Effect
507
A16.2 CASE I: THE MAXWELL EQUATIONS IN VACUUM

In vacuum infinitely remote from matter, D 0 and D 0, and the Maxwell equations (A16.1.1) to (A16.1.4) are reduced to
D D0
A16.2.1
B D0
B
D0
E C
t
D
H
D0
t
A16.2.2
A16.2.3
A16.2.4
Also, since in vacuum P D 0 and M D 0 the constitutive relationships, given by

eqs. (A16.1.6) and (A16.1.7) assume the following simple forms
D D 0 E
A16.2.5
B
0
A16.2.6
and
H D
To find an equation for E we take the curl of eq. (A16.2.3) and obtain
E C
B
D0
t
A16.2.7
Then, using the vector identity

A D A r2 A
A16.2.8
which is valid only in cartesian coordinates, eq. (A16.2.7) becomes

E r 2 E C
B
D0
t
A16.2.9
Now it follows from eqs. (A16.2.5) and the Maxwell equation (A16.2.1) that in vacuum
E D0
A16.2.10
Also, using the Maxwell equation (A16.2.4), introducing eqs. (A16.2.5) and (A16.2.6)
and remembering that space operators are independent of time operators, we find that
B
2 E
D 0 0 2
t
t
A16.2.11
Thus, using this result and eq. (A16.2.10), eq. (A16.2.9) reduces to
r2 E D 0 0
2 E
t2
A16.2.12
508
The Raman Effect
Similarly, starting from eq. (A16.2.4) we find the corresponding equation for H,
r2 H D 0 0
2 H
t2
A16.2.13
It follows from eqs. (A16.2.5) and (A16.2.6) that the corresponding equations for D and
B have the same forms as eqs. (A16.2.12) and (A16.2.13), respectively. Thus
r2 D D 0 0
2 D
t2
A16.2.14
r2 B D 0 0
2 B
t2
A16.2.15
and
The four equations (A16.2.12) to (A16.2.15) are differential equations for an unattenuated
wave travelling through vacuum with a speed c0 given by
c0 D 0 0 1/2
A16.2.16
Thus the four field vectors E, D, B and H can propagate as waves in vacuum.
The quantity c0 is, of course, measurable and is found to have the value c0 D 2.998
108 m s1 . As the value of 0 is defined to be exactly 4 107 H m1 , 0 can be
calculated to be 8.854 1012 F m1 which coincides within experimental error with
the value determined by experiment. This agreement provides satisfying proof of the
validity of the Maxwell equations.
The solutions of the differential equations (A16.2.12) and (A16.2.15) and also the
properties of the electromagnetic waves which they represent are dealt with in detail in
Chapter A17.
A16.3 CASE II: THE MAXWELL EQUATIONS IN A LINEAR

MEDIUM WITH r D 0 AND s D 0
As D 0 and D 0 we can again use the special forms of the Maxwell equations
(A16.2.1) to (A16.2.4). However, as the medium now contains matter which behaves
linearly, the constitutive relationships are given by eqs. (A13.7.6) and (A13.10.22), which
are repeated below for convenience:
D D E
A16.3.1
B

A16.3.2
and
H D
In the above equations the linear polarization and linear magnetization are subsumed in
and , respectively.
509
As eqs. (A16.3.1) and (A16.3.2) have the same form as eqs. (A16.2.5) and (A16.2.6),
but with and replacing 0 and 0 , it is evident that the Maxwell equations will yield
the following equations for the fundamental fields E and B:
2 E
t2
2 B
r2 B D 2
t
r2 E D
A16.3.3
A16.3.4
These are differential equations for an unattenuated wave travelling through a linear,
isotropic and homogeneous medium which is non-conducting and has no free charges,
with a speed c given by
c D 1/2
A16.3.5
The refractive index of the medium n, is given by
nD
c0
D r r 1/2
c
A16.3.6
In a non-magnetic medium r D 1, and then

n D r1/2
A16.3.7
A16.4 CASE III: THE MAXWELL EQUATIONS IN A

NON-LINEAR MEDIUM WITH r D 0 AND s D 0
We shall consider the case of a material medium which is non-linear in its response to
the electric field strength E, but is linear in its response to the magnetic field strength H.
The constitutive relationships are then given by eqs. (A13.7.11) and (A13.10.22), which
are repeated below for convenience:
D D E C P NL
and
H D
B

A16.4.1
A16.4.2
We reiterate that the linear polarization and linear magnetization are subsumed in and
, respectively.
Since D 0 and D 0, we can again use the special forms of the Maxwell equations (A16.2.1.) to (A16.2.4). Inserting eq. (A16.4.1) into eq. (A16.2.4) we obtain
H D
P NL
E
C
t
t
A16.4.3
510
The Raman Effect
Using eq. (A16.4.2) this becomes

B D
P NL
E
C
t
t
A16.4.4
and taking the time derivative we have
B
2 P NL
2 E
D 2 C
t
t
t2
A16.4.5
If we replace 0 0 by in eq. (A16.2.11) and compare the result with eq. (A16.4.5) we
see that the non-linear electric polarization produces an extra term 2 P NL /t2 in the
expression for B /t.
We now take the curl of the Maxwell equation (A16.2.3) and use eq. (A16.4.5) to
obtain
2 E
2 P NL
E D 2
A16.4.6
t
t2
When dealing with a vacuum or a linear medium we were able to make simplifications by
using the vector identity given by eq. (A16.2.8) and introducing the special property of
a vacuum or linear medium that E D 0. However in a non linear-medium, it follows
from combining eq. (A16.4.1) with the Maxwell equation (A16.2.1) that, instead of the
simple relationship, we now have
E C P NL D 0
A16.4.7
and so further progress is not possible without making some assumptions.

Of course, if we assume the non-linear polarization is small, so that
E 0
A16.4.8
then we may take the vector identity given by eq. (A16.2.8) to be a good approximation
and use it to simplify eq. (A16.4.6), obtaining
r2 E D
2 E
C P NL
t2
A16.4.9
Comparison of eq. (A16.4.9) with eq. (A16.3.3) shows that for a non-linear medium an
extra term P NL contributes to r2 E in the approximation of small non-linearity. However,
in general it is preferable to retain the completely general eq. (A16.4.6) and introduce
specific relationships between P NL and E, as required.
The corresponding differential equation for B can be obtained as follows. We rewrite
the Maxwell equation (A16.2.4) in terms of B by using eq. (A16.4.2) which applies when
the system has a linear response to a magnetic field and obtain.
B D
D
t
A16.4.10
511
Taking the curl of this equation we have

D
B D
t
A16.4.11
Then using the vector identity

B D B r 2 B
A16.4.12
and noting from the Maxwell equation (A16.2.2) that

B D 0
eq. (A16.4.11) reduces to

D
r B D
t
2
A16.4.13
Introducing eq. (A16.4.1) into eq. (A16.4.13) and using the Maxwell equation (A16.2.3)
we obtain
2 B
A16.4.14
r2 B D 2 P NL
t
t
Comparison of eq. (A16.4.14) with eq. (A16.3.4) shows that in a non-linear medium an
extra term /t P NL contributes to r2 B .
A16.5 CASE IV: THE MAXWELL EQUATIONS IN A LINEAR

MEDIUM WITH r 6D 0 AND s 6D 0
We must now use the general forms of the Maxwell equations (A16.1.1) to (A16.1.4).
Initially, we follow the same procedures as were used in Case I, to obtain eq. (A16.2.9).
We take the curl of eq. (A16.1.3), introduce the vector identity of eq. (A16.2.8) and obtain
B
D0
A16.5.1
t
This is identical with eq. (A16.2.9) because eq. (A16.1.3) and eq. (A16.2.3) are identical,
and both are independent of and .
We then take eq. (A16.1.4), substitute for D and H using eqs. (A16.3.1) and (A16.3.2),
and introduce the relationship for J given by eq. (A16.1.5). We obtain
E r 2 E C
E
C E
A16.5.2
t
From eq. (A16.1.1) and eq. (A16.3.1) it follows that

E D
A16.5.3
Then introducing eq. (A16.5.3) and the time derivative of eq. (A16.5.2) into eq. (A16.5.1)
we obtain

2 E
E
C
A16.5.4
r2 E D 2 C
t
t
B D
512
The Raman Effect
The first term on the right of this equation comes from the displacement current density
D/t, the second term from the conduction current density E and the third from the
free charge density . In the absence of free charges eq. (A16.5.4) becomes
2 E
E
C

A16.5.5
t2
t
We can obtain the equation for B in a similar manner. We rewrite eq. (A16.1.4) in terms
of B using eq. (A16.4.2) and obtain
r2 E D
D
C J
A16.5.6
t
Then taking the curl of eq. (A16.5.6) and introducing the vector identity of eq. (A16.4.12)
we have
D
B r2 B D
A16.5.7
C J
t
Introducing the Maxwell equation (A16.2.2), eq. (A16.5.7) reduces to
B D
D
A16.5.8
J
t
Since the medium is linear in E we can combine eq. (A16.3.1) with the Maxwell equation (A16.2.3) to give
B
A16.5.9
D D
t
Also introducing eq. (A16.1.5) into the Maxwell equation (A16.2.3) we obtain
r2 B D
B
A16.5.10
t
Finally, substituting eqs. (A16.5.9) and (A16.5.10) into eq. (A16.5.8) we obtain
J D
2 B
B
C
A16.5.11
2
t
t
Again the first term on the right of this equation comes from the displacement current
and the second from the conduction current. This equation is similar to eq. (A16.5.5) for
E except that there are only two terms on the right since there are no free magnetic poles.
Clearly eqs. (A16.5.4) and (A16.5.11) reduce to eqs. (A16.3.3) and (A16.3.4) if both
and are zero.
It should be noted that the first-order derivatives in eqs. (A16.5.4) and (A16.5.11)
arise from the conduction current and lead to damping or dissipation of the wave as it
propagates.
r2 B D
513
A17
Monochromatic Plane Harmonic
Waves in Vacuum and in a
Non-absorbing Linear Medium
What Emanations,
Quick vibrations
And bright stirs are there?
Henry Vaughan
A17.1 GENERAL WAVE EQUATION IN VACUUM

The general homogeneous second-order differential equation for an unattenuated wave
propagating through an isotropic medium with speed v is
1 2
A17.1.1
v2 t2
where , the quantity propagated, can be either a scalar or a vector.
In Chapter A16 we obtained from the Maxwell equations in vacuum, eqs. (A16.2.12)
and (A16.2.15) which are homogeneous, second-order differential equations for E and B.
For convenience we repeat them below:
r2 D
2 E
t2
2 B
r2 B D 0 0 2
t
r2 E D 0 0
A17.1.2
A17.1.3
514
The Raman Effect
From comparison with eq. (A17.1.1) it is evident that each of eqs. (A17.1.2) and (A17.1.3)
corresponds to a differential equation for a wave propagating through vacuum with a speed
which we define as c0 . Thus we have
0 0 c02 D 1
A17.1.4
Equations (A17.1.2) and (A17.1.3) may then be rewritten as

r2 E D
1 2 E
c02 t2
A17.1.5
1 2 B
A17.1.6
c02 t2
There are many types of solution to these wave equations. We shall consider in
detail just one special case which forms the basis for our treatment of light-scattering
phenomena, namely a monochromatic, plane, harmonic, electromagnetic wave propagating along one direction. The significance of these important, property-defining italicized
labels qualifying the word wave will emerge in the development that follows.
r2 B D
A17.2 MONOCHROMATIC PLANE HARMONIC

ELECTROMAGNETIC WAVE IN VACUUM
The solution of the wave equation for the vector E is now treated in section A17.2.1, and
the solution for the vector B in section A17.2.2. Section A17.2.3 deals with the energy
in an electromagnetic wave.
A17.2.1 Solution of the wave equation for the vector E

We first choose the direction of propagation to be the z axis and then, in so far as its
propagation is concerned, E becomes a function of z and t only. We may express this by
writing
E D E z, t
A17.2.1
Now a plane wave has the property that for any plane normal to the direction of propagation the quantity being propagated has the same value over that plane at a given time.
Therefore, for the case we are considering, the vector E is constant over the xy plane
so that
Ey
Ex
D
D0
A17.2.2
x
y
Then combining eq. (A17.2.2) with the Maxwell equation (A16.2.10) we have
Ez
D0
A17.2.3
E D
z
Consequently, Ez cannot be a function of z. As we are only interested in waves and not
in uniform fields, we can take E to be entirely oscillatory, and set
Ez D 0
A17.2.4
Monochromatic Plane Harmonic Waves in Vacuum
515
It follows that for a plane wave propagating along the z axis the vector E is confined to
the xy plane and is given in general by
E D e x Ex C e y Ey
A17.2.5
that is, it has transverse but not longitudinal components.

Equation (A17.1.5) now becomes
2 E
1 2 E
D
z2
c02 t2
A17.2.6
This equation, we re-emphasize, describes a plane wave propagating along the z axis with
the vector E in the xy plane.
The solution of eq. (A17.2.6) for a plane wave which is harmonic in time and
monochromatic, and is associated with one frequency which we designate as the angular
frequency , is given by

z
E D E0 cos t
C
A17.2.7
c0
for propagation along the positive z axis where E0 is the amplitude with components Ex0
and Ey0 . The argument of the cosine term, ft z/c0 C g, is called the phase angle,
and is the phase angle at t D 0 and z D 0. We shall often set equal to zero.
For propagation along the negative z axis the solution is

z
E D E0 cos t C
C
A17.2.8
c0
To simplify subsequent developments we shall now set Ey0 D 0 so that E D ex Ex , and
E0 D ex Ex0 . Then for propagation along the positive z axis we have

z
Ex D Ex0 cos t
C
A17.2.9
c0
and for propagation along the negative z axis we have

z
Ex D Ex0 cos t C
C
c0
A17.2.10
In what follows we shall only consider propagation along the positive z axis. The
variation of Ex with z along the positive z axis at a particular time t is shown in Fig. A17.1.
The variation of Ex with t at a given point has the same form.
When dealing with just one monochromatic plane harmonic wave it is not usually
necessary to add frequency labels to the field symbols. In situations where distinguishing
frequency labels are necessary, as when more than one electromagnetic wave is incident
on a material system, we shall write E and E0 , for example, to indicate that these
quantities are associated with an electromagnetic wave of circular frequency .
Angular frequency, amplitude and phase angle are defined and discussed later in this section.
516
The Raman Effect

y
By
Ex
= m01(E B)
z
Figure A17.1 Variation of E D ex Ex and B D ey By with z at a particular instant for a

plane electromagnetic wave travelling along the z axis (n0 D ez ).
Ex0 , which is always greater than zero, is the maximum value of Ex and is called the
amplitude. Clearly Ex D Ex0 when cosft z/c0 C g has its maximum value of unity.
The angular frequency is the number of vibrations in 2 seconds. The relationship
between and the frequency
, which is defined as the number of vibrations in one
second, is
D 2
A17.2.11
Ex remains unchanged in magnitude when t is replaced by t C T, where T, the period, is
defined by
1
2
TD D
A17.2.12
Ex also remains unchanged when z is replaced by z C 0 where

c0
2 c0
A17.2.13
D
D c0 T
0 is called the wavelength of the wave in vacuum. The number of wavelengths per unit
length in vacuum is the wavenumber
Q , and thus
0 D
Q D
1
D
0
c0
A17.2.14
In the S.I. system the unit of length is the metre so that the S.I. unit for the wavenumber
Q is m1 . However we adopt the almost universal usage of defining the wavenumber
Q
as the number of wavelengths per centimetre in vacuum, so that the unit for
Q is cm1 .
It follows from eqs. (A17.2.11) and (A17.2.14) that
D 2 c0
Q
A17.2.15
517
We now introduce the scalar quantity k0 given by

k0 D
2
2
D
D
0
c0
c0
A17.2.16
and then define a wave vector k0 , of magnitude k0 , which points in the direction of
propagation of the wave. Thus for propagation along z

2
k0 D k0 ez D
ez
A17.2.17
0
where ez , a unit vector along z, is a special case of a general propagation vector n0 defined
below.
If we introduce the scalar k0 into eq. (A17.2.9) and set equal to zero we obtain
Ex D Ex0 cost k0 z
A17.2.18
We may adapt this equation for the general case where the direction of propagation of the
wave is not parallel to one of the axes of the coordinate system x,y,z, but along a general
direction defined by the unit vector n0 . The wave vector k0 is now defined as
k0 D k0 n0
A17.2.19
with k0 still given by eq. (A17.2.16).

If we now define s as the distance from the origin along this general direction of propagation of the wave, eq. (A17.2.18) is replaced by
E D E0 cost k0 s
A17.2.20
The electric field vector E is, of course, orthogonal to n0 , that is

n0 E D 0
A17.2.21
but subject to this condition E may have any direction.

We now introduce r the position vector of some point in space, not necessarily on the
propagation direction of the wave, and defined such that
s D r n0
A17.2.22
Introducing eqs. (A17.2.19) and (A17.2.22), into eq. (A17.2.20) we obtain the general
equation
A17.2.23
E D E0 cost k0 r
We note that k0 r has its maximum value of k0 s when the vector r is parallel to the
direction of propagation, defined by n0 and is zero when r is orthogonal to n0 .
The equation of the plane perpendicular to k0 (or equivalently n0 ) is
k0 r D constant
This choice of notation is explained in Section A17.4.
A17.2.24
518
The Raman Effect
This plane is the locus of all points whose projection on to the k0 direction is a constant.
It follows that for each value of the displacement s there exists a plane for which k0 r
is constant.
Consequently an equation of the form
r D A cosk0 r
A17.2.25
in which A is the amplitude of r, represents a set of planes over which r varies in
space sinusoidally. The function r repeats itself in space after each displacement of
0 in the direction of k0 since

s
k0 r D 2
A17.2.26
0
This behaviour is illustrated in Fig. A17.2 where just a few of the infinite number of
planes are shown. These planes have an infinite extent since no limits have been placed
on r and thus the disturbance occupies all space.
For a given point in space where r is constant the function r is constant so that the
planes are motionless. When r is combined with a sinusoidal time function cos t, the
disturbance travels along the k0 direction. This is illustrated by eq. (A17.2.23) in which
the function E0 cosk0 r is combined with E0 cos t to give the dependence of E on
t and r.
The surfaces joining all points of equal phase at any given time are called wavefronts
or wave surfaces. The wave function will only have a constant value over the wavefront
if the amplitude has a fixed value at every point on the wavefront. Waves that satisfy this
condition are said to be homogeneous; those which do not are said to be inhomogeneous.
A17.2.2 Solution of the wave equation for the vector B

We now consider the solution of the wave equation for the vector B, with the direction
of propagation chosen to lie along the z axis. The initial procedures parallel those used
for the vector E. Thus we have
B D B z, t
A17.2.27
and since we are considering a plane wave, the vector B is constant over the xy plane
so that
By
Bx
D
D0
A17.2.28
x
y
Combining eq. (A17.2.28) with the Maxwell eq. (A16.2.2) we find that
rB D
Bz
D0
z
A17.2.29
Consequently Bz cannot be a function of z and taking B to be entirely oscillatory we

may set
A17.2.30
Bz D 0
519
y(r )
+A
l0
Displacement in the
direction of r = k 0 = n0
(a)
A
y
A
y
k0
(b)
Figure A17.2 (a) The function r D A cosk0 r), with constant amplitude A, propagating along the direction of k0 D n0 D r. (b) Some of the infinite number of planes (or
wavefronts) for each of which k0 r and hence r has a constant value (k0 r D 2 r/).
The density of the shading is a measure of r. Also, in fact, the planes have an infinite
extent since no limits have been placed on r but only limited areas are illustrated here.
Thus for a plane wave propagating along the z axis, the vector B is confined to the xy
plane and is given in general by
B D ex Bx C ey By
A17.2.31
1 2 B
2 B
D
z2
c02 t2
A17.2.32
Equation (A17.1.6.) now becomes
520
The Raman Effect
To proceed further we must now seek the relationship between E and B. We invoke
the Maxwell equation (A16.2.3) and, for the plane wave we are considering, this yields
Ey
Ex
Bx
By
ey
D ex
C ey
z
z
t
t
Expressing eq. (A17.2.7) in component form and differentiating we obtain

z
Ex
D Ex0 sin t
C
z
c0
c0
ex
and

z
Ey
C
D Ey0 sin t
z
c0
c0
A17.2.33
A17.2.34
A17.2.35
We now introduce these derivatives into eq. (A17.2.33) and integrate with respect to time,
taking B to be entirely oscillatory. We obtain

1
z
[ex Ey0 C ey Ex0 ] cos t
C D ex Bx C ey By
A17.2.36
c0
c0
This may be recast in the form
BD
1
[ez E ]
c0
A17.2.37
which shows that B is normal to both E and the direction of propagation ez .

Two special cases should be noted. When E is parallel to the x axis, that is Ex 6D 0 and
Ey D 0, eq. (A17.2.36) yields

Ex
1
z
By D
D Ex0 cos t
C
A17.2.38
c0
c0
c0
and when E is parallel to the y axis, that is Ex D 0 and Ey 6D 0,

Ey
1
z
Bx D
D Ey0 cos t
C
c0
c0
c0
A17.2.39
The relative directions of E and B (and H) are shown in Fig. A17.3 for propagation along
(a) the positive z axis and (b) the negative z axis.
For propagation along the negative z axis eq. (A17.2.37) is replaced by
B D
1
ez E
c0
and as a consequence eqs. (A17.2.38) and (A17.2.39) are replaced by

Ex
1
z
D Ex0 cos t C
C
By D
c0
c0
c0
and

Ey
1
z
Bx D
D Ey0 cos t C
C
c0
c0
c0
A17.2.40
A17.2.41
A17.2.42
521
By = c1 Ex
0
Ey
Ex
z
Bx = c1 Ey
0
x
(a)
y
Bx = c1 Ey
0
Ey
x
Ex
By = c1 Ex
x
(b)
Figure A17.3 Relative directions of E and B for propagation along (a) the positive z axis
(ez ) and (b) the negative z axis (ez ).
Thus the phase of B relative to E is reversed for propagation along the negative z axis,
as is shown in Fig. A17.3.
It follows that for the plane electromagnetic wave defined by eq. (A17.2.9) the corresponding equation for the associated magnetic induction component is

z
By D By0 cos t
C
A17.2.43
c0
and for the associated magnetic field strength component

z
Hy D Hy0 cos t
C
c0
with
Ex
c0
By0 D
Ex 0
c0
A17.2.45
Ex
0 c0
Hy0 D
Ex 0
0 c0
A17.2.46
By D
and
Hy D
A17.2.44
The variation of By with z along the positive z axis at a particular time is also shown in
Fig. A17.1.
522
The Raman Effect
An alternative formulation of the relation between components of B and E, using the

alternating or LeviCività tensor, is as follows:
B D
1
n0 E
c0
A17.2.47
where n0 is the magnitude of the projection on the positive cartesian axis of the unit
vector n0 which lies along the direction of propagation. If the direction of propagation
coincides with the positive cartesian axis then n0 has the value C1 but its presence is
still necessary to identify the propagation direction.
To summarize, the vectors E and B are mutually perpendicular and oriented so that
their vector product E B points in the direction of propagation. E and B have the same
relative magnitudes at all points at all times.
It is instructive to consider the magnitude of B and E, in an electromagnetic wave. For
example, if the magnitude of Ex is 103 V m1 then, as Ex /By D c0 , the magnitude of By
is 3.3 106 T. The corresponding magnitude of Hy is 2.6 A m1 since Hy D By 1
0 .
A17.2.3 Energy considerations for a plane harmonic electromagnetic

wave in vacuum
The energy in electromagnetic radiation arises from the electric and magnetic fields associated with it. There are two contributions, electric energy and magnetic energy, each
defined as energy per unit volume with units of J m3 . According to classical theory,
for the plane wave defined by eqs. (A17.2.9) and (A17.2.38) the instantaneous electric
energy density elec is given by
A17.2.48
elec D 12 0 E2x
and the instantaneous magnetic energy density mag by
2
mag D 12 1
0 By
A17.2.49
These energy densities are in phase and equal since using eqs. (A17.1.4) and (A17.2.38)
1
2
2 0 Ex
1 1 2
2 0 By
D 0 0 c02 D 1
A17.2.50
The total instantaneous energy total , at a given position along the direction of propagation
and at a given instant of time, is thus given by
1 2
2
2
total D 12 0 E2x C 12 1
0 By D 0 Ex D 0 By
A17.2.51
The variations with z of the electric, magnetic and total energy densities at a given instant
are shown in Fig. A17.4.
For radiation of the wavelength range considered in this book, the periodic time is
many orders of magnitude less than the response time of any detector. For example, for
See Section A17.4 page 530.

r
523
(b)
(c)
(a)
Figure A17.4 Energy densities as a function of z for a plane electromagnetic wave

travelling along the z axis: (a) elec D mag ; (b) total D elec C mag ; and (c) the time-averaged
total energy density total .
a wavelength of 500 nm the periodic time is 1.7 1015 s, whereas the response time
of a photomultiplier may be of the order of 108 s. Consequently, only time-averaged
properties of radiation can be measured. For the plane wave we are considering, E2x and
By2 , the time averages over one cycle of the squares of the electric field strength and the
magnetic induction are given by
E2x D 12 E2x0
and
By2 D 12 By20
A17.2.52
Thus, total , the time-averaged total energy density, is given by

2
2
1
1 1 2
1
1 1 2
2
total D 12 0 E2x C 12 1
0 By D 4 0 Ex0 C 4 0 By0 D 2 0 Ex0 D 2 0 By0
A17.2.53
Equation (A17.2.53) can be rewritten in the useful forms:

total /J m3 D 4.427 1012 Ex0 /V m1 2
A17.2.54
or
total /J m3 D 3.979 105 By0 /T2
A17.2.55
The total energy associated with the electromagnetic wave can be considered as travelling with the average density total , as shown in Fig. A17.4, provided that the amplitude
remains constant. Thus the amount of radiation energy which passes through a surface
of unit area perpendicular to the propagation direction in one second, can be obtained by
multiplying total by the speed of propagation, in this case c0 . The resultant quantity is
properly termed the irradiance of the surface I , and has units of W m2 . Thus, expressed
in terms of the electric field strength, I is given by
I D c0 0 E2x D 12 c0 0 E2x0
A17.2.56
524
The Raman Effect
and in terms of the magnetic induction by

1 2
1
2
I D c0 1
0 By D 2 c0 0 By0
A17.2.57
These equations can be rewritten in the forms:

I /W m2 D 1.327 103 Ex0 /V m1 2
A17.2.58
I /W m2 D 1.193 1014 By0 /T2
A17.2.59
and
As it is often necessary to calculate the electric field and magnetic induction from the
irradiance, the following alternative forms of eqs. (A17.2.58) and (A17.2.59) will prove
useful:
A17.2.60
log10 fEx0 /V m1 g D 0.5 log10 fI /W m2 g C 1.439
and
log10 fBy0 /Tg D 0.5 log10 fI /W m2 g 7.038
A17.2.61
The irradiance I must not be confused with the radiant intensity of a point source,
which is the ratio of the time-averaged power d, in a given direction contained in a
conical beam of solid angle d about this direction, to the solid angle d. Thus
ID
d
d
A17.2.62
and has units of W sr1 . The power radiated by a point source into a conical beam of a
given orientation in space and of solid angle d is constant if I is constant. Thus, since
d D dA/r 2 , where dA is the area of the cross-section of the cone perpendicular to the
cone axis and at distance r from the point source, the irradiance produced by a point
source is given by Id/dA D I/r 2 and decreases as r 2 . This may be contrasted with a
parallel beam of electromagnetic radiation where the irradiance is the same at all points
along the propagation direction. It is not uncommon in the literature to find the irradiance
of a small element of the wavefront of electromagnetic radiation from, for example, a
dipole source, described as the intensity, although it is being calculated using 12 c0 0 E20
and is an explicit function of r 2 .
In experimental situations, particularly in light-scattering spectroscopy, instead of
measuring power directly the radiant intensity of the scattered radiation is often determined
by counting the number of photons of a particular frequency s or wavenumber
Q s falling
on the detector in a given period of time. If s is the number of photons of such scattered
radiation in an element of solid angle d falling on the detector in unit time, that is the
photon rate, then
sh s
s hcQ
s
D
A17.2.63
ID
d
d
and thus s is related to the radiant intensity as follows
s D
Id
Id
D
h s
hcQ
s
A17.2.64
525
In a particular experimental situation d is usually kept constant and then s is proportional to I/hs D I/hcQ
s . s has units of photon counts per second.
Similar considerations apply to the irradiance I of the incident radiation of wavenumber
i (or wavenumber
Q i ). If i is the number of photons of incident radiation falling on an
area dA of the detector in unit time, then as
I D
ih i
i hcQ
i
D
dA
dA
A17.2.65
I dA
I dA
D
h i
hcQ
i
A17.2.66
it follows that
i D
In a particular experimental situation dA is usually kept constant so that i is proportional

to I /hi D I /hcQ
i . i has units of photon counts per second.
The above considerations show that when photon counting is used additional
wavenumber (or frequency) factors arise which must be taken into account in analysing
experimental data.
An alternative approach to the calculation of the flow of energy involves the Poynting
vector. The mutually perpendicular electric field and magnetic induction vectors are so
oriented that their vector product points in the direction of propagation, as shown in
Fig. A17.1. This vector product is called the Poynting vector and is denoted by the
symbol . Thus
D 1
A17.2.67
0 E B
For electromagnetic waves, the Poynting vector may be regarded as representing a flow
of energy per unit time, per unit area, in the direction of propagation of the wave. The
magnitude of the Poynting vector thus corresponds to the irradiance. For example, for the
plane wave under consideration, the instantaneous value of the Poynting vector inst at a
given value is given by
inst D 1
0 E B
D 1
0 Ex By ex ey
D 1
0 Ex By ez
2
D 0 c0 E2x ez D c0 1
0 By ez
A17.2.68
where ex , ey and ez are unit vectors along the positive x, y and z directions. The timeaveraged value over one cycle of the Poynting vector is given by
2
1 2
1
D c0 0 Ex ez D 12 c0 0 E20 ez D c0 1
0 By ez D 2 c0 0 By0 ez
A17.2.69
It is instructive to consider the peak values of the electric field strength (i.e., Ex0 values)
associated with the irradiances that can be produced by various sources of radiation. With
conventional sources, even those which might be described as powerful, only relatively
low values of the irradiance can be achieved and the associated electric field strengths are
quite small. For example, the irradiance produced by solar radiation at the Earths surface
526
The Raman Effect
is about 1.4 103 W m2 , corresponding to an electric field strength of 103 V m1 . In
lasers, the energy is propagated in an essentially non-diverging beam of small cross-section
of the order of 105 m2 which can be further reduced by focusing. Thus, even though
the total power in a CW gas laser beam may be only of the order of watts, irradiances
of the order of 104 108 W m2 can be achieved, depending on the type of laser and the
extent to which the laser beam is concentrated by focusing. The corresponding electric
field strengths are of the order of 3 103 to 3 105 V m1 . A laser operating in the
giant-pulse mode, with a pulse duration of the order of 2 108 s, energy of 2 J, and
an unfocused beam of diameter 3 103 m, gives an irradiance of 1013 W m2 and an
electric field strength of 108 V m1 . A 100-fold reduction in beam radius increases the
irradiance to 1017 W m2 and thus the peak electric field strength rises to 1010 V m1 . The
fields binding outer electrons in atoms and molecules are of the order of 1010 1012 V m1 ,
and so it is not surprising that a giant-pulse laser, when focused to a sufficiently small
spot, causes ionization of air. Even larger irradiances, of the order of 1022 W m2 , can
be produced with giant-pulse laser systems using very short pulses and several stages
of amplification. The electric field strength then exceeds 1012 V m1 and is sufficient to
affect atomic nuclei; for example, neutrons have been expelled from lithium deuteride
with such lasers.
The magnetic induction associated with intense laser radiation is no less impressive.
Whereas the magnetic induction associated with solar radiation at the earths surface is
only about 4 106 T, at an irradiance of 1022 W m2 the magnetic induction exceeds
104 T. For a plane electromagnetic wave propagating along z, the dependence of Ex0 on
the irradiance is shown graphically in Fig. A 17.5.
A17.3 THE EXPONENTIAL REPRESENTATION OF HARMONIC

WAVES
So far we have used the sinusoidal representation of an harmonic wave. Although this
is convenient for discussing the periodic properties of waves, the exponential representation is far superior for handling mathematical manipulations of wave equations. We
now discuss this exponential representation which we shall normally use when treating
radiation classically.
The exponential representation is based on the following relationships introduced in
Chapter A6, Section 6.4.
exp i D cos C i sin
A17.3.1
and
exp i D cos i sin
A17.3.2
The corresponding inverse relationships are

cos D 12 exp i C exp i
Chapter A6 treats exponential representation.
A17.3.3
Neutron
emission
n
527
log10{/Ex / V m 1}
0
10
Electron
emission
log10{ / W m2}
2
Sun
Gas laser
10
12
14
16
Q-switched
laser
18
20
22
Laser
amplifier
Figure A17.5 Dependence of Ex0 on the irradiance I of a plane electromagnetic wave

propagating along z.
and
i
A17.3.4
sin D exp i exp i
2
In the previous section we have shown that the sinusoidal representation of a plane
harmonic wave with a real amplitude E0 and set equal to zero which is propagating along
the positive direction is given by eq. (A17.2.23). This is repeated below for convenience:
E D E0 cost k0 r
A17.3.5
Using the exponential representation eq. (A17.3.5) may be rewritten as

E D Re E0 exp it k0 r
A17.3.6
where Re is an operator which means that the real part is to be taken. The equivalence
of eqs. (A17.3.5) and (A17.3.6) follows straightforwardly from eq. (A17.3.2). However,
eq. (A17.3.6) could equally well have been written as
E D Re E0 exp it k0 r
A17.3.7
The choice of eq. (A17.3.6) rather than eq. (A17.3.7) is a matter of convention. Both
equations satisfy the requirement that for propagation along the positive direction t and
k0 r must have opposite signs. However, the sign used with i is a matter of choice.
528
The Raman Effect
The negative sign selected in eq. (A17.3.6) leads to a positive photon momentum. This
is the universal convention in quantum mechanics and so we shall use it throughout.
Some works on classical optics, however, use a plus sign with i. It should be noted that
following accepted practice the operator symbol Re will usually be omitted when the
exponential representation is being used, since it is normally clear from the context that
it is implicit. We shall, however, retain it in this section for clarity.
It is convenient to introduce at this point several useful properties of the exponential
representation. Phase shifts of multiples of /2 are dealt with easily in the exponential
notation. For example for a phase shift of /2 we have
exp i C /2 D exp i exp i /2 D i exp i
A17.3.8
since
exp i /2 D i
A17.3.9
Similarly for a phase shift of /2 we have

exp i /2 D i exp i
A17.3.10
It should be noted that the sign of i associated with a phase shift depends on the sign in
the original exponential term. If exp i had been chosen then i would correspond to a
phase shift of /2 and i to a phase shift of /2. It is useful to note that
Re i exp i D sin
A17.3.11
Re i exp i D sin
A17.3.12
and
Since cos C /2 D sin we see that eq. (A17.3.11) expresses the fact that i corresponds to a phase change of /2. Similarly, since cos /2 D sin we see that
eq. (A17.3.12) expresses the fact that i corresponds to a phase change of /2. These
phase-shift relationships are important in the treatment of circularly polarized radiation
which is discussed in detail in Chapter A21.
Differential operators also take on simple forms. Thus as
fexp it k0 rg D ifexp it k0 rg

A17.3.13
t
the operator /t operating on exp it k0 r can be replaced by the factor i and
the operator n /tn by in . For the function exp it k0 r the operator /t D i,
and the operator n /tn D in .
Similarly, since
fexp it k0 zg D ik0 fexp it k0 zg

A17.3.14
z
for the function exp it k0 z, the operator /z D ik0 and the operator n /zn D
ik0 n ; and for the function exp it k0 z, the operator /z D ik0 and the operator
n /zn D ik0 n .
529
Relationships of the type given by eq. (A17.3.14) are important in dealing with the
electric field strength gradient tensor a typical component of which is
Ezx D
Ex
z
A17.3.15
Thus for an electromagnetic wave propagating along the positive z axis with Ex0 6D 0 and
Ey0 D 0, for which the exponential representation is
Ex D Re Ex0 exp it k0 z
we have
Ezx D
Ex
D Re ik0 Ex0 exp it k0 z
z
A17.3.16
A17.3.17
As k0 D /c0 this may be rewritten as

Ezx D
Re iEx0 exp it k0 z

c0
A17.3.18
If this is expressed in the sinusoidal representation using eq. (A17.3.12) we have

Ezx D
Ex sint k0 z
c0 0
A17.3.19
We now consider an alternative method of representing an harmonic wave which

uses exponential forms but is based on identity relationships. It follows directly from
eq. (A17.3.3) that an equivalent form of eq. (A17.3.5) is the following:
E D 12 E0 fexp it k0 r C exp it k0 rg
A17.3.20
where E0 is real. The numerical factor 1/2 in eq. (A17.3.20) should be noted. Its presence
ensures that the real amplitude E0 has the same definition in both the sinusoidal and
exponential representations. However, this is not a universal convention and as a result
formulae in some texts may involve different numerical factors. The lack of consistency
in such numerical factors is rather unsatisfactory and readers should be on their guard
when comparing formulae in different texts.
We have so far taken the electric field strength amplitude to be real. A more general
form of eq. (A17.3.20) in which the amplitudes are complex is
E D 12 fEQ0 exp it k0 r C EQ0 exp it k0 rg
A17.3.21
where EQ0 is a complex amplitude and EQ0 its complex conjugate. Complex amplitudes
can arise in several ways. For example if is not set equal to zero, eq. (A17.3.20) has to
be replaced by
E D 12 E0 fexp it k0 r C C exp it k0 r C g
The electric field strength gradient is introduced in Chapter 13.
A17.3.22
530
The Raman Effect
This can be rewritten as
E D 12 fEQ0 exp it k0 r C EQ0 exp it k0 rg
A17.3.23
where
and
EQ0 D E0 exp i
A17.3.24
EQ0 D E0 exp i
A17.3.25
Thus we see that if the angle is included the amplitude is necessarily complex.
We now consider a general complex amplitude EQ0 and its complex conjugate EQ0 and
express these in terms of their real and imaginary parts, E0R and E0I , respectively:
EQ0 D E0R C iE0I
EQ0 D E0R iE0I
A17.3.26
A17.3.27
Substituting into eq. (A17.3.23) we obtain

E D 12 [E0R fexp it k0 r C exp it k0 rg
C iE0I fexp it k0 r exp it k0 rg]
A17.3.28
Then using eqs. (A17.3.3) and (A17.3.4) we have

E D E0R cost k0 r C E0I sint k0 r
A17.3.29
Thus, when the amplitude is complex the wave consists of two parts. One part, with
amplitude E0R has the normal sinusoidal dependence on t and r; the other has an amplitude
E0I and is shifted in phase by /2 relative to the first part. This phase shift may be
deduced either from eq. (A17.3.28) having regard to the sign of i in relation to the
exponential terms or from eq. (A17.3.29) since cos /2 D sin .
A17.4 MONOCHROMATIC PLANE HARMONIC WAVE IN A

HOMOGENEOUS, ISOTROPIC AND LINEAR MEDIUM
We now consider the propagation of electromagnetic waves in a material medium which
is homogeneous, isotropic, and linear, has no free charges D 0 and is non-conducting
D 0. In Chapter A16, we derived eqs. (A16.3.3) and (A16.3.4) which are second-order
homogeneous differential equations for E and B in such a medium. For convenience they
are repeated below:
2 E
t2
2 B
r2 B D 2
t
r2 E D
A17.4.1
A17.4.2
531
From the discussion in Section A17.1 it follows that each of these equations corresponds
to a differential equation for an unattenuated wave propagating through the material
medium whose properties are defined above, with a speed 1/2 which we define as
c. Thus we have
A17.4.3
c2 D 1
The wave speed c, is less than c0 , the speed in vacuum. The refractive index n of the
medium, is given by

1
1
2
c0
D r r 2
A17.4.4
D
nD
c
0 0
where r D /0 is the relative permittivity and r D /0 is the relative permeability.
In a non-magnetic medium D 0 or r D 1 and so
1
2
D r 1/2
A17.4.5
nD
0
It is important to appreciate that n, and hence and r , are functions of the frequency of
the wave, that is they exhibit the phenomenon of dispersion. This means that eq. (A17.4.5)
will only hold if n and are measured at the same frequency.
Comparison of eqs. (A17.4.1) and (A17.4.2) with eqs. (A17.1.2) and (A17.1.3) indicates that the results we obtained in Sections A17.1 and A17.2 for an electromagnetic
wave in vacuum can be carried over to the case of the material medium considered here,
provided we make the appropriate modifications for the changed speed of propagation.
Thus, a plane electromagnetic wave in the material medium is entirely transverse and
the dependence of E on r and t is given, in the form of exponential representation used
in eq. (A17.3.23) by
E D 12 fEQ0 exp it k r C EQ0 exp it k rg
A17.4.6
where we use complex amplitudes to maintain generality. The wave vector is now k
(instead of k0 ), and the direction of this wave vector is perpendicular to the plane
containing E and B. Its magnitude is given by
2
2
D
D
A17.4.7
kD

c
c
Using eq. (A17.4.4) this may be rewritten as
kD n
A17.4.8
c0
and introducing eq. (A17.2.16) this becomes
k D k0 n
A17.4.9
Since n exhibits dispersion, so does k. In some situations it is desirable to emphasize the

dependence of k and n on . In such cases eq. (A17.4.8) can be written as
A17.4.10
k D n
c0
532
The Raman Effect
and eq. (A17.4.9) as

k D k0 n
A17.4.11
Similarly, the frequency dependence of and r can be emphasized by writing and

r when needed.
In treating the wave vector in a material medium it is convenient to base its definition
on a propagation vector n, which is not a unit vector but has a magnitude equal to the
refractive index in the direction of propagation. Thus
k D n
A17.4.12
c0
In this convention the magnitude of k or k is always /c0 . However, the propagation
vector is no longer a unit vector but a vector whose magnitude changes with the medium
and the frequency. Thus
n D n0 n
A17.4.13
where n is a unit vector in the direction of propagation. In vacuum n D n0 since
n D 1. We shall use this unit vector n0 , which we have already introduced ad hoc
in section A17.3 to define the propagation direction of an electromagnetic wave in the
general case in a vacuum or in a medium. In specific cases we can set n0 D ex , ey or ez .
In the case of an electromagnetic wave propagating along a general direction defined
by n0 , it is useful to introduce n0 , the projection of the unit vector n0 along the positive
cartesian axis or, equivalently, the cosine of the angle between n and . For example
in the case of a plane electromagnetic wave with electric field vector E and frequency ,
travelling with speed c0 in free space in a general direction defined by the unit vector n0 ,
we have for the direction

E D E0 cos t n0 r0
A17.4.14
c0
Where r0 is the distance from the origin along the direction defined by n0 , and E is the
component which is orthogonal to n0 .
In this chapter the amplitude has been treated as a constant quantity. In some situations
the amplitude may be a slowly varying function of the displacement along the propagation
direction, or a slowly varying function of time or even both. These situations will not be
treated here as they do not arise in this book.
533
A18
The Transition Polarizability
Tensor afi
Wherever a sign is present, ideology is
present too.
Valentin Volosinov
A18.1 INTRODUCTION
In Chapter 4 we gave as eq. (4.2.11) the following result for fi , the component
of a general transition polarizability afi :

fi D
C
A18.1.1
h r6Di,f ri 1 ir
rf C 1 C ir
The purpose of this Appendix is to discuss more fully two matters relating to eq. (A18.1.1)
These are the restrictions r 6D i, f on the summation over all intermediate states jri and
the relative signs of the phenomenological damping term ir in the denominators of
eq. (A18.1.1).
A18.2 THE RESTRICTIONS r 6D i , f

We now consider the two additional terms which arise in the summation over r in
eq. (A18.1.1) when we remove the restriction r 6D i, f. The treatment we use is based on
that given by Mortensen and Svendsen (1981).
534
The Raman Effect
If we let r D i and set i D 0 we have the additional term

1
hfjpO jiihijpO jii hfjpO jiihijpO jii
rDi
fi D

C
h
1
if C 1
If we let r D f and set f D 0 we have a further term

1 hfjpO jfihfjpO jii hfjpO jfihfjpO jii
rDf
C
fi D
h
fi 1
1
A18.2.1
A18.2.2
Summing these two terms we obtain, after a modest algebraic rearrangement, the contribution to fi which results from the removal of the restriction r 6D i, f, namely

hfjpO jfi hijp jii
1
rDi,f
fi
D
hfjpO jii
h
1

hfjpO jfi hijpO jii
hfjpO jii
A18.2.3
C
fi 1
It is evident from eq. (A18.2.3) that the initial and final states i and f can contribute in
principle as intermediate states to fi . We now consider the conditions that must be
rDi,f
met for fi
to be non-zero. We confine detailed discussion to vibrational Raman
scattering.
rDi,f
to be non-zero two conditions must be met. The transition jfi
jii must
For fi
be infrared active and the transition must have associated with it a change of permanent
electric dipole moment. More explicitly these conditions are hfjpO jfi hijp jii 6D 0 and
hfjpO jii 6D 0 and/or hfjpO jfi hijpO jii 6D 0 and hfjpO jii 6D 0.
Neither of the terms in eq. (A18.2.3) can become resonant because neither 1 nor
fi 1 can be zero in Raman scattering. Furthermore, as 1 jfi j we can set
fi 1 1 which leads to the approximate equality
fi fi
A18.2.4
Thus the additional terms contribute only to the antisymmetric components of the
scattering tensor. Such components are normally negligible in non-resonant Raman
scattering.
The above properties lead to restrictions on the symmetry of participating vibrational
transitions. We consider first a transition from the ground vibrational state which is totally
symmetric to a totally symmetric final state. As hfjpO jfi and hfjpO jii must be simultaneously non-zero with 6D , the scattering molecule must have two different cartesian
coordinates that are totally symmetric. This is only possible for the point groups C1 and
The treatment of the transition polarizability tensor used here is based on the Born Oppenheimer
approximation. Beyond the Born Oppenheimer approximation the non-adiabatic corrections to antisymmetric
transition polarizability tensors may be only about two orders of magnitude smaller than the symmetric ones
(Liu and Buckingham, 1993).
The Transition Polarizability Tensor afi
535
Cs which have just one symmetry element, a mirror plane. Thus, generally speaking, only
transitions from the vibrational ground state to non-symmetric vibrational states need to
rDi,f
be considered in relation to fi
Mortensen and Svendsen (1981) have made an order of magnitude calculation of the
change of permanent dipole associated with the fundamental vibrational transition of
the antisymmetric stretching mode, 3 b1 of the water molecule. They found that the
contribution to the vibrational intensity from this source was only about 106 of that
arising from the states r 6D i, f.
rDi,f
on the grounds that
Thus it is justifiable to disregard the contributions from fi
they are very small although formally they should be included in the sum over states r.
A18.3 THE RELATIVE SIGNS OF ir

In eq. (A18.1.1) the term ir has a negative sign in the denominator of the first term and
a positive sign in the denominator of the second term. The use of opposite signs for the
ir terms is in accord with the formula quoted by Placzek (1934) in his classic treatment
of Rayleigh and Raman scattering. However, Placzek does not derive the formula but
claims that Weisskopf (1931) was his source.
In the subsequent literature many authors have used a formula for fi which differs
from the one given above in that the ir terms have negative signs in both denominators.
For convenience we shall refer to this as the same-sign convention to distinguish it from
the opposite-sign convention of the Placzek formula.
The existence of these two conventions seems to have gone unnoticed until recently,
presumably because the relative signs of the terms is usually of little significance. In
resonance Raman scattering the second term in eq. (A18.1.1) can usually be neglected;
and in normal, non-resonance Raman scattering the r are small relative to the frequency
terms in the denominators. However it is unsatisfactory that, in a fundamental formula
for Raman scattering, an unresolved ambiguity should have persisted for so long.
However, recent treatments of a variety of optical phenomena, such as the Pockels effect
and harmonic generation for example, have brought renewed interest in this problem of
relative signs. Several authors, for example Andrews, Naguleswaran and Stedman (1998),
have claimed that the same-sign convention is the correct one. However, recently Buckingham and Fischer (2000) have shown that the same-sign convention yields unphysical
results and that the opposite-sign convention is the correct one for susceptibilities and
hence polarizabilities of all orders.
The problem has also been addressed by Hassing, Mortensen and Svendsen (2000)
specifically for the case of fi . They use arguments based on fundamental principles,
namely the causality principle and the principle of equivalence between two interaction
operators, to show that the sign of ir must follow the sign of 1 in the denominators,
so supporting the opposite-sign convention.
We have used the opposite-sign convention in this book. The several references cited
provide detailed literature references for those interested in the relevant detailed arguments.
536
The Raman Effect
REFERENCES
Andrews, D. L., Naguleswaran, S. and Stedman, G. E. (1998). Phys. Rev. A 57, 4925.
Buckingham, A. D. and Fischer, P. (2000). Phys. Rev. A 61, 035801.
Hassing, S., Mortensen, O. S. and Svendsen, E. N. (2000). Proceedings of XVIIth Conference on
Raman Spectroscopy, Shu-Lin Zhang and Bang-Fen Zhu (eds.), 110. John Wiley and Sons:
Chichester.
Liu, Fan-chen and Buckingham, A. D. (1993). Chemical Physics Letters, 207, 325.
Placzek, G. (1934). Rayleigh-Streuung und Raman-Effekt, in Handbuch der Radiologie, (ed.),
E. Marx, 6, 205374. Academische Verlag: Leipzig.
Mortensen, O. S. and Svendsen, E. N. (1981). J. Chem. Phys. 74, 3185.
Weisskopf, V. (1931). Ann. Physik. 9, 23.
537
A19
ClebschGordan Coefficients
and Wigner 3-j and 6-j Symbols
The setting-up is always worst:
Such heaps of things you want at first.
Lewis Carroll
A19.1 INTRODUCTION
The theory of angular momentum coupling in quantum mechanics, long established as
a powerful technique in both atomic and nuclear physics, is now well recognized as
providing the basis for elegant treatments of some aspects of molecular spectroscopy. As
far as the subject matter of this book is concerned, angular momentum coupling theory is
particularly useful for the treatment of irreducible tensors, and for the evaluation of the
matrix elements which arise in both normal and resonance rotation and vibrationrotation
Raman scattering.
In this Appendix we outline the treatment of the coupling of angular momenta in
terms of the ClebschGordan coefficients and Wigner 3-j and 6-j symbols. Our purpose
is simply to enable the reader to understand the application of these concepts in the
main text, particularly Chapters 6 and 8. Thus we shall confine ourselves to establishing
the nomenclature, giving the basic concepts and definitions, largely without proof, and
providing key formulae which are relevant to the main text. To assist the reader we also
introduce some simple examples at appropriate stages.
It is important to appreciate that a variety of nomenclatures and choices of phase have
been used in the literature, and that inconsistencies are not uncommon. The problems that
538
The Raman Effect
can beset the unwary when combining information from different sources in the literature
have been listed and discussed by Silver (1976).
A19.2 CLEBSCHGORDAN COEFFICIENTS

We consider a compound system consisting of two sets of states, each set with its
associated vector angular momentum operator. The first set is defined to consist of
2j1 C 1 states with a typical member of the set being labelled jj1 m1 i. The vector
angular momentum operator for this set is JO1 with JO 1z its component along the spacefixed z axis. The second set is defined to consist of 2j2 C 1 states with a typical member
being labelled jj2 m2 i. The vector angular momentum operator for this set is JO2 , with JO 2z
its component along the space-fixed z axis.
The direct product jj1 m1 ijj2 m2 i constitutes a representation of the system in which
the states jj1 m1 i and jj2 m2 i are uncoupled. These product states are simultaneous eigenfunctions of the operators JO12 , JO 1z , JO22 and JO 2z . Thus we have
JO12 jj1 m1 ijj2 m2 i D j1 j1 C 1jj1 m1 ijj2 m2 i
JO 1z jj1 m1 ijj2 m2 i D m1 jj1 m1 ijj2 m2 i
A19.2.2
JO22 jj1 m1 ijj2 m2 i
A19.2.3
D j2 j2 C 1jj1 m1 ijj2 m2 i
JO 2z jj1 m1 ijj2 m2 i D m2 jj1 m1 ijj2 m2 i
A19.2.1
A19.2.4
Here j1 can be 0, 1, 2, . . . with m1 D j1 , j1 1 . . . j1 (that is there are 2j1 C 1 values

of m for a given j) and similarly for j2 and m2 . In eqs. A19.2.1 to A19.2.4. the operator
symbols denote the dimensionless ratio (angular momentum)/h.
This uncoupled representation jj1 m1 ijj2 m2 i spans a space of dimension
2j1 C 12j2 C 1. It will be helpful even at this early stage to introduce a specific
example, namely two sets of uncoupled states, one with j1 D 1 and the other with j2 D 1.
It follows that m1 and m2 can each take the values 1, 0 and 1. There are thus nine possible
product states spanning a space of nine dimensions. The uncoupled states are listed as a
column matrix in Table A19.1.
This example of nine uncoupled states j1m1 ij1m2 i has been chosen not merely because
of its relative simplicity, but also because of its importance in Raman spectroscopy. It
transpires that the three components of a vector in the spherical basis , such as V (with
D 1, 0 or 1) or V (with D 1, 0 or 1) transform as the angular momentum states
j1i and j1 i. Thus the nine product states j1ij1 i correspond to those given in
Table A19.1 for j1m1 ij1m2 i.
We now consider a coupled representation consisting of a set of states which we
denote as jjmi. The coupled states jjmi are simultaneous eigenfunctions of the angular
momentum operators JO12 , JO22 , JO 2 D JO1 C JO2 2 and JO z D JO 1z C JO 2z . Thus for the coupled
In general j values can be integral or half-integral, but we only consider integral values in this Chapter.
Chapter A9, Section A9 treats vectors in the spherical basis.
ClebschGordan Coefficients and Wigner 3-j and 6-j Symbols
539
Table A19.1 The uncoupled product states jj1 m1 ijj2 m2 i and the
resulting coupled states jjmi presented as column matrices
jjmi
jj1 m1 ijj2 m2 i
j22i
j11ij11i
j11ij10i
j21i
j10ij11i
j11i
j11ij1 1i
j20i
j10ij10i
j10i
j1 1ij11i
j00i
j10ij1 1i
j2 1i
j1 1ij10i
j1 1i
j2 2i
j1 1ij1 1i
representation we have in place of eqs. (A19.2.2) to (A19.2.4)

JO12 jjmi D j1 j1 C 1jjmi
JO22 jjmi D j2 j2 C 1jjmi
A19.2.5
O2
A19.2.7
J jjmi D jj C 1jjmi

JO z jjmi D mjjmi
A19.2.6
A19.2.8
In these equations the conditions relating to the values of j when j1 j2 are as follows:
j1 C j2 j j1 j2
A19.2.9
or equivalently
j D j1 C j2 ,
j1 C j2 1 . . . j1 j2
A19.2.10
Thus
jmax D j1 C j2
A19.2.11
jmin D j1 j2
A19.2.12
The conditions relating to the values of m are

m D m1 C m2
A19.2.13
jm1 j j1
A19.2.14
jm2 j j2
A19.2.15
with
540
The Raman Effect
The coupled representation jjmi spans a space of dimension 2j C 1 for each j value.
As it can be shown that
jmax

2j C 1 D 2j1 C 12j2 C 1
A19.2.16
jmin
the coupled representation spans a space of the same dimension as the uncoupled representation.
We now illustrate the operation of these conditions by considering the coupled states
jjmi which can be formed from the nine uncoupled states j1m1 ij1m2 i in the example
introduced earlier.
As j1 D 1 and j2 D 1 three values of j are possible, namely j D j1 C j2 D 2, j D
j1 C j2 1 D 1 and j1 j2 D 0. When j D 2, the allowed values of m are 2, 1, 0,
1 and 2. These result from algebraic sums of m1 and m2 values as follows: m D 2
from m1 D 1 C m2 D 1; m D 1 from m1 D 1 and m2 D 0 and also from m1 D 0 and
m2 D 1; m D 0 from m1 D 1 and m2 D 1, from m1 D 1 and m2 D 1 and from m1 D 0
and m2 D 0; and so on. The resulting nine coupled states are also listed in Table A19.1
where they are arranged as a column matrix with the states of the same m value grouped
together, in the order m D 2, m D 1 . . . etc.; and for each m value the states are listed
in order of decreasing j value. The column matrix of the uncoupled product states in
Table A19.1 has its entries ordered so that a row number of a product state for which
m1 C m2 D m corresponds to one of the row numbers of a coupled state jjmi. The ordering
for a given m value is, of course, arbitrary. There are a total of nine coupled states
as expected, because from eq. (A19.2.16) the coupled representation spans a space of
dimension 2j1 C 12j2 C 1 D 9.
Each of the coupled states jjmi is a linear combination of the uncoupled states so that
we can write

Cj1 j2 j; m1 m2 mjj1 m1 , j2 m2 i
A19.2.17
jjmi D
m1 ,m2
where we have introduced the slightly simpler notation for the uncoupled states, writing
jj1 m1 , j2 m2 i for jj1 m1 ijj2 m2 i. The double summation over m1 and m2 ensures that all
possible product states are included.
The coefficients C are usually called ClebschGordan coefficients, and we shall adopt
that name here. It will often be convenient to use C-coefficient as an abbreviation for
ClebschGordan coefficients. Other names used in the literature are vector coupling coefficients, vector addition coefficients and Wigner coefficients. In eq. (A19.2.17) the notation
used for labelling the C-coefficients cites all the quantum numbers. Although this is clear
and unambiguous, the notation can be simplified when the inter-relation of the quantum
numbers, which results from the properties of the C-coefficients is taken into account. We
shall return to this a little later, when the relevant properties of the C-coefficients have
been established.

The inverse of the transformation defined in eq. (A19.2.17) is given by

Cj1 j2 j; m1 m2 mjjmi
jj1 m1 , j2 m2 i D
541
A19.2.18
j,m
The C-coefficients are scalar products of the states jj1 m1 , j2 m2 i and the state jjmi.
We may regard the state jjmi as a vector in the [2j1 C 12j2 C 1]-dimensional state
spanned by the orthogonal product states jj1 m1 , j2 m2 i. Then the contribution of the state
jj1 m1 , j2 m2 i to the vector jjmi is just the projection of the vector on this state, that is
the scalar product of jjmi and jj1 m1 , j2 m2 i.
It follows from these considerations of scalar products that the following identification
may be made:
Cj1 j2 j; m1 m2 m hj1 m1 , j2 m2 jjmi hjmjj1 m1 , j2 m2 i
A19.2.19
Many authors choose to employ these bra-ket forms of the vector coupling coefficients
rather than the ClebschGordan form which we use here. The transformation between
the two conventions is readily made using eq. (A19.2.19).
A C-coefficient is zero unless the following two conditions on the quantum numbers
are satisfied simultaneously:
(i)
m1 C m2 D m
A19.2.20
(ii)
j1 C j2 j j1 j2
A19.2.21
with j1 j2 assumed, or equivalently,

j D j1 C j2 , j1 C j2 1 j1 j2
A19.2.22
The latter condition is usually called the triangle condition and is often written as
j1 j2 j. This condition is a consequence of the vector addition of the angular momenta
vector operators, that is JO D JO1 C JO2 . The triangle condition is symmetric in the three
angular momenta. The above conditions are equivalent to saying that when coupling the
states jj1 m1 i and jj2 m2 i, m1 and m2 add algebraically but j1 and j2 add vectorially.
It must be recognized that the definition of the C-coefficients implies an order of the
addenda. The consequence of such operations as interchanging j1 and j2 or exchanging
an addendum with a resultant (for example combining j1 and j to give j2 ) are determined
by the symmetry properties of the C-coefficients which we shall deal with later.
We now consider a specific case of eq. (A19.2.17) relating to the example introduced
earlier, namely the relationship between the nine coupled states jjmi and the nine product
states jj1 m1 ijj2 m2 i when j1 D 1 and j2 D 1. This relationship is given by eq. (A19.2.23)
in the form of a matrix equation with the coupled and product states as 9 1 column
matrices (see Table A19.1) and the C-coefficients as a 9 9 matrix. The matrix of the
j
C-coefficients is related to the matrix which gives the m in terms of the . This
latter matrix is defined in eq. (A14.5.2), and entries in it can be obtained from the corresponding entries in the matrix of the C-coefficients by multiplying the latter by the
Note that this matrix has the symbol C with entries Cjm, . The term C-coefficient is always used to
denote a Clebsch-Gordan coefficient.
542
The Raman Effect
phase factor 1 .
A19.2.23
Some refinement of the nomenclature for C-coefficients introduced in eq. (A19.2.17)
is possible. The uncoupled product states are characterized simply by m1 and m2 for a
given choice of j1 and j2 ; and the coupled states are characterized by j and m, but with j
limited to vectorial sums of j1 and j2 . However, the three projection quantum numbers are
related by m1 C m2 D m so that it is only necessary to cite two of them. With these observations in mind, a C-coefficient can be adequately labelled as Cj1 j2 j; m1 m2 instead of
Cj1 j2 j; m1 m2 m. However, we shall normally use the full set of six quantum numbers.
We see from eq. (A19.2.23) that the matrix of the C-coefficients is factored into submatrices, one for each value of m. Now m D 2 can be achieved in only one way and so the
associated submatrix is of order one. Similarly m D 1 can be achieved in two ways and
so the associated submatrix is of order two and so on for m D 0, 1 and 2.
The factorization of the matrix of C-coefficients can be interpreted as follows. The set of
product functions jj1 m1 ijj2 m2 i form a basis for the reducible representation Dj1 Dj2
which is of dimension 2j1 C 12j2 C 1. The coupled states jjmi form the basis for a
reduced representation and this basis is obtained using the ClebschGordan coefficients.
In the specific example under consideration the reducible representation is D1 D1
and has dimensions 9 9; and the reduced representation consists of two irreducible
representations of order one m D 1, two of order two m D 2, and one of order three
m D 0. The C-coefficients are taken to be real. Thus the matrix of the C-coefficients is
a real unitary or orthogonal matrix and so its inverse is its transpose.
The following sum relationships hold:

Cj1 j2 j; m1 m2 mCj1 j2 j0 ; m1 m2 m0 D jj0 mm0
A19.2.24
m1
m2
which expresses the orthogonality condition on the rows of the matrix of C-coefficients;
and

Cj1 j2 j; m1 m2 mCj1 j2 j; m10 m20 m D m1 m10 m2 m20
A19.2.25
j,m
which expresses the orthogonality on the columns of the matrix of C-coefficients.
Chapter A10, Section A10.5 treats reducible and irreducible representations.
543
The submatrices of the matrix of C-coefficients, each of which is characterized by one

value of m, are also orthogonal. Thus for the rows

Cj1 j2 j; m1 , m m1 , mCj1 j2 j0 ; m10 , m m1 , m D jj0
A19.2.26
m1
for a fixed m, and for the columns

Cj1 j2 j; m1 , m m1 , mCj1 j2 j; m10 , m m10 , m D m1 m10
A19.2.27
for a fixed m.
The orthonormal properties just given can be verified using the matrix and submatrices
defined in eq. (A19.2.23).
The orthogonal nature of the matrix of the C-coefficients means that
hj1 j2 jmjj1 m1 , j2 m2 i D hj1 m1 , j2 m2 jj1 j2 jmi
A19.2.28
We now consider some symmetry relationships for the C-coefficients. Here and subsequently we replace j, m by j3 m3 to emphasize the symmetry properties. Three independent
symmetry relationships for the C-coefficients are
Cj1 j2 j3 ; m1 m2 m3 D 1j1 Cj2 j3 Cj1 j2 j3 ; m1 , m2 , m3
j1 Cj2 j3
A19.2.29
Cj2 j1 j3 ; m2 m1 m3
A19.2.30
1/2
2j3 C 1
Cj1 j3 j2 ; m1 , m3 , m2 A19.2.31
D 1j1 m1
2j2 C 1
The conditions j1 j2 j3 and m1 C m2 D m3 operate in the above equations; otherwise
the C-coefficients are zero.
Equation (A19.2.29) shows that an overall change in the sign of the projection quantum
numbers is equivalent to a change of phase. Equation (A19.2.30) shows that the same
change of phase is produced by interchanging j1 and j2 (the two j values of the two
states that are being combined) together with their two projection quantum numbers m1
and m2 The third symmetry relationship, eq. (A19.2.34), is concerned with the effect of
interchanging the resultant j3 and m3 with an addendum, in this case j2 and m2 . In this
case in addition to a phase change 1j1 m1 , the square root of a statistical weight ratio,
namely [2j3 C 1/2j2 C 1]1/2 is introduced. Thus this interchange lacks symmetry and
this reflects the special role of j and m.
Some other useful relationships that can be derived from eqs. (A19.2.32) to (A19.2.34)
are given below:

1/2
j2 Cm2 2j3 C 1
Cj3 j2 j1 ; m3 , m2 , m1 A19.2.32
Cj1 j2 j3 ; m1 m2 m3 D 1
2j1 C 1

1/2
j1 m1 2j3 C 1
Cj3 j1 j2 ; m3 , m1 , m2 A19.2.33
Cj1 j2 j3 ; m1 m2 m3 D 1
2j2 C 1

1/2
j2 Cm2 2j3 C 1
Cj1 j2 j3 ; m1 m2 m3 D 1
Cj2 j3 j1 ; m2 m3 m1
A19.2.34
2j1 C 1
D 1
544
The Raman Effect
We see for example that when j3 is interchanged with j1 in eq. (A19.2.32) the associated projection quantum numbers m3 and m1 change sign, but the projection quantum
number associated with the angular momentum which is not interchanged with j3 remains
the same. The phase factor depends on the quantum numbers not involved in the interchange. The statistical weights in the square root involve j3 in the numerator and the
j-number which is interchanged with j3 in the denominator. Note that in eqs. (A19.2.33)
and (A19.2.34) there are two transpositions of the j-numbers, and they appear as a cyclic
permutation of j1 , j2 , j3 .
Formulae for the Clebsch-Gordan coefficients hj1 m1 , j2 m2 jjmi are given in the
Reference Table A19.1 for j2 D 0, j2 D 1 and j2 D 2. These results are important for
rotation and vibration-rotation Raman scattering as they are directly related to the PlaczekTeller factors b0 , b1 and b2 .
A19.3 WIGNER 3-j SYMBOLS

The symmetry relations for the C-coefficients are not particularly easy to remember, but
Wigner has devised a symbol closely related to the C-coefficients which has very simple
symmetry properties and sets out the j and m values with admirable clarity. This symbol
which is known as the Wigner 3-j symbol is defined as follows:

j1 j2 j3
D 1j1 j2 m3 2j3 C 11/2 Cj1 j2 j3 ; m1 m2 m3
A19.3.1
m1 m2 m3
We see that a 3-j symbol simply represents another numerical coefficient which differs
from the corresponding C-coefficient by a phase factor 1j1 j2 m3 and a weight factor
2j3 C 11/2 . This is illustrated in Table A19.2 which gives examples of values of
Table A19.2 Some C-coefficients, their corresponding 3-j symbols and numerical
values.
Cji j2 j3 ; m1 m2 m3
C112; 1 10
C112; 000
C112; 110
1
p
6
2
p
6
1
p
6
numerical valuea

j1 j2 j3
mi m2 m3
numerical valueb
1 1 2
1 1 0
1
p
30
1 1 2
0 0 0
2
p
30
1 1 2
1 1 0
1
p
30
Values are from the j3 D j D 2, m3 D m D 0 row of the matrix of C-coefficients given in

eq. (A19.2.23).
b
Values obtained from C-coefficients using eq. (A19.3.1). These values may be verified
using the explicit formula for the 3-j symbol given by eq. (A19.3.9).
Chapter 6, Section 6.3.2 deals with the Placzek-Teller factors.
545
C-coefficients taken from the matrix equation (A19.2.23) and the values of the corresponding 3-j symbols calculated using eq. (A19.3.1). Note however that the 3-j symbols
are not matrix elements of a unitary transformation.
The conditions for the non-vanishing of a 3-j symbol should be carefully noted.
Since the C-coefficient as defined in eq. (A19.3.1) vanishes unless m1 C m2 D m3 , the
3-j symbol vanishes unless m1 C m2 C m3 D 0. Also, the triangular condition on the j
values, namely j1 j2 j3 applies for the 3-j symbols as for the C-coefficients. The
satisfaction of the triangular condition is implied in the formulae that follows.
As transition polarizabilities for rotational transitions can be expressed in terms of 3-j
symbols involving rotational quantum numbers, the selection rules for rotational transitions are determined by the conditions for non-vanishing of the relevant 3-j symbols. The
relative intensities of rotational transitions are determined by the squares of the magnitudes of the 3-j symbols. The satisfaction of the triangular condition is implied in the
formulae that follows.
The 3-j symbol has easily remembered symmetry properties. An even permutation of
columns leaves the value of the symbol unchanged:

j1 j2 j3
j2 j3 j1
j3 j1 j2
D
D
A19.3.2
m1 m2 m3
m2 m3 m1
m3 m1 m2
An odd permutation of columns multiplies the value of the symbol by 1j1 Cj2 Cj3 :

j1 j2 j3
j2 j1 j3
j1 j3 j2
1j1 Cj2 Cj3
D
D
m1 m2 m3
m2 m1 m3
m1 m3 m2

j3 j2 j1
D
A19.3.3
m3 m2 m1
There are six possible orderings of the symbols j1 , j2 and j3 in the 3-j symbol, and the
values of the remaining five 3-j symbols can be obtained from any one 3-j symbol by
using the above permutation symmetry properties. A further property is that

j1 j2 j3
j1
j2
j3
j1 Cj2 Cj3
D 1
A19.3.4
m1 m2 m3
m1 m2 m3
It follows from this that a 3-j symbol with all m values equal to zero vanishes unless
j1 C j2 C j3 is even.
The orthogonality properties of the C-coefficients lead to corresponding orthogonality
properties for the 3-j symbols. Thus the orthogonality condition expressed by
eq. (A19.2.2) means that

j
j2 j3
j1 j2 j03
1
D 2j3 C 11 j03 j3 m30 m3
A19.3.5
m1 m2 m3
m1 m2 m30
m1 ,m2
and

j3 ,m3
2j3 C 1
j1
m1
j2
m2
j3
m3
j1
m10
j2
m20
j3
m3
D m1 m10 m2 m20
A19.3.6
546
The Raman Effect
If we sum over m3 in eq. (A19.3.5) we have

j
j2
1
m1 m2
m1 ,m2 ,m3
j3
m3
D1
A19.3.7
because there are 2j3 C 1 values of m3 . This is a very useful result.

General algebraic formulae for 3-j coefficients are complicated and their evaluation
tedious. However, computer programmes are available for their numerical evaluation.
In theoretical developments in this book specific algebraic formulae for 3-j coefficients
are required and these will be introduced as they are needed in the main text. We conclude
this section by presenting the following general results which can prove useful:

j
j 1
D 1jm m[2j C 1j C 1j]1/2
A19.3.8
m m 0

1jm 2[3m2 jj C 1]
j
j 2
D
A19.3.9
m m 0
[2j C 32j C 22j C 12j2j 1]1/2

1/2
j1 j2 j3
J/2 j1 C j2 j3 !j1 C j3 j2 !j2 C j3 j1 !
D 1
0 0 0
j1 C j2 C j3 C 1!
J/2!
A19.3.10
J/2 j1 !J/2 j2 !J/2 j3 !
where J D j1 C j2 C j3 and is even. This symbol vanishes if J is odd, a result that follows
from eq. (A19.3.4). In passing we note that any symbol vanishes if it has two identical
columns and j1 C j2 C j3 is odd, a consequence of eq. (A19.3.3). Another very useful
general formula is

j1
j2 0
D 1j1 m1 2j1 C 11/2 j1 j2 m1 m2
A19.3.11
m1 m2 0
Extensive tables of algebraic expressions for 3-j symbols are available in the literature.
References to some of these are given in the reading list for this chapter.
A19.4 WIGNER 6-j SYMBOLS

Wigner 6-j symbols arise in the coupling of three angular momenta. Consider a compound
system consisting of three sets of states, each set with its associated angular momentum
operator. Typical members of each of the three sets will be labelled as jj1 m1 i, jj2 m2 i and
jj3 m3 i. They are eigenfunctions of the completely uncoupled representation for which
the vector angular momentum operators are as follows. For the set jj1 m1 i, JO1 with JO 1z its
component along the space fixed z axis; for the set jj2 m2 i, JO2 (and JO 2z ); and for jj3 m3 i,
JO3 (and JO 3z ).
When these three angular momenta are coupled, specifying the magnitude j of the
resulting vector and m its z component does not characterize uniquely the coupled state
547
because this also depends on the order of addition of the angular momenta. This is in
contrast to the coupling of two angular momenta considered in Section A19.2.
We now examine two alternative coupling schemes: (i) combine j1 with j2 to give j12
and then combine j12 and j3 to give the resultant j. (ii) combine j2 and j3 to give j23
and then combine j23 with j1 to give j. We see that the final state jjmi can be reached
via either an intermediate state jj12 i or an intermediate state jj23 i. As the values of the
angular momenta of the intermediate states are not in general unique, we have for a given
final state jjmi two different sets of states which differ in the values of j12 and j23 .
The eigenfunctions for the final coupled states reached via schemes (i) and (ii) can be
written as
scheme (i) : jj1 j2 j12 j3 jmi jj12 j3 jmi
A19.4.1
scheme (ii) : jj1 j23 j23 jmi jj1 j23 jmi
A19.4.2
Each of these eigenfunctions is a linear combination of the eigenfunctions of the

completely uncoupled representation which we may write as
jj1 m1 ijj2 m2 ijj3 m3 i jj1 m1 , j2 m2 , j3 m3 i
A19.4.3
The two representations of the same set of states arising from schemes (i) and (ii) must
be physically equivalent. They are therefore connected by a unitary transformation just
as the coupled and uncoupled representations of the addition of two angular momenta are
so related. We may express this by writing
jj1 j23 jmi D

hj12 j3 jjj1 j23 j0 ijj12 j3 j0 m0 ijj0 mm0
A19.4.4
j12
Here each expansion coefficient hj12 j3 jjj1 j23 j0 i hj12 j3 jmjj1 j23 j0 m0 i is a scalar
product between the eigenfunctions taken from the two coupling systems. Because of
orthogonality this scalar product vanishes unless j D j0 and m D m0 . Moreover, its value
is independent of m because a scalar product does not depend on the orientation of the
coordinate system.
The coefficients in the unitary transformation are real and are usually called recoupling
coefficients because they relate to the transformation between two different couplings,
which in this case are scheme (i) and scheme (ii). In scheme (i) with the coupled state
2
and JO32 are diagonal; and in scheme (ii)
eigenfunctions jj12 j3 jmi the operators JO 2 , JO z , JO12
2
are
with the coupled state eigenfunctions jj1 j23 jmi the operators JO 2 , Jz , JO12 and JO23
diagonal.
Explicit expressions can be developed for the recoupling coefficients but we shall
not treat this here. We merely note that such expressions involve the contraction of the
product of four ClebschGordan coefficients. As the latter are known in principle all the
recoupling coefficients for the addition of three angular momenta can be obtained.
548
The Raman Effect
We now introduce the Wigner 6-j symbol. This is directly related to the recoupling
coefficients and is defined as follows:

j1 j2 j12
D 1j1 Cj2 Cj3 Cj [2j12 C 12j23 C 1]1/2 hj12 j3 jjj1 j23 ji
j3 j j23
A19.4.5
Note that the braces f g are invariably used for 6-j symbols, to distinguish them from
3-j symbols for which the brackets ( ) are used.
The 6-j symbols have many useful symmetry properties and are consequently much
more convenient to use than the related recoupling coefficients. For the discussions that
follow it will be advantageous to introduce a 6-j symbol defined as follows.

j1 j2 j3
A19.4.6
j4 j5 j6
With this definition the 6-j symbol relates to the following two coupling schemes:
(a) couple j1 and j2 to give j3 and then couple j3 with j4 to give j5 ; (b) couple j2 and
j4 to give j6 and then couple j6 with j1 to give j5 . In the new notation, eq. (A19.4.5) is
replaced by the following:

j1 j2 j3
D 1j1 Cj2 Cj4 Cj5 [2j3 C 12j6 C 1]1/2 hj3 j4 j5 jj1 j6 j5 i A19.4.7
j4 j5 j6
The 6-j symbol is zero unless each of the four triads j1 , j2 , j3 , j3 , j4 , j5 , j2 , j4 , j6
and j6 , j1 , j5 satisfies the triangle condition. Thus the 6-j symbol can be represented
by an irregular tetrahedron as shown in Fig. A19.1. The six j numbers are represented by
the edges of the tetrahedron; and the three edges of each face correspond to the triangular
condition for a particular triad.
The 6-j symbol is invariant under interchange of any two columns. Thus for example

j1 j2 j3
j2 j1 j3
D
A19.4.8
j4 j5 j6
j5 j4 j6
It is also invariant under interchange of any two numbers in the bottom row with the
corresponding two numbers in the top row. Thus, for example,

j1 j2 j3
j1 j5 j6
D
A19.4.9
j4 j5 j6
j4 j2 j3
j1
j2
j3
j6
j5
j4
Figure A19.1 An irregular tetrahedron; the six j numbers form the

edges and each face corresponds to a 3j symbol.
549
As a result of these symmetry properties the sequence of numbers j1 , j2 , . . . j6 in the 6-j

symbol can be put in 24 equivalent forms. If we regard the tetrahedron in Fig. A19.1 as
being regular, the 24 equivalent permutations of the numbers j can be obtained by means
of the 24 symmetry transformations of a regular tetrahedron.
Useful consequences of the four triangle conditions are that
1j1 Cj2 Cj3 D 1j1 j2 j3
A19.4.10
12j1 Cj2 Cj3 D 1
A19.4.11
and
The elements

1/2
[2j3 C 12j6 C 1]
j1
j4
j2
j5
j3
j6
form a real orthogonal matrix where j3 , j6 are considered to be the matrix indices. Thus

j1 j2 j3
j1 j2 j3
2j3 C 12j6 C 1
D j6 j06
A19.4.12
j4 j5 j6
j4 j5 j06
j3
A 6-j symbol may be written as the combination of four 3-j symbols. Thus

j1 j2 j3
D
1j1 m1 Cj2 m2 Cj3 m3 Cj4 m4 Cj5 m5 Cj6 m6
j4 j5 j6
all m

j1
j2
j3
j1
j5 j6
m1 m2 m3

m1 m5 m6

j2
j6 j4
j3
j4 j5
A19.4.13
m2 m6 m4
m3 m4 m5
Using the symmetry properties of the 3-j symbols and also the fact that m1 C m2 C m3
sums to zero and that j1 C j2 C j3 sums to an integer we may rewrite eq. (A19.4.13) as

j1 j2 j3
D
1j4 m4 Cj5 m5 Cj6 m6
j4 j5 j6
all m

j1 j2 j3
j1
j5 j6
m1 m2 m3
m1 m5 m6

j4 j2
j6
j4 j5 j3
A19.4.14
m4 m2 m6
m4 m5 m3
In fact, only two of the six summation indices in eqs. (A19.4.13) and (A19.4.14) are
independent, as the magnetic quantum numbers in each of the four 3-j symbols must
sum to zero. Hence if one index, say , is omitted from the sum, the right-hand side of
eqs. (A19.4.13) and (A19.4.14) must be multiplied by 2j C 1.
550
The Raman Effect
Another type of
be replaced by the

j1
j4
relationship is that which allows the product of three 3-j symbols to

product of one 6-j symbol and one 3-j symbol. Thus

j2 j3
j5 j1 j6
j5 j6
m5 m1 m6

j1 j2
j3
D
1j1 Cj2 j3 Cj4 Cj5 Cj6 m1 m4
m1 m2 m3
m2 ,m3 ,m4

j4 j5 j3
j2 j4
j6
A19.4.15
m4 m5 m3
m2 m4 m6
It should be noted that because of the properties of the 6-j and 3-j symbols, in particular
their symmetries, the above relationships can be recast in a variety of alternative but
equivalent forms. Plus c a change, plus cest la meme chose.
Algebraic expressions can be derived for a wide range of 6-j symbols. One example
which finds application in Chapter 8 is the following:

j1
j2
j3
1 j3 1 j2 1

1/2
ss C 1s 2j1 1s 2j1
D 1s
A19.4.16
2j2 12j2 2j2 C 12j3 12j3 2j3 C 1
where
A19.4.17
s D j1 C j2 C j3
Extensive tables of algebraic expressions for 6-j symbols are available in the literature.
References to a selection of these are given in the reading list for this chapter.
REFERENCE

for j2 D 0, 1 and 2.
551
Formulae for the ClebschGordan coefficients hj1 m1 , j2 m2 jjmi
j2 D 0
hj1 m1 , 0 0jj mi
D m1 m j1 j
j2 D 1
hj1 m 1, 1 1jj1 C 1 mi
hj1 m 1, 1 1jj1 mi
hj1 m 1, 1 1jj1 1 mi
hj1 m, 1 0jj1 C 1 mi
hj1 m, 1 0jj1 mi
1
j1 C mj1 C m C 1 2
D
2j1 C 12j1 C 2
1
j1 C mj1 m C 1 2
D
2j1 j1 C 1
1
j1 mj1 m C 1 2
D
2j1 2j1 C 1
1
j1 m C 1j1 C m C 1 2
D
2j1 C 1j1 C 1
hj1 m, 1 0jj1 1 mi
hj1 m C 1, 1 1jj1 C 1 mi
hj1 m C 1, 1 1jj1 mi
hj1 m C 1, 1 1jj1 1 mi
[j1 j1 C 1] 2
1
j1 mj1 C m 2
D
j1 2j1 C 1
1
j1 mj1 m C 1 2
D
2j1 C 12j1 C 2
1
j1 mj1 C m C 1 2
D
2j1 j1 C 1
1
j1 C m C 1j1 C m 2
D
2j1 2j1 C 1
j2 D 2
hj1 m 2, 2 2jj1 C 2 mi
hj1 m 2, 2 2jj1 C 1 mi
hj1 m 2, 2 2jj1 mi
hj1 m 2, 2 2jj1 1 mi
1
j1 C m 1j1 C mj1 C m C 1j1 C m C 2 2
2j1 C 12j1 C 22j1 C 32j1 C 4
1
j1 C m 1j1 C mj1 C m C 1j1 m C 2 2
D
j1 C 1j1 C 22j1 2j1 C 1
1
3j1 C m 1j1 C mj1 m C 1j1 m C 2 2
D
j1 C 12j1 12j1 2j1 C 3
1
j1 C m 1j1 mj1 m C 1j1 m C 2 2
D
j1 j1 C 12j1 22j1 C 1
552
The Raman Effect

hj1 m 2, 2 2jj1 2
hj1 m 1, 2 1jj1 C 2
hj1 m 1, 2 1jj1 C 1
hj1 m 1, 2 1jj1 mi
hj1 m 1, 2 1jj1 1
hj1 m 1, 2 1jj1 2
(Continued )
1
j1 m 1j1 mj1 m C 1j1 m C 2 2
mi
D
2j1 22j1 12j1 2j1 C 1
1
j1 m C 2j1 C m C 2j1 C m C 1j1 C m 2
mi
D
j1 C 1j1 C 22j1 C 12j1 C 3
1
2
j1 C mj1 C m C 1
mi
D j1 2m C 2
j1 C 1j1 C 22j1 2j1 C 1
1
2
3j1 m C 1j1 C m
D 1 2m
j1 2j1 12j1 C 22j1 C 3
1
2
j1 mj1 m C 1
mi
D j1 C 2m 1
j1 1j1 2j1 C 12j1 C 2
1
j1 m 1j1 mj1 m C 1j1 C m 1 2
mi
D
j1 1j1 2j1 12j1 C 1
1
3j1 m C 1j1 m C 2j1 C m C 1j1 C m C 2 2
D
j1 C 22j1 C 12j1 C 22j1 C 3
1
3j1 m C 1j1 C m C 1 2
Dm
j1 j1 C 1j1 C 22j1 C 1
hj1 m, 2 0jj1 mi
hj1 m, 2 0jj1 1 mi
hj1 m, 2 0jj1 2 mi
hj1 m C 1, 2 1jj1 C 2 mi
hj1 m C 1, 2 1jj1 C 1 mi
hj1 m C 1, 2 1jj1 1 mi
hj1 m C 1, 2 1jj1 2 mi
3m2 j1 j1 C 1
1
[j1 j1 C 12j 12j1 C 3] 2
1
2
3j1 mj1 C m
D m
j1 1j1 j1 C 12j1 C 1
1
3j1 m 1j1 mj1 C m 1j1 C m 2
D
j1 2j1 22j1 12j1 C 1
1
j1 mj1 m C 1j1 m C 2j1 C m C 2 2
D
j1 C 1j1 C 22j1 C 12j1 C 3
1
2
j1 mj1 m C 1
D j1 C 2m C 2
j1 j1 C 22j1 C 12j1 C 2
1
2
3j1 mj1 C m C 1
D 2m C 1
j1 2j1 12j1 C 22j1 C 3
1
2
j1 C mj1 C m C 1
D j1 2m 1
j1 1j1 2j1 C 12j1 C 2
1
j1 m 1j1 C m 1j1 C mj1 C m C 1 2
D
j1 1j1 2j1 12j1 C 1

hj1 m C 2, 2 2jj1 C 2
hj1 m C 2, 2 2jj1 C 1
hj1 m C 2, 2 2jj1 1
hj1 m C 2, 2 2jj1 2
553
(Continued )
1
j1 m 1j1 mj1 m C 1j1 m C 2 2
mi
D
2j1 C 12j1 C 22j1 C 32j1 C 4
1
j1 m 1j1 mj1 m C 1j1 C m C 2 2
mi
D
j1 j1 C 12j1 C 12j1 C 4
1
3j1 m 1j1 mj1 C m C 1j1 C m C 2 2
D
j1 2j1 12j1 C 22j1 C 3
1
j1 m 1j1 C mj1 C m C 1j1 C m C 2 2
mi
D
j1 1j1 2j1 C 12j1 C 2
1
j1 C m 1j1 C mj1 C m C 1j1 C m C 2 2
mi
D
2j1 22j1 12j1 2j1 C 1
These formulae may be used to obtain the Placzek Teller factors bj using eq. (A6.3.20). Note that
(a) hj1 m1 , j2 m2 jjmi hJi Ki , jKjJf Kf i
(b) The formulae are presented in the form
hj1 m m2 , j m2 jj1 C j mi hJi Kf K, j KjJf Kf i
(c) The Placzek Teller factors bj are expressed in terms of Ki so that m D Ki C 1 when K D 1,
m D Ki 1 when K D 1 and so on.
555
A20
Sources of Electromagnetic
Radiation
We all know what light is, but it is not easy
to tell what it is.
Samuel Johnson
A20.1 INTRODUCTION
The most important source of electromagnetic radiation is the oscillating electric dipole.
Other sources of electromagnetic radiation include the oscillating magnetic dipole and
higher order oscillating electric and magnetic multipoles as, for example, the oscillating
electric quadrupole. The intensity of radiation from an oscillating electric dipole is usually
several orders of magnitude greater than that from other kinds of oscillating multipoles.
Thus the major contributions to light-scattering arise from oscillating electric dipoles.
However, magnetic dipoles and electric quadrupoles are important in some special cases
as for example in scattering from chiral systems. In subsequent sections we present a
quantitative treatment of the properties of oscillating electric dipoles, oscillating magnetic
dipoles and oscillating electric quadrupoles.
A20.2 THE OSCILLATING ELECTRIC DIPOLE AS A SOURCE

The electric dipole has been defined in Chapter A13. If such a dipole oscillates harmonically with frequency then electromagnetic radiation of frequency is produced. Before
we discuss the properties of such radiation, we consider first the several ways in which
molecular systems can give rise to oscillating electric dipoles.
556
The Raman Effect
In the absence of an external electric field, if a molecule has an unsymmetrical electron

distribution, it will have a so-called permanent dipole. During a vibration, the magnitude
of this may change and then there will result a dipole oscillating at the frequency of the
vibration and the emission of radiation with this frequency. If the electron distribution is
symmetrical in a molecule in the absence of an external electric field it has no permanent
dipole in its equilibrium configuration. However, it is still possible for unsymmetrical
vibrations to disturb the symmetry of the electron distribution and produce a dipole
which again oscillates with the frequency of the vibration and emits radiation with this
frequency.
Electric dipoles may also be induced in molecules by an external electric field, and
such dipoles can oscillate and produce radiation. The frequencies of such dipoles are
related to the frequency of the electric field and the frequency of the molecular vibration.
If the electric field is a static one, the induced dipole oscillates with the frequency of
the molecular vibration. If the electric field is an oscillating one, as for example in
electromagnetic radiation, induced dipoles result which, provided the electric field strength
is not too great, oscillate at the frequency of the electric field and also at frequencies which
are the sum of, and the difference between, the electric field frequency and the molecular
frequency.
We shall be particularly concerned with such induced electric dipoles. They are the principal source of light scattering and can exist even if the system has no permanent dipole
moment. However, for our immediate purpose we can maintain generality as we can express
the harmonic variation of any electric dipole oscillating with frequency by
p D p0 cos t
A20.2.1
where p is the instantaneous value of the electric dipole vector, (magnitude p), and p0 its
amplitude (magnitude p0 ).
Using the standard procedures of electromagnetic theory, it is relatively straightforward
to calculate the electric and magnetic field strengths as a function of time and position
produced by such a harmonically oscillating electric dipole in free space. For the dipole
orientation and coordinate system defined in Fig. A20.1, the electric field strength and
magnetic induction vectors at some point with the polar coordinates r, and , are
given by
2 p sin
A20.2.2
ej
ED
40 c02 r
and
0 2 p sin
ek
A20.2.3
BD
4c0 r
where

r
p D p0 cos t
A20.2.4
c0
In eqs. (A10.2.2) and (A10.2.3) ej and ek are unit basis vectors in the polar coordinate
system, and ft r/c0 g is called the phase angle.
Chapter A9, Section A9.3 treats polar coordinates and related polar basis vectors.
Sources of Electromagnetic Radiation
557
z
E
ei
ek
B
r
ej
q
P
f
x
y
Figure A20.1 Orientation of E, B and vectors for an oscillating electric dipole oriented
as shown (r : ft r/c0 g D 0). The unit polar vectors ei , ej and ek at a given point
lie along the direction of the vectors , E and B , respectively (see Fig. A9.2).
In deriving these formulae, it has been assumed that the dimensions of the electric
dipole are small compared with r and , and that r is much greater than , where is
the wavelength of the radiation of frequency . These conditions are readily satisfied for
dipoles of molecular origin and radiation in the visible region, under practical conditions
of observation.
It can be seen that both E and B propagate as waves with velocity c0 and vary inversely
as the distance r. They both vary as 2 ; and as far as amplitude is concerned, they also
both have the same angular dependence which is determined by sin . Thus, the amplitudes
of E and B have maximum values in the equatorial plane, are zero along the axis of the
dipole, and have axial symmetry about the dipole axis as shown in Fig. A20.2, which
is a plot of the amplitude of E (or B) as a function of and . However, as far as
direction is concerned, comparison of eqs. (A20.2.2) and (A20.2.3) shows that whereas
B is azimuthal, E lies in a plane passing through the polar axis. The relative orientations
of E and B at a particular point are shown in Fig. A20.1. The ratio of the magnitudes of
E and B is c0 and of E and H is 0 c0 D 377
.
It is clear from these considerations that the vectors E and H produced by an oscillating dipole are related to each other exactly as in a plane electromagnetic wave. An
oscillating dipole with frequency is therefore a source of electromagnetic radiation of
frequency .
Chapter A17 treats plane electromagnetic waves.
558
The Raman Effect

z
x
y
Figure A20.2 Polar diagrams of sin (outer surface) and of sin2 (inner surface) showing,
respectively, the angular distributions of the amplitude of E or B and the magnitude of for
an oscillating electric dipole situated at the origin and oriented along the z axis. The radial
distance from the origin to a surface is proportional to the magnitude of the corresponding
quantity in that direction.
Now that we have related the magnitude of the oscillating dipole to the electric field
strength of the radiation, the energy density and power per unit area of the radiation at
the point defined by r, and may be expressed in terms of the magnitude of the dipole.
According to eq. (A17.2.51) the instantaneous total energy density total at a distance r
along the direction of propagation of the plane electromagnetic wave radiated by an
oscillating electric dipole is given by
total D 0 E2
A20.2.5
where E is the instantaneous magnitude of the electric field strength. Provided there is
no other contribution to E, as for example from magnetic dipole or electric quadrupole
terms, we may introduce eq. (A20.2.2) into eq. (A20.2.5) and obtain for the instantaneous
total energy density

4
1
2 2
2 Q 4 2 2
p
sin

D
p sin
A20.2.6
total D
162 0 c0 4 r 2
0 r 2
The time-averaged total energy density total is then given by

4
1
2 Q 4 2 2
2
2
total D
p0 sin D
p0 sin
322 0 c0 4 r 2
20 r 2
A20.2.7
In these equations p is the instantaneous magnitude and p0 the magnitude of the amplitude of the oscillating electric dipole moment. We recall that total is the time-averaged
total energy density passing through a surface of unit area perpendicular to the propagation direction and has units of J m3 . Substitution of eqs. (A20.2.2) and (A20.2.3)
559
in eq. (A17.2.68) gives, for the instantaneous value of the Poynting vector, inst in the
propagation direction defined by the polar basis vector ei

4
1
2 2
2 c0 Q 4 2 2
inst D
p sin ei D
p sin ei
A20.2.8
162 0 c0 3 r 2
0
r2
Using eqs. (A20.2.2) and (A20.2.3) with eq. (A17.2.69) gives for the time-averaged value
of the Poynting vector

4
2c
1

Q 4 2 2
0
2
2
D
p
sin
e
D
p0 sin ei
A20.2.9
0
i
322 0 c0 3 r 2
20 r 2
is the time-averaged flow of energy per unit area per unit time in the propagation
direction. It has units of W m2 and corresponds to the irradiance of the surface. This
quantity is also given by the product c0 total .
The Poynting vector is entirely radial, as is shown in Fig. A20.1. Since its angular
dependence is determined by sin2 , its magnitude is zero along the axis of the dipole, is
a maximum in the equatorial plane, and has axial symmetry about the dipole axis, as can
be seen from Fig. A20.2.
The Poynting vector varies as r 2 , a condition which is necessary for conservation of
energy, since under steady-state conditions the energy flow through any given solid angle
must be the same for all r. The Poynting vector also depends on the fourth power of the
frequency (or wavenumber Q ) and on the square of the magnitude of the amplitude of
the oscillating dipole moment.
We may use eq. (A20.2.9) for the time-averaged value of the Poynting vector to calculate the time-averaged energy per second or time-averaged power from an oscillating
dipole which passes through an area dA at distance r from the centre of the dipole
where r makes an angle with the dipole direction (or, alternatively, the power into the
corresponding solid angle d
).
For simplicity we denote the time-averaged power by d (rather than d). Thus
d D j jdA
A20.2.10
and substituting for j j, using eq. (A20.2.9), we obtain

d D
Since
4 p0 2 sin2 dA
2 c0 Q 4 p0 2 sin2 dA
D
322 0 c0 3 r 2
20 r 2
dA
D d
r2
A20.2.11
A20.2.12
where d
is an element of solid angle, we may also write
d D
2 c0 Q 4 p0 2 sin2 d
4 p0 2 sin2 d
D
322 0 c0 3
20
A20.2.13
560
The Raman Effect
To obtain the total power radiated by the dipole we must integrate over d
. Thus, for
example

4 p0 2
sin2 d
A20.2.14
D
322 0 c0 3
and putting
d
D sin d d
A20.2.15
we have
4 p0 2
D
322 0 c0 3
2
sin3 d d
A20.2.16
The integral has the value 8/3 and hence

43 c0 Q 4 2
4 p0 2
D
p0
A20.2.17
120 c0 3
30
It follows from eqs. (A17.2.62) and (A20.2.13) that the radiant intensity I produced by
an oscillating electric dipole in a particular direction is given by
D
ID
d
D k0 4 p0 2 sin2 D k0Q Q 4 p0 2 sin2
d
where
k0 D
k0Q D
1
322 0 c0 3
2 c0
A20.2.18
A20.2.19
A20.2.20
20
The radiant intensity defined in eq. (A20.2.18) is often described rather loosely, as the
intensity of a source in a given direction.
If photon rates are used as a measure of radiant intensity then equation (A20.2.18)
must be replaced by the following
D k0
3 2 2
Q 3 2 2
p0 sin d
p0 sin d
D k0Q
h
hc0
A20.2.21
where is the number of scattered photons of energy h D hc0 Q in a solid angle d
falling on the detector per second.

Measurements of radiated power usually involve collection over a finite solid angle.
Since d is angle-dependent, the total power in a given finite solid angle
is then
given by integrating eq. (A20.2.13) over the appropriate range of and values. That is,
C C
4 p0 2
sin3 d d
D
322 0 c0 3

2 c0 Q 4 p0 2 C C 3
sin d d
A20.2.22
D
20

Chapter A17 Section A17.2.3 treats photon rates.
561
This can lead to a rather complicated angular dependence of in certain situations

in light scattering, particularly where contributions to the scattering from two dipoles at
right angles to each other have to be considered.
Equations (A20.2.11), (A20.2.13), (A20.2.18), and (A20.2.21) can be expressed in the
useful forms:
dA
r2
2
28 4
2
d D 1.671 10 Q p0 sin d
d D 1.671 1028 Q 4 p0 2 sin2

I D 1.671 1028 Q 4 p0 2 sin2
2
A20.2.24
A20.2.25
D 8.414 10 Q p0 sin d
50 3
A20.2.23
A20.2.26
where is in W, I in W sr1 , Q in cm1 , p0 in C m, in photon counts per second and

d
D dA/r 2 in sr.
A20.3 THE OSCILLATING MAGNETIC DIPOLE AS A SOURCE

The magnetic dipole has been defined in Chapter A12. An oscillating magnetic dipole
is also a source of electromagnetic radiation, and it can be shown that provided the
dimensions of the magnetic dipole are small compared with r, and that r is much greater
than , the electric field strength E is given by
1/2 2

r
0
m0
ED
cos t
sin ek
A20.3.1
0
4c0 2 r
c0
the magnetic induction B by
BD

r
0 2 m0
t

sin ej
cos
4c0 2 r
c0
A20.3.2
and the time-averaged Poynting vector

D
0 4 m02
sin2 ei
322 c0 3 r 2
A20.3.3
where m0 is the amplitude of the oscillating magnetic dipole.

Comparing eqs. (A20.3.1) and (A20.3.2) with eqs. (A20.2.2) to (A20.2.4) we see that
m0
A20.3.4
Emagnetic dipole D Belectric dipole
p0
and
Bmagnetic
dipole
D 0 0
m0
Eelectric
p0
dipole
A20.3.5
It can be seen from eqs. (A20.3.1) and (A20.3.2) that, for magnetic dipole radiation,
E is azimuthal and B lies in a plane passing through the polar axis, in contrast to the
562
The Raman Effect

B
z
E
Figure A20.3 Relative orientations of E, B and for a magnetic dipole along the z axis.
The unit polar vectors ei , ej , and ek are defined in Fig. A20.1
situation for an electric dipole. Also, the sign of E for the magnetic dipole is opposite to
that of B for the electric dipole.
The relative orientations of E, B and for magnetic dipole radiation are shown in
Fig. A20.3.
A20.4 THE OSCILLATING ELECTRIC QUADRUPOLE AS A

SOURCE
The electric quadrupole has been defined in Chapter A13. An oscillating quadrupole is
also a source of electromagnetic radiation. We illustrate the properties of an electric
quadrupole by considering the special case of a linear quadrupole. Such a quadrupole
is made up of two dipoles of opposite polarity arranged in a line to give three charges,
CQ, 2Q and CQ as shown in Fig. A20.4 where the dipoles lie along the z axis. Although
the electric dipole moment of such a charge distribution is zero the electric quadrupole
moment zz is not.
The following expressions result for E and B provided the dimensions of the electric
quadrupole are small compared with r and and r is much greater than :

3 zz0
r
ED
sin cos ej
A20.4.1
sin t
40 c0 3 r
c0
and

r
0 3 zz0
sin t
sin cos ek
BD
4c0 2 r
c0
A20.4.2
where zz0 is the amplitude of the oscillating electric quadrupole. E and B for electric
quadrupole radiation have some features in common with E and B for electric dipole
radiation. The relative orientations of E and B and the dependence on r 1 are the same
in both cases. However, the angular distribution is different. For electric quadrupole
radiation, both E and B are zero, not only along the axis of the quadrupole D 0 or
563
z
+Q
+s
2Q
s
y
+Q
Figure A20.4 Linear electric quadrupole formed of two dipoles of opposite polarity, one
above the other. The dipole centred at s/2 has a moment of Qs and the dipole centred
at Cs/2 has a moment of CQs .
z
Figure A20.5 Radiation pattern for a vertical oscillating electric quadrupole at the origin.
The amplitude of E or of B in any given direction is proportional to the distance between
the origin and the outer surface in that direction. The inner surface is a similar plot of the
magnitude of . There is no field along the axis or along the equator of the quadrupole. The
maximum field intensity occurs along the surface of a cone at 45 to the axis.
564
The Raman Effect
where neither constituent dipole radiates but also along the equator D /2 where the
fields of the constituent dipoles are equal and opposite; the maximum field strength occurs
along the surface of a cone at 45 to the z axis. A plot of the amplitudes of E and B is
shown in Fig. A20.5
The dependence on is also different. For the electric quadrupole, E and B depend
on 3 .
The time-averaged Poynting vector is given by
2
6 zz
0
D
sin2 cos2 ei
2
5
2
32 0 c r
A20.4.3
is again radial and a plot of the magnitude of is included in Fig. A20.5. depends
on 6 .
A20.5 SCATTERING FROM CHIRAL MOLECULES

In light-scattering the radiation has its origin in oscillating multipoles induced in the
molecule by incident electromagnetic radiation. For the great majority of cases it is sufficient to consider only the induced electric dipole as a source and then the intensity of the
scattered radiation can be found using eq. (A20.2.18).
However, for a limited number of cases, as for example scattering by chiral systems, it
is necessary to consider not only the induced electric dipole but also the induced magnetic
dipole and the induced electric quadrupole. Since all three are properties of one and the
same molecule, the electric fields they radiate are phase-related and thus their amplitudes
must be added before the scattered power and intensity can be calculated. It should be
noted that the sign of the electric field vector associated with electromagnetic radiation
from an oscillating electric dipole is the opposite of the sign of the electric field vector
associated with radiation from an oscillating magnetic dipole or an oscillating electric
quadrupole. This situation arises in the case of scattering by chiral molecules and is
developed in detail in Chapter 10.
565
A21
Polarization of Electromagnetic
Radiation
La luce e lallontanarsi dalloscurita
Lombra e il celarsi della luce.
Leonardo da Vinci
A21.1 INTRODUCTION
In this chapter we consider various aspects of the polarization of electromagnetic radiation.
We treat first the several states of polarization of monochromatic radiation and introduce
the Stokes parameters. We then consider the states of polarization of quasi-monochromatic
radiation. We conclude by discussing the changes in the state of polarization produced by
scattering of electromagnetic radiation incident upon a material system.
A21.2 STATES OF POLARIZATION: MONOCHROMATIC

RADIATION
A21.2.1 Linear polarization
Let us consider a plane electromagnetic wave which is propagating along the z axis and
has the electric vector parallel to the x axis and the magnetic vector parallel to the y axis.
Such a wave, and the radiation associated with it, is often described as plane polarized,
In this chapter it is convenient to use electric vector as a shorthand for electric field vector E and magnetic
vector for the magnetic induction vector B.
566
The Raman Effect

y
x
Ex 0
Figure A21.1 A linearly polarized plane electromagnetic wave: direction of vibration of

electric vector x direction; plane of vibration of electric vector xz plane; propagation
direction z; amplitude Ex0 .
but this description is complete only if the plane is defined. Historically, the plane of
polarization was taken to be the plane containing the direction of propagation and the
magnetic vector. The direction of the magnetic vector was then referred to as the direction
of polarization. However, the plane of polarization has also sometimes been defined in
terms of the electric vector. Modern practice in terminology, which will be followed here,
is to avoid the terms plane of polarization and direction of polarization. Instead, direction
of vibration and plane of vibration are used to denote, respectively, the direction of a
field vector and the plane containing the field vector and the direction of propagation,
the vector in question being specified in each case. Thus, for the plane wave shown in
Fig. A21.1 the direction of vibration of the electric vector is the x direction and the plane
of vibration of the electric vector is the xz plane. The wave is said to be linearly polarized.
A21.2.2 Elliptical and circular polarization

Waves with other states of polarization exist. They may be regarded as the result of
combining together two plane waves of the same frequency propagating along the same
direction, say the z axis, with the direction of vibration of the electric vector in the x
direction for one wave and in the y direction for the other wave. The resultant state of
polarization depends on the amplitudes and phases of the two plane waves.
Using eq. (A17.2.9), we may write for two such waves
Ex D Ex0 cos C x D Re Ex0 expfi C x g
A21.2.1
Ey D Ey0 cos C y D Re Ey0 expfi C y g
A21.2.2
and
where

z
D t
c0
A21.2.3
Polarization of Electromagnetic Radiation
567
After some mathematical manipulation, it can be shown that at a given position along
the z axis, the locus of the points whose coordinates are Ex , Ey is, in general, an ellipse
whose characteristics depend on the amplitudes, Ex0 and Ey0 , and D y x the difference in phase of the two waves. In consequence, such waves are said to be elliptically
polarized.
In general, the major and minor axes of the ellipse lie not along the initially chosen x
and y directions, but along a new set of rectangular axes, x 0 and y 0 , where the x 0 direction
is the direction of the major axis of the ellipse and makes an angle 0 < with the
x direction (see Fig. A21.2). Then we may write for the relationship between the electric
field components Ex0 , Ey 0 and the components Ex , Ey
Ex0 D Ex cos
C Ey sin
Ey 0 D Ex sin
A21.2.4
C Ey cos
A21.2.5
and for the time dependence of Ex0 and Ey 0 ,

Ex0 D Ex00 cos C 0 D Re Ex00 expfi C 0 g
0
A21.2.6
0
Ey 0 D Ey00 sin C D Re iEy00 expfi C g
A21.2.7
where Ex00 and Ey00 are the semi-minor axes of the ellipse in the x 0 , y 0 axis system,
Ex00 Ey00 , and 0 is a phase factor.
Following the traditional nomenclature, the polarization is said to be right-handed when,
to an observer looking in the direction from which the radiation is coming, the end point of
the electric vector would appear to describe the ellipse in the clockwise sense. Likewise,
the polarization is said to be left-handed if, under the same conditions of observation,
y
2Ey 0
2Ex
Figure A21.2 The vibrational ellipse for the electric vector of an elliptically polarized
wave propagating along the positive z axis, (out of the paper).
568
The Raman Effect
the end point of the electric vector appears to describe the ellipse anticlockwise. In
eq. (A21.2.7) the upper and lower signs refer to right and left elliptical polarization
respectively as may be established by considering the values of Ex0 and Ey 0 for z D 0
(a) when t D 0 and (b) when t D /2 (i.e. a quarter of a period later).
The relationships between the various quantities in the two axis systems may be found
by appropriate manipulation, and are most conveniently expressed using two auxiliary
angles, 0 /2 and /4 /4, defined as follows:
tan D
and
tan D
Ey0
Ex 0
A21.2.8
Ey00
A21.2.9
Ex00
The numerical value of tan represents the ratio of the axes of the ellipse and the upper
and lower signs refer to right and left elliptical polarization, respectively. (Note that C
and i refer to right elliptical polarization). The required relationships are then
E2x0 C E2y0 D E2x0 C E2y 0
0
tan 2
A21.2.10
D tan 2 cos
A21.2.11
sin 2 D sin 2 sin
A21.2.12
We see that, in general, the polarization ellipse can be characterized by three independent quantities: e.g. the amplitudes Ex0 and Ey0 and the phase difference , or the major
and minor axes Ex00 and Ey00 and the angle which specifies the orientation of the ellipse,
or I the irradiance and the angles and .
The various cases of elliptical polarization are shown in Fig. A21.3. Two special cases
of elliptical polarization are of importance, namely, when the ellipse degenerates into a
straight line or a circle.
q = 0
p
0 < q < 2
p
q = 2
2 <q<p
q = p
3p
p < q < 2
3p
q = 2
3p
2 < q < 2p
Figure A21.3 Elliptical polarization with various values of the phase difference .
569
It can be seen from Fig. A21.3 that the ellipse will reduce to a straight line if
and then
D y x D m m D 0, 1, 2 . . .
A21.2.13
Ey
Ey
D 1m 0
Ex
Ex 0
A21.2.14
This is clearly a case of linear polarization. If the axis system is realigned so that, say,
the x axis lies along the line of vibration of the resultant electric vector, there is just
one component of the electric vector, namely Ex0 . There is also only one component of
the magnetic vector, the orthogonal component By . This, then, corresponds exactly to the
linearly polarized wave we treated above. It can be seen that any linearly polarized wave
can be considered to be the sum of two component waves propagating along the same
direction, with their electric vectors in phase and linearly polarized in directions which
are perpendicular to each other and the direction of propagation.
The ellipse degenerates into a circle if
Ey0 D Ex0 D E0
A21.2.15
and
D y x D m/2 m D 1, 3, 5 . . .
A21.2.16
We then have the case of circular polarization. The polarization is right-handed if

sin > 0
A21.2.17
D /2 C 2m m D 0, 1, 2 . . .
A21.2.18
so that
and thus
Ex D E0 cos C x D Re E0 expfi C x g
A21.2.19
and
Ey D E0 cos C x C /2 D E0 sin C x D Re iE0 expfi C x g
A21.2.20
The polarization is left-handed if
sin < 0
A21.2.21
so that
D /2 C 2m m D 0, 1, 2 . . .
A21.2.22
and thus
Ex D E0 cos C x D Re E0 expfi C x g
A21.2.23
Ey D E0 cos C x /2 D E0 sin C x D Re iE0 expfi C x g
A21.2.24
and
570
The Raman Effect
Thus, in the exponential notation,

Ey
D expfi/2g D i
Ex
A21.2.25
where the upper and lower signs correspond to right and left circularly polarized radiation,
respectively. More generally in this notation, for right-handed elliptical polarization the
ratio Ey /Ex has a negative imaginary part, whereas for left-handed elliptical polarization
the imaginary part is positive.
The time dependence at a particular point along the propagation direction of the electric
vector of radiation in a general polarization state can be expressed as
E D 12 EQ 0 exp it C EQ 0 exp it
A21.2.26
where E is always real but the amplitude EQ 0 (conjugate complex EQ 0 ) can be complex.
For example eq. (A21.2.26) yields E D ex E0 cos t (linear polarization) if EQ 0 D E0 ex and
E0 D 12 E0 ex cos t ey sin t (right circular polarization) if EQ 0 D E0 ex iey .
A21.2.3 Stokes parameters

It is convenient to characterize the states of polarization by Stokes parameters which
all have the same dimensions. For a plane monochromatic wave there are four Stokes
parameters. In the case of a wave propagating along the positive z axis with real amplitudes
these are
S0 D E2x0 C E2y0
A21.2.27
E2x0
A21.2.28
S1 D
E2y0
S2 D 2Ex0 Ey0 cos
A21.2.29
S3 D 2Ex0 Ey0 sin
A21.2.30
The Stokes parameters have units of V2 m2 . Only three of them are independent quantities
as they are related by the identity
S02 D S12 C S22 C S32
A21.2.31
For complex amplitudes, eqs. (A21.2.27) to (A21.2.30) take the form

Q x C E
Q y0 EQ y
S0 D EQ x0 E
0
0
A21.2.32
Q x E
Q y0 EQ y
S1 D EQ x0 E
0
0
Q
Q
Q
Q
S2 D Ex0 Ey0 C Ey0 Ex0
A21.2.33
Q x0 EQ y
iE
0
A21.2.35
S3 D
Q x
Q y0 E
E
0
A21.2.34
We see that S0 is related to the irradiance of the radiation I , as follows:

S0 D
2I
D 20 c0 I
c0 0
A21.2.36
571
S1 , S2 and S3 can be shown to be related to the angles
and as follows:
S1 D S0 cos 2 cos 2
A21.2.37
S2 D S0 cos 2 sin 2
A21.2.38
S3 D S0 sin 2
A21.2.39
Hence
1
S0
A21.2.40
c0 0 S0 D
2
20 c0
1
D tan1 S2 /S1
A21.2.41
2
S3
1
A21.2.42
D tan1 2
2
S1 C S22 1/2
It follows from eqs. (A21.2.27) to (A21.2.30) that for linearly polarized radiation (
zero or an integral multiple of ), S3 D 0; for right-handed circular polarized radiation
Ex0 D Ey0 , D /2, S1 D S2 D 0 and S3 D S0 ; and for left-handed circular polarized
radiation Ex0 D Ey0 , D /2, S1 D S2 D 0 and S3 D S0 .
Equations (A21.2.37) to (A21.2.39) indicate that a simple geometrical representation is
possible for all states of polarization: S1 , S2 and S3 may be regarded as the cartesian coordinates of a point P on a sphere of radius S0 such that 2 and 2 are the spherical angular
coordinates of this point (see Fig. A21.4). Thus, to every possible state of polarization of
I D
S3
P
S0
2c
0
2y
S2
y
S1
Figure A21.4 Poincares representation of the state of polarization of a monochromatic

wave (the Poincare sphere).
572
The Raman Effect
a plane monochromatic wave of a given irradiance (S0 constant) there corresponds one
point on the sphere, and vice versa. This geometrical representation was first introduced
by Poincare and is called the Poincare sphere. Since is positive or negative according
to whether the polarization is right-handed or left-handed, right-handed polarization is
represented by points on the sphere above the equatorial plane and left-handed polarization by points below this plane. Linearly polarized radiation D 0 is represented by
points on the equatorial circumference. Right-handed circular polarization D /4 is
represented by the north pole and left-handed circular polarization D /4 by the
south pole.
A21.2.4 Stokes parameters for scattered radiation

Stokes parameters are used in Chapters 5 and 10 to obtain general formulae for the
intensity and polarization properties of scattered radiation which cover all polarization
states of the incident radiation. We now illustrate how the definitions of Stokes parameters
given in Section A21.2.3 are adapted for the generation of such formulae.
Consider, for example, the case of the illumination-observation geometry where the
incident radiation is along the positive z axis (n0i D ez ) and observation is along the
positive x axis (n0s D ex ) so that D /2. The relevant amplitudes of the electric field
strength of the scattered radiation are then Esy0 ?s and Esz0 jjs . The associated Stokes
parameters for the scattered radiation are S0s /2 and S1s /2 and adapting eqs. (A21.2.27)
and (A21.2.28) for the geometry under consideration we have
S0s /2 D Esy0 2 C Esz0 2
A21.2.43
S1s /2 D Esy0 2 Esz0 2
A21.2.44
Esy0 2 D 12 fS0s /2 C S1s /2g
A21.2.45
Esz0 2 D 12 fS0s /2 S1s /2g
A21.2.46
and
The next stage is to convert Esy0 2 and Esz0 2 to intensities. Using eqs. (A17.2.62),
(A20.2.20) and (A17.2.69) it follows that Iy the intensity of the scattered radiation propagating along the direction defined by ex and linearly polarized with Esy0 6D 0 is given
by
Esy0 2 c0 0 x 2
Iy D
A21.4.47
2
where x is the distance from the origin O. We recall from eq. (A20.2.2) that Es 2 depends
on x 2 so that Iy , which is the scattered power per unit angle is independent of x.
We see from Chapter 5, Section 5.6 that the Stokes parameters for scattered radiation
given there are defined in terms of Ei0 , the amplitude of the electric field of the incident
radiation. To convert Ei0 2 to I , the irradiance of the incident radiation, we use the
relation
2I
A21.4.48
Ei0 2 D
c0 0
573
It follows from eqs. (A21.2.47) and (A21.2.48) that the transformation from formulae
like eqs. (5.6.1) and (5.6.2) which involve Es0 2 and Ei0 2 to formulae with I and I
simply involves discarding x 2 and substituting I for Ei 2 . For example, I/2; ?s , pi
is obtained from Esy0 2 using eq. (A21.2.45), substituting for S0s /2 and S1s /2 using
eqs. (5.6.1) and (5.6.2) and making the conversions to I and I .
These procedures are readily adapted to other illumination-observation geometries and
other states of polarization of the scattered radiation.
A21.3 STATES OF POLARIZATION: QUASI-MONOCHROMATIC

RADIATION
Strictly monochromatic radiation is always completely polarized, i.e. the end point of the
electric (and also the magnetic) vector at each point in space moves periodically around
an ellipse which may, in special cases, reduce to a circle or a straight line. In practice,
we usually have to consider radiation that is only nearly monochromatic, containing
frequencies in a small range centred on what is apparently a monochromatic frequency.
Such waves are called quasi-monochromatic and can be represented by a superposition
of strictly monochromatic waves of various frequencies.
Quasi-monochromatic radiation has an extra degree of freedom in its permitted polarizations. At one extreme, the resultant electric vector can have the properties of a completely
monochromatic wave and it is then said to be completely polarized. At the other extreme,
the resultant electric vector can have no preferred directional properties and the radiation is said to be completely unpolarized or natural. In many instances, particularly with
radiation produced by scattering, the radiation is partially polarized and consists of a
completely polarized part and a completely unpolarized part.
Natural light has the property that the intensity of its components in any direction
perpendicular to the direction of propagation is the same; and, moreover, this intensity
is not affected by any previous retardation of one of the rectangular components relative
to another, into which the light may have been resolved. Natural light of irradiance I is
equivalent to the superposition of any two independent linearly polarized waves each of
irradiance I /2 with their electric vectors vibrating in two mutually perpendicular directions at right angles to the direction of propagation. The independence of these two waves
means that there exists no specific phase relationship between them. In consequence, the
results of the interaction of natural light of irradiance I with a system can be calculated
by considering the effects of two independent orthogonal electric field amplitudes, Ex0
and Ey0 , where
1
I
1
c0 0 E2x0 D c0 0 E2y0 D
A21.3.1
2
2
2
Natural light of irradiance I is also equivalent to two independent circularly polarized
waves, one right-handed and the other left-handed, each of irradiance I /2, or indeed to
the superposition of any two independent orthogonal polarization states.
The polarization properties of such quasi-monochromatic radiation can be calculated
following the procedures used in Section A21.2.2, but eqs. (A21.2.1) to (A21.2.3) must
574
The Raman Effect
be replaced by
Ex D Ex0 t cosf C x tg
A21.3.2
Ey D Ey0 t cosf C y tg
A21.3.3

z
D t
A21.3.4
c0
where is the mean frequency of the radiation. In these equations the amplitudes and
the phases are now time-dependent. As a consequence, for the Stokes parameters of a
quasi-monochromatic wave, time-averaged quantities such as
with
Ex0 t and Ex0 tEy0 t cos t with t D y t x t
must be used in eqs. (A21.2.27) to (A21.2.30), and similarly for eqs. (A21.2.32) to
(A21.2.35).
In general, the Stokes parameters are now four independent quantities and only in the
special case of complete polarization does the relationship of eq. (A21.2.31) exist. For
completely unpolarized or natural radiation, S1 D S2 D S3 D 0.
As already emphasized, complete polarization implies a strictly monochromatic wave
for which the amplitudes and phases of the component waves are independent of time,
and hence eqs. (A21.2.27) to (A21.2.30) or (A21.2.32) to (A21.2.35) define the Stokes
parameters.
The Stokes parameters may be determined as follows: S0 from I the irradiance of the
radiation; S1 from the difference in irradiance of radiation transmitted by analysers that
accept linear polarization with azimuth D 0 and D /2, respectively; S2 from the
difference in irradiance of radiation transmitted by analysers which accept linear polarization with azimuths D /4 and D 3/4, respectively; and S3 from the additional
irradiance transmitted by a device that accepts right circularly polarized radiation over
that transmitted by a device that accepts left circularly polarized radiation.
We now consider the decomposition of a partially polarized wave into mutually independent unpolarized and polarized portions, using Stokes parameters. Since, for an unpolarized wave, S1 D S2 D S3 D 0, we can write the following sets of Stokes parameters
for the polarized and unpolarized waves:
unpolarized wave : S0 S12 C S22 C S32 1/2 , 0, 0, 0
polarized wave :
S12
S22
S32 1/2 , S1 , S2 , S3
A21.3.5
A21.3.6
where S0 , S1 , S2 and S3 are the Stokes parameters of the undecomposed partially polarized
radiation. We can now define a degree of polarization P of the original radiation as
I polarized
A21.3.7
PD
I total
Using eqs. (A21.3.5) and (A21.3.6), we obtain
PD
S12 C S22 C S32 1/2

S0
A21.3.8
575
For natural radiation P D 0 and for completely polarized radiation P D 1. We note also
that in place of eq. (A21.2.39) we now have
sin 2 D
S12
S3
C C S32 1/2
S22
A21.3.9
but the angle is still defined by eq. (A21.2.41).

The Stokes parameters provide a very convenient basis for a unified treatment of a
radiation problem which embraces all polarization states. Such a general treatment can
then be reduced to specific cases of polarization, as required. For example, we note that
an alternative representation of the four Stokes parameters for radiation of irradiance I
and degree of polarization P is
S0 D E20
A21.3.10
S1 D PE20 cos 2 cos 2
A21.3.11
PE20 cos 2 sin 2

PE20 sin 2
A21.3.12
I D 12 c0 E20
A21.3.14
S2 D
S3 D
A21.3.13
where
Thus if a problem, such as the scattering of radiation, is treated for the general case
of partially polarized incident radiation defined by the four Stokes parameters given by
eqs. (A21.3.10) to (A21.3.14), the results for specific cases of polarization can be obtained
by inserting into the formulae resulting from the general treatment the appropriate values
of P, and . For example, for linearly polarized radiation, P D 1, D 0 and is the
angle made by the electric vector with the x axis. For right circularly polarized radiation,
P D 1, D /4 and D 0; and for left circularly polarized radiation, P D 1, D /4
and D 0. For right elliptically polarized radiation P D 1, and
and determine the
orientation and ellipticity of the ellipse. For the corresponding left elliptically polarized
radiation, we have P D 1, and . For natural light, P D 0.
A21.4 CHANGE OF POLARIZATION: DEPOLARIZATION

RATIOS, REVERSAL COEFFICIENTS AND DEGREES OF
CIRCULARITY
When electromagnetic radiation interacts with a system, there is often a change in the state
of polarization. For example, incident natural radiation can produce scattered radiation
which is either completely or partially polarized; changes in the state of polarization are
also observed if the incident radiation is plane polarized or circularly polarized. These
changes are important parameters particularly in Rayleigh and Raman scattering because
they may be correlated with the symmetry of the scattering species.
576
The Raman Effect
We now consider how quantitative expression can be given to such polarization changes.
We first define a scatter plane as the plane containing the direction of propagation of
the incident radiation and the direction of observation (provided their directions do not
coincide). Then for linearly polarized incident radiation with the electric vector parallel
to the scatter plane we define a depolarization ratio ; jji for a direction of observation
lying in the scatter plane and making an angle with the direction of propagation of the
incident radiation as the intensity ratio
; jji D
I; ?s , jji
I; jjs , jji
A21.4.1
where I; ?s , jji and I; jjs , jji define the angle and relationships of the electric
vectors of the scattered and incident radiation to the scatter plane. Similarly, if the
electric vector of the incident radiation is perpendicular to the scatter plane, we can
define a depolarization ratio given by
; ?i D
I; jjs , ?i
I; ?s , ?i
A21.4.2
and for incident natural radiation we can define a depolarization ratio given by
; ni D
I; jjs , ni
I; ?s , ni
A21.4.3
It should be noted that in the older literature, degree of depolarization is used instead of
depolarization ratio. However IUPAC recommends depolarization ratio and this avoids
confusion with P, defined earlier on page 574. Figure A21.5 illustrates how these three
depolarization ratios are related to the scattered intensities for radiation incident along the
z axis of a cartesian coordinate system and observation along the x direction.
Normally in Rayleigh and Raman scattering, the depolarization ratios as defined above
have values such that 0 1 for gases and liquids. If D 0, the scattered radiation is
completely polarized; if D 1, the radiation is natural; and if 0 < < 1, the radiation is
partially polarized. In certain special cases, which arise when the scattering system has
an absorption band close to the frequency of the incident radiation, values considerably
greater than unity are observed. Such values are generally described as anomalous
degrees of depolarization. For scattering at 90 to the direction of the incident radiation,
it happens that /2; jji D 1 and hence for this geometry /2; jji can be equally well
defined by the intensity ratio I/2; jjs , jji /I/2; ?s , jji and thus the same definition
could be used for /2; jji , /2; ?i and /2; ni . However, this would lead to
> 1 for certain geometries. The definition in eq. (A21.4.1) reserves the term anomalous
depolarization for situations where values of > 1 arise only from scattering associated
with the antisymmetric part of the scattering tensor (see Chapter 7). A special case is
when is infinity, and this is termed inverse polarization.
Chapter 5, Section 5.3 defines the scatter plane in detail.

Chapter 5, Section 5.4.1 and Reference Table 5.1 define such intensity symbols in detail.
We use as a general symbol for depolarization ratio.
577
Iz
Iz
Iy
Iy
x
y
Ey
Ex
Scatter plane xz
Observation along x
r(p/2; ||i) =
Scatter plane xz
Observation along x
I(p/2; s, ||i)
I(p/2; ||s, ||i)
Iy
r(p/2; i) =
Iz
(a)
I(p/2; ||s, i) Iz
=
I(p/2; s, i) Iy
(b)
Iz
Iy
x
y
Ex
Ey
Scatter plane xz
Observation along x
r(p/2; ni) =
I(p/2; ||s, ni)

I(p/2; i, ni)
Iz
Iy
(c)
Figure A21.5 Measurement of (a) /2; jji , (b) /2; ?i and (c) /2; ni for illumination along the z axis and observation along the x axis.
578
The Raman Effect
With incident circularly polarized radiation, a change in the handedness of the radiation
can take place on scattering. We can define a reversal factor P for the radiation scattered
at an angle to the direction of the incident radiation by
P D
L s,
R i
R s,
L i
I;
I;
D
D P C 1
R s,
R i
L s,
L i
I;
I;
A21.4.4
Note that P D P C 1 since a vector turning in one sense when viewed in the
forward direction is seen to turn in the opposite sense when viewed in the backward
direction.
Another useful property is the degree of circularity C of the scattered radiation
which is defined by
C ; pi D
R s , pi I;
L s , pi
I;
D C C , pi
I; totals , pi
A21.4.5
where I; totals , pi is the total intensity of the scattered radiation and pi is the state of
R i D C ;
L i . The
polarization of the incident radiation. It should be noted that C ;
depolarization ratios, reversal coefficients and degrees of circularity can also be expressed
in terms of the Stokes parameters of the scattered radiation. For example,

I /2; jjs , pi
S /2 S1 /2

D 0
A21.4.6
s
i
S0 /2 C S1 /2
I /2; ? , p

and thus /2; jji , /2, ?i and /2; ni can be related to the Stokes parameters. Also,
S0 0 jS3 0j
D P 1
P 0 D
A21.4.7
S0 0 C jS3 0j
and
C 0; pi D
S3 0
D C ; pi
S0 0
A21.4.8
It should be noted that the Stokes parameters of the scattered radiation in eqs. (A21.4.6)
to (A21.4.8) depend on the state of polarization of the incident radiation.
We note that P 0 and in principle can range from 0 (no reversal) to 1 (complete
R s , pi D 0 and I;
L s , pi D
reversal). However, C ; pi can range from 1 with I;
s
s
i
s
i
s
i
L , p D 0 and I;
R , p D I; total , pi . When
I; total , p , to C1 with I;
R s , pi D I;
L s , pi . For
C , pi D 0 the scattered radiation is natural, that is I;
i
i
R D 1 but C ;
L D 1; and for no reversal C ;
R i D 1
complete reversal, C ;
i
L D 1. In practice the range of P and C values may be more restricted
but C ;
because symmetry does not always permit complete reversal on scattering.
Chapter 5, Section 5.4.1, and Reference Table 5.1 define intensity symbols for circularly polarized incident
and scattered radiation.
579
Further Reading
1 INTRODUCTION
This Section gives a selection of books, review articles and original papers which relate
to the topics treated in this book. It does not purport to be complete; rather it reflects
the material the author has consulted frequently in writing this book. Most of the items
cited have detailed bibliographies and so provide useful routes to the extensive literature
devoted to the Raman effect.
This Section is organised on the following basis. An initial General Section is followed
by more specific references for individual chapters.
2 GENERAL
2.1 Reviews, Journals and Conference Proceedings
The series Advances in Infrared and Raman Spectroscopy, Volumes 112, retitled
Advances in Spectroscopy, Volume 13 onwards, edited by R. J. H. Clark and R. H. Hester
(publishers initially Heyden & Son, London and subsequently John Wiley & Sons,
Chichester) contains many review articles some of which are cited later.
The six volumes of Specialist Periodical Reports on Molecular Spectroscopy, edited by
D. A. Long et al. (published by The Royal Society of Chemistry, London) also contain
review articles on Raman spectroscopy some of which are cited later.
The Journal of Raman Spectroscopy, Editor, W. Kiefer, Editor Emeritus, D. A. Long,
(published by John Wiley & Sons, Chichester) contains not only original papers but also
580
Further Reading
regular Special Issues reviewing developments in areas of Raman spectroscopy of current

interest. Some Special Issues are cited later.
The Proceedings of the International Conferences on Raman Spectroscopy, (published
by John Wiley & Sons, Chichester), provide useful accounts of current research.
2.2 Books
Anderson, A. (ed.), (1971 & 1973). The Raman Effect, Vols. 1 and 2, Marcel Dekker: New York.
Press: Cambridge.
Chalmers, J. and Griffiths, P. R. (eds.), (2002). Handbook of Vibrational Spectroscopy, John Wiley &
Sons: Chichester.
Jeffreys, H. and Jeffreys, B. S. (1962). Methods of Mathematical Physics, Cambridge University
Press: Cambridge.
Long, D. A. (1977). Raman Spectroscopy, McGraw Hill: New York.
Margenau, H. and Murphy, G. M. (1943). The Mathematics of Physics and Chemistry, D. van
Nostrand: New York.
Schrader, B. (ed.), (1995). Infrared and Raman Spectroscopy, VCH: Weinheim.
This book has a good list of earlier books dating from 1931.
3 SPECIFIC CHAPTER REFERENCES

Chapter 1
Kirkwood, J. C., Ulness, D. J. and Albrecht, A. C. (2001). Raman Spectroscopy, in Encyclopedia of Chemical Physics and Physical Chemistry, J. H Moore and N. D. Spencer (eds.), IOP
Publishing Ltd.: Bristol.
Krishnan, R. S. (1971). Historical Introduction, in A. Anderson (ed.), The Raman Effect, Marcel
Dekker: New York.
Long, D. A. (1988). Early History of the Raman Effect, International Reviews in Physical Chemistry 7, 317.
Long, D. A. (ed.), (1989). The Renaissance of Raman Spectroscopy, Chemistry in Britain, 25,
589.
Venkataraman, G. (1988). Journey into Light: The Life and Science of C. V. Raman, Indian
Academy of Sciences: Bangalore.
Chapter 3
Gussoni, M. (1980). Infrared and Raman Intensities from Electro-Optical Parameters, in Advances
in Infrared and Raman Spectroscopy, R. J. H. Clark and R. E. Hester (eds.), 6, 61. Heyden:
London.
Further Reading
581
Chapter 4
Lee, D. and Albrecht, A. C. (1985). A Unified View of Raman, Resonance Raman, and Fluorescence Spectroscopy (and their Analogues in Two-Photon Absorption), in Advances in Infrared
and Raman Spectroscopy, R. J. H. Clark and R. E. Hester (eds.), 12, 179. Wiley: Chichester.
Schrotter, H. W. and Klockner, H. W. (1979). Raman Scattering Cross Sections in Gases and
Liquids in Raman Spectroscopy of Gases and Liquids, A. Weber (ed.), Springer Verlag: Berlin.
Tang, J. and Albrecht, A. C. (1970). Developments in the Theories of Vibrational Raman Intensities in Raman Spectroscopy: Theory and Practice, H. A. Szymanski (ed.), 2, 33. Plenum Press:
New York.
Chapter 5
Chase, D. B. and Rabolt, J. F. (eds.), (1994). Fourier Transform Raman Spectroscopy, Academic
Press: New York.
Laserna, J. J. (ed.), (1996). Modern Techniques in Raman Spectroscopy, John Wiley & Sons:
Chichester.
McCreery, R. L. (2000). Raman Spectroscopy for Chemical Analysis, John Wiley & Sons: New
York.
Turrell, G. and Corset, J. (1996). Raman Microscopy: Developments and Applications, Academic
Press: New York.
Special Issue on Raman Microscopy and Imaging (1996). J. Raman Spectrosc. 27, 559636.
Special Issue on Raman Spectroscopy in Art, Medicine and Archaeology (1997). J. Raman Spectrosc. 28, 77198.
Special Issue on Raman Micro-Spectrometry and Materials Science (1999). J. Raman Spectrosc.
30, 867963.
Chapter 6
Brodersen, S. (1979). High-Resolution Rotation-Vibrational Raman Spectroscopy in Raman Spectroscopy of Gases and Liquids, A. Weber (ed.), Springer Verlag: Berlin.
Edwards, H. G. M. (1971). High Resolution Raman Spectroscopy of Gases in Essays in Structural
Chemistry, D. A. Long, L. A. K. Staveley and A. J. Downs (eds.), Macmillan: London.
Edwards, H. G. M. and Long, D. A. (1973). Rotation and Vibration-Rotation Raman and Infrared
Spectra of Gases, in Chemical Society Specialist Periodical Reports, D. A. Long, R. F. Barrow
and D. J. Millen (eds.), No. 29, Molecular Spectroscopy, 1. Chemical Society: London.
Edwards, H. G. M (1975). Pure Rotation and Vibration-Rotation Raman and Infrared Spectra
of Gases in Chemical Society Specialist Periodical Reports, D. A. Long, R. F. Barrow and
D. J. Millen (eds.), No. 29, Molecular Spectroscopy, 3. Chemical Society: London.
Heise, H. M. and Schrotter, H. W. (1995). Rotation-Vibration Spectra of Gases in Infrared and
Raman Spectroscopy, B. Schrader (ed.), 253. VCH: Weinheim.
Herzberg, G. (1950). Molecular Spectra and Molecular Structure, I. Spectra of Diatomic Molecules,
D. van Nostrand: New York.
Herzberg, G. (1945). Molecular Spectra and Molecular Structure, II. Infrared and Raman Spectra
of Polyatomic Molecules, D. van Nostrand: New York.
Hollas, J. M. (1998). High Resolution Spectroscopy, 2nd. edition, John Wiley & Sons: Chichester.
Jones, W. J. (2000). Can. J. Phys. 78, 327.
582
Further Reading
King, G. W. (1964). Spectroscopy and Molecular Structure, Holt, Rinehart and Winston: New
York.
Papousek, D. and Aliev, M. R. (1982). Molecular Vibration-Rotation Spectra: Theory and
Application of High Resolution Infrared, Microwave and Raman Spectroscopy of Polyatomic
Molecules, Elsevier: Amsterdam.
Schrotter, H. W. (1995). Raman Spectra of Gases in Infrared and Raman Spectroscopy,
B. Schrader (ed.), 277. VCH: Weinheim.
Stoicheff, B. P. (1959). High Resolution Raman Spectroscopy in Advances in Spectroscopy,
H. W. Thompson (ed.), 1, 91. Interscience: New York.
Weber, A. (1973). High Resolution Studies of Gases, in The Raman Effect, A. Anderson (ed.),
M. Decker Inc: New York.
Weber, A. (1979). High-Resolution Rotational Raman Spectra of Gases in Raman Spectroscopy of
Gases and Liquids, A. Weber (ed.), Springer Verlag: Berlin.
Weber, A. (2002). Raman Spectroscopy of Gases in Handbook of Vibrational Spectroscopy,
J. Chalmers and P. R. Griffiths (eds.), John Wiley & Sons: Chichester.
Chapters 7 and 9
Clark, R. J. H. (1975). Resonance Raman Spectra of Inorganic Molecules in Advances in Infrared
and Raman Spectroscopy, R. J. H. Clark and R. E. Hester (eds.), 1, 143. Heyden: London.
Clark, R. J. H. and Dines, T. J. (1982). Electronic Raman Spectroscopy in Advances in Infrared
and Raman Spectroscopy, R. J. H. Clark and R. E. Hester (eds.), 9, 282. Heyden: London.
Clark, R. J. H. (1984). Raman and Resonance Raman Spectroscopy of Linear Chain Complexes
in Advances in Infrared and Raman Spectroscopy, R. J. H. Clark and R. E. Hester (eds.), 11,
95. Wiley: Chichester.
Clark, R. J. H. and Dines, T. J. (1986). Resonance Raman Spectroscopy and its Application to
Inorganic Chemistry, Angew. Chem. Int. Ed. Engl. 25, 131.
Clark, R. J. H. and Stewart, B. (1979). The Resonance Raman Effect, Structure and Bonding 36,
2.
Hamaguchi, H. (1985). The Resonance Raman Effect and Depolarization in Vibrational Raman
Scattering in Advances in Infrared and Raman Spectroscopy, R. J. H. Clark and R. E. Hester
(eds.), 12, 273. Wiley: Chichester.
Kiefer, W. (1995). Resonance Raman Spectroscopy in Infrared and Raman Spectroscopy,
Lee, D. and Albrecht, A. C. (1985). A Unified View of Raman, Resonance Raman, and Fluorescence Spectroscopy (and their Analogues in Two-Photon Absorption) in Advances in Infrared
Myers, A. B. and Mathies, R. A. (1987). Resonance Raman Intensities: A Probe of Excited-State
Structure and Dynamics in Biological Applications of Raman Spectroscopy, T. G. Spiro (ed.),
2, 3. Wiley: New York.
Rousseau, D. L., Friedman, J. M. and Williams, P. F. (1979). The Resonance Raman Effect in
Raman Spectroscopy of Gases and Liquids, A. Weber (ed.), Springer Verlag: Berlin.
Special Issue Resonance Raman Scattering, Part A (1997). J. Raman Spectrosc. 28, 379466.
Special Issue Resonance Raman Scattering, Part B (1998). J. Raman Spectrosc. 29, 8571002.
Further Reading
583
Chapter 8
Ziegler, L. D. (1986). J. Chem. Phys. 84, 6013.
Ziegler, L. D., Chung, Y. C., Wang, P. and Zhang, Y. P. (1989). J. Chem. Phys. 90, 4125.
Ziegler, L. D., Chung, Y. C., Wang, P. and Zhang, Y. P. (1989). Resonance Rotational Raman
and Hyper-Raman Scattering: A Probe of Sub-Picosecond Dynamics in Advances in Infrared
Chapter 10
Barron, L. D. (1978). Raman Optical Activity in Advances in Infrared and Raman Spectroscopy,
R. J. H. Clark and R. E. Hester (eds.), 4, 271. Heyden: London.
Barron, L. D. and Svendsen, E. N. (1981). Asymmetric Light Scattering and Time Reversal in
Advances in Infrared and Raman Spectroscopy, R. J. H. Clark and R. E. Hester (eds.), 8, 322.
Heyden: London.
Barron, L. D. and Vrbancich, J. (1985). Magnetic Raman Optical Activity in Advances in Infrared
Barron, L. D. and Hecht, L. (1993). Biomolecular Conformational Studies with Vibrational
Raman Optical Activity in Advances in Infrared and Raman Spectroscopy, R. J. H. Clark and
R. E. Hester (eds.), 21, 235. Wiley: Chichester.
Hoffman, G. G. (1995). Vibrational Optical Activity in Infrared and Raman Spectroscopy,
Nafie, L. A. (1984). Experimental and Theoretical Advances in Vibrational Optical Activity in
Advances in Infrared and Raman Spectroscopy, R. J. H. Clark and R. E. Hester (eds.), 11, 49.
Wiley: Chichester.
Nafie, L. A. and Che, D. (1994) Theory and Measurement of Raman Optical Activity in Modern
Nonlinear Optics, pt. 3, M. Evans and S. Kielich (eds.), Wiley: New York.
Chapter A10
Fano, U. and Racah, G. (1959). Irreducible Tensorial Sets, Academic Press: New York.
Simmonds, J. G. (1982). A Brief on Tensor Analysis, Springer-Verlag: New York.
Chapters A11 and A12

Craik, D. (1995). Magnetism, Wiley: Chichester.
Dobbs, E. R. (1984). Electricity and Magnetism, Routledge and Kegan Paul: London.
Chapter A13
Buckingham, A. D. (1959). Molecular Quadrupole Moments, Quarterly Reviews, 13, 183.
Chapter A14
Mortensen, O. S. and Hassing, S. (1980). Polarization and Interference Phenomena in Resonance
Raman Scattering in Advances in Infrared and Raman Spectroscopy, R. J. H. Clark and
R. E. Hester (eds.), 6, 1. Heyden: London.
584
Further Reading

Corson, D. R. and Lorrain, P. (1962). Introduction to Electromagnetic Fields and Waves, Freeman:
San Francisco.
Dobbs, E. R. (1985). Electromagnetic Waves, Routledge and Kegan Paul: London.
Jackson, J. D. (1975). Classical Electrodynamics, Wiley: New York.
Chapter A19
Edmonds, A. R. (1957). Angular Momentum in Quantum Mechanics, Princeton University Press:
Princeton.
Rose, M. E. (1957). Elementary Theory of Angular Momentum, John Wiley & Sons: New York.
Zare, R. N. (1988). Angular Momentum, John Wiley & Sons: New York.

Born, M. and Wolf, E. (1975). Principles of Optics (fifth edition), Pergamon Press: Oxford.
Mathieu, J.-P. (1975). Optics, Pergamon Press: Oxford.
C.V. Raman. A copy of a portrait by Homi Bhabha, the famous Indian theoretical physicist who also had a very considerable reputation as a painter. When Bhabha handed it to Raman with the remark A scientist painted by a scientist,
Raman answered No, an artist painted by an artist.
A detail of Waterlilies by Claude Monet.
white lead and lapis-lazuli blue
(a)
lapis-lazuli blue
vermilion
(b)
Plate 5.1MTwo illuminated capitals (capital letters) from the treatise of St Augustine on the Trinity. In (a) two areas
coloured with lapis-lazuli blue and two with a superposition of white lead and lapis-lazuli blue are indicated by the
arrows. In (b) two areas coloured with vermilion are indicated by arrows.
Plate 5.2MA cross-section (enlarged approximately 10 times) of a sandstone sample from East Beacon Ridge. Six distinct layers: a, b, c, d, e, and f are identified.
585
Index
Note: Page references followed by f represents a figure, n represents a footnote and t

represents a table.
A
AIII term active modes 295
AV term Raman scattering 293296
AVI term 224227, 247250
transformation to time-dependent expression
263
AVI term Raman scattering
from molecules with totally symmetric
modes 231239
non-totally symmetric modes 239241
AVI term scattering 290
involving excited Jahn-Teller coupling 240
molecular symmetry change of the resonant
excited state 239240
absorption energy 55
absorption spectra 235f, 302
adiabatic approximation see
Born-Oppenheimer approximation
adiabatic polarizability 6667, 79
adjoint matrix 375376n
alternating tensors 427, 522
ammonia, Raman spectra of 283287
Ampères law 504
amplitude 369, 473, 517518

angular frequency 45, 14, 45, 517518
angular momentum 388
combination rules 485
coupling 541, 548
orbital 173, 466
transformation matrices 155
vectors 394, 396
vibrational 198
anharmonic oscillators 125, 269
anharmonic vibrational functions 121
anharmonic wave functions, symmetry
properties of 123
anharmonicity 4344, 122, 124
anisotropic invariants 158
anisotropy 36, 495498
anti-Stokes bands 5, 47, 182
anti-Stokes lines 6, 124
anti-Stokes Raman frequency 12
anti-Stokes Raman scattering 7, 47, 52, 118,
120, 133, 170171
anti-Stokes scattered intensity 119
anti-Stokes spectrum 180181
586
anti-Stokes transitions 119, 214, 323
associate matrix 376n
asymmetric top molecules 171, 186, 211212
axial vectors 394395, 424
B
BIII and CIII term scattering 291292
BV term scattering 296297
BVI term scattering, involving vibronic
coupling 241245
BVI terms 227229, 250262
symmetric requirement of 250
back-scattering radiation 107108, 110, 113
band shape 15
band width 15, 236
basis vectors 384385, 407408
Biot-Savart law 442, 446
Bolzmann distribution law 119
bond polarizability 3940, 128
Born-Oppenheimer (adiabatic) approximation
68, 74, 121, 153, 228, 246, 272, 298, 327
and transition polarizability tensors 6162,
535n
bra states, perturbed 70
Brillouin scattering 17
C
CV term scattering 297
CVI term 223, 229
carbon tetrachloride (CCl4 )
Rayleigh and Raman spectra of 48f
Stokes Raman spectra of 112f
cartesian axis system and related coordinate
systems, right-handed 342347
cartesian basis 8687
cartesian basis vectors 384
cartesian coordinate system 344
cartesian coordinates 92, 369, 407
cartesian tensor components 164
centrifugal stretching 177, 194
centrosymmetric molecules 204
chiral molecules, scattering from 564
chiral tensors, invariants of 315
circular intensity differential (CID) 315
circular polarization 569
left and right 9394, 132t
reversal of 107108
Index
circular polarized radiation 92, 314n
incident 106109, 114115
left-handed 565, 567
right-handed 113, 565, 567
circularity, degree of 109113114, 116, 136,
319, 323, 574577
classical electromagnetic waves 4
classical polarizability tensors 28
classical and quantum mechanics, comparison
of 2829
classical scattering tensors 35
classical theory, limitations of 45
Clebsch-Gordan coefficients 160, 340
and Wigner 3-j and 6-j symbols 541554
Clebsch-Gordan symbols 483
closure theorem 63, 66
coherent anti-Stokes hyper-Raman scattering 3
coherent anti-Stokes Raman scattering (CARS)
3, 1113, 17
coherent Stokes hyper-Raman scattering 3
coherent Stokes Raman scattering (CSRS) 3,
1113
colours, analysis of 128
column matrix 393
combination bands 43, 123
combination tones 44
complex conjugate matrix 376
complex coordinate systems 346347
complex numbers and quantities 365372
complex transpose matrix 376
Coriolis coupling 210
Coriolis forces 121
Coriolis interactions 198, 203, 205, 210
Coulomb force 435, 445
Coulombs law 442
current density vectors 440
cylindrical coordinate system 344345
cylindrical coordinates 407408
D
DIII term scattering 292
DV term scattering 297
DVI terms 223, 229230
degenerate vibrations 206
depolarization
change of 574577
degree of 100103, 109113, 115
Index
incident linear polarized radiation 138
incident natural radiation 139
Stokes parameters and 114
depolarization ratios 100n, 120, 213214, 248,
302
intensities and 279283
reversal coefficients and 168169
tensor decomposition 136
derived polarizability tensors 32, 35, 39, 112,
117, 119, 471
diatomic molecules 3638, 174186, 231
isotropic averages of polarizability tensor
components 216t
Raman spectra of 125, 170, 182
Raman vibrational transition 269
transitions of 214
dielectric constant 460n
dielectric polarization 458n
difference bands, intensities of 123
dipole electronic transitions 226f
dipole lengths 84
dipole moment operators, electric 50, 52, 66
dipole moment vectors, electric 480
dipole moments 41, 43
adiabatic permanent electric 66
electric 22, 370, 450, 454
electronic transition 228
first-order induced transition electric 4950
permanent 38
pure electronic transition electric 68
real induced transition electric 50, 52
dipole transition vectors 274
dipoles
amplitude of induced 127
permanent 40, 556
direct product 390, 392, 417
direct vector product 474, 480
direction cosines 351354, 422
isotropic averages of products of 355358
spherical components and basis vectors and
409412
direction of polarization 566
direction vectors 384
discrete resonance Raman scattering 56,
221222
divergence theorem 438, 446447
DNA, vibrational spectra of 43
587
Doppler-limited resolution 213
dual circular polarization (DCP) 316, 333, 334t
E
eigenvectors, eigenvalues and 377379
Einstein summation convention 349350, 354,
422, 477
elastic scattering 7, 17
electric current 441
electric current density 504505
electric dipole amplitude 422
electric dipole emission 309
electric dipole transition moments 57, 307
electric dipole vectors 31, 32, 452
permanent 455
electric dipoles 24, 310, 450451, 463, 554
induced 4344, 456, 564
linear induced 34
permanent 454
electric displacement 504
electric field, time-dependent 50
electric field strength 3, 435436
amplitude of 127, 422
gradient tensor 531
electric field vectors 32, 452
electric fields 4, 22, 24, 54, 461
amplitudes 50
electric displacement 460461
molecular dipoles in 454457
point charges 451454, 462466
electric and magnetic fields, electric charges
449470
electric permittivity of vacuum 433
electric polarizability component 468
electric polarization, induced 370371
electric quadrupoles 21, 463, 465
mechanism 324
operators 28, 303
perturbations 311
quantum mechanics and 2728
transition moment 25, 307
electric Rayleigh and Raman optical activity
327
electric vector, time-dependent 22
electric vectors 95, 108109
polarization 566567
of radiation 98, 570
588
electrical anharmonicity 44, 118
electrical harmonic approximation 32
electrical harmonicity 117, 123
mechanical and 118119
electromagnetic radiation 4, 25, 31, 34, 574
photon description 7
polarization of 100, 565577
sources of 555564
wavelength of 27
electromagnetic theory 397, 506, 556
electromagnetic waves 527
of circular frequency 517
electromagnetism 505
electronic absorption frequency 6263
electronic orbital and spin angular momenta
208210
electronic Raman band 294
electronic Raman scattering 49, 64, 289, 291,
297, 300, 302
intensities and polarization properties of
301302
electronic Raman spectra, selection rules in
297300
electronic resonance Raman scattering 82
electronic transition dipoles 69, 70, 74
electronic transitions 76
electronic/vibronic resonance Raman scattering
77, 8182
electrostatics 433438
energy ladders 197
energy level diagrams 78, 8283
equilibrium mean polarizability 164
equilibrium polarizability tensors 45, 109, 117,
134, 169, 473
ethene
AVI term scattering 296
symmetry operations in 240
Euler angles 489
and the rotation operator 359363
excitation frequency 233
excitation profiles 234, 236238, 301302
excitation wavelength, effect of changing 286
excitation wavenumber 97
excited electronic states 73
excited state potential, displacement of 237
excited vibronic and rovibronic states 57
exciting radiation, frequency of 75
Index
F
Faraday law of magnetic induction 504
field gradient quadrupole polarizability 465
field vectors 566
fine structure constant 126
first-order electric susceptibility 371, 459
first-order induced electric dipoles 3134
fluorescence 17
fluorescent light, frequency 6
forward scattering 10, 92, 108
Fourier transform 266
fourth-rank tensors 455
frequency 4
frequency conditions 53, 153
frequency denominators 5556, 81
frequency dependence, of first-order induced
electric dipoles 3435
fundamental transitions 236
fundamental vibrations, selection rules for
3643
G
gases, Raman spectroscopy of 213
Gausss law 437438
Gausss theorem 504
H
hamiltonian
electric 70
electronic 6869
interaction 27, 50, 79, 81, 121, 303, 449,
452, 461, 467
harmonic approximation 228
harmonic generation 535
harmonic oscillator approximation 86, 116,
119, 124
harmonic oscillator functions 117, 123
harmonic oscillator wave functions 118, 121
harmonic oscillators 123, 238
harmonic potential functions 224, 263
harmonic vibration wavenumbers 124
harmonic vibrational functions 121
harmonic wave functions 122
harmonic waves
exponential represention of 528532
monochromatic plane, in a non-absorbing
linear medium 515
Index
harmonicity, mechanical 32
hermitian matrix 376
hermitian properties 5960, 79, 81, 309310,
322
Herzberg-Teller coupling 6875, 81, 228229,
253254, 327
heteronuclear diatomic molecules 38, 172, 213
pure rotation 174175
vibration rotation 175179
homonuclear diatomic molecules 232
nuclear spin degeneracy 179180
hot bands 122, 301
hyper-Raman scattering 3, 1011, 2324, 27,
29
energy balance sheet 89t
hyper-Raman spectra 43
hyper-Rayleigh and hyper-Raman scattering,
energy transfer model 1011
hyper-Rayleigh scattering 3, 1011, 2324,
27, 29
hyperpolarizability tensors 23, 28, 423,
455456
I
ideal gas 44, 314, 457
illumination, with polarized radiation 214
illumination-observation geometry 45, 85, 96,
132, 137143t, 311312
definition of 8994, 133
intensity formulae and polarization 97113
intensity of vibrational Raman scattering
119, 233
improper rotations 424
incident circular polarization (ICP) 316, 323,
335
incident circular polarized radiation 140141t,
577
incident electromagnetic radiation 22, 24, 50,
127, 564
time-dependent electric field of 314
incident exciting radiation 55
incident linear polarizations 99
incident linear polarized radiation, intensities
and depolarization ratios for 137t138t
incident natural radiation 134, 574575
589
intensities and depolarization ratios for 139t
incident photon, absorption of 57
incident polarized radiation 26
intensities for 329t330t
normalized circular intensity differentials for
331t
incident radiation 4, 52, 89, 95, 105, 134, 154,
575
angular dependence 94
circularly polarized 106109, 315
direction of 98
dual 320
electric field strength 113
electric vector of 102
frequency of 45
intensities and depolarization ratios 103
irradiance of 96, 99, 104
linearly polarized 98102
polarization 85, 86, 9798, 101, 132, 142t
Rayleigh scattering of 47
Stokes parameters for 113116
wavenumber of 96
incident and scattered radiation 314
polarization of 314
incoherent light scattering 1929
infrared activity 36, 38, 41, 43
infrared spectroscopy 127
intensity 81, 114, 207, 301, 319
distribution 180186
expressions 154
formulae 8889, 101103, 109, 313318
of fundamental and overtone bands 233
and isotropic invariants 322324
measurements 110
of optically active Raman scattering
321326
profiles 243
of Raman transitions of symmetric top
molecules 203
of scattered radiation 90, 9497
intensity
sums and differences 315, 319320,
328329
units of 136
of vibrational Raman scattering 119, 167
interaction energy 454, 461
590
interference effects 236237
inverse matrix 375
inverse polarization 576
irradiance 10, 113, 570, 572
of exciting radiation 96
of incident radiation 95, 106, 114115
irreducible tensors 154155, 164, 298,
428432, 541
and their components 484
irreducible transformations 429
irreducible transition polarizability components
154156, 272
irrotational vector 400
isotropic averages 105, 110, 313, 499502
isotropic invariants 158
isotropic polarizabilities 37
J
Jahn-Teller coupling 228, 239, 296
Jahn-Teller effects 227
K
ket states, perturbed 70
Klein condition 55f
Kronecker delta 350, 354, 358, 425, 462,
487488
L
l-type doubling 205
lanthanide ions, electronic Raman spectra of
298
lapis-lazuli blue, Raman spectra of 129f
Laplaces equation 438
lasers 17, 20, 526
atomic transitions 15
Levi-Cività tensor 426, 522
light-scattering 339
by liquids 17
classical theory of 2224
experiments 1920, 102
optically active 316
spectra 43
spectroscopy 526
survey of 318
linear angular momentum vector 388
linear electric susceptibility 461
linear molecules 204208, 215
Index
Raman spectra of 170
rotation and vibration-rotation Raman
spectra of 204207
symmetric 40, 44
linear polarization 565566, 569
of monochromatic radiation 132t
linear polarized radiation 99
lines, shapes and width 15
Lorentz condition 447
Lorentz force, on a point charge moving in a
magnetic field 445446
Lorentzian, frequency function 252, 253f
Lorentzian distribution 236
M
macroscopic polarization, and electric
susceptibilities 457459
magnetic dipoles 309
induced 466, 564
mechanism 324
operators 303, 309
quantum mechanics and 2728
radiation 561
transition moments 25, 307
magnetic fields 24, 327, 442n
interaction of charges in motion with a static
466471
strength 504
magnetic flux 447
magnetic forces 441442
magnetic induction 442444, 505, 528, 561
divergence of 446
magnetic moment operators 28
magnetic multipole moments 21, 449, 466
magnetic quadrupoles, induced 564
magnetic Rayleigh and Raman optical activity
327
magnetic Rayleigh and Raman scattering 468
magnetic susceptibility 469
magnetic vectors 565
magnetostatics 439447
matrices
elements 541
properties of 373380
real unitary 404
Maxwell equations 446, 516, 522
in vacuum and in media 505514
Index
menthol, optically active Raman spectrum of
324
mercury arcs 15, 20, 45
molecular dipole 454
molecular polarizability 39, 41
molecular rotations 45
molecular rotors 29
molecular transition frequency 56
molecules
identification of using Raman spectra
128
moment of inertia 173, 183, 186
scattering efficiencies of 95
symmetry of 43, 166, 224, 227
momentum vectors 389
monochromatic plane harmonic waves
515536
exponential representation 528532
in homogeneous isotropic and linear
medium 532534
in vacuum 516528
monochromatic radiation 565572
multipole moments, electric 449
N
natural incident radiation 102103
natural light 572
natural radiation 132
nitrogen molecule, reduced transition tensor
for 84
non-forward hyper-Rayleigh scattering 11
non-forward Rayleigh scattering 98
non-linear molecules 41
moments of inertia of 186
non-resonance Raman scattering 537
identification of 7577
non-zero bond dipole derivatives 39, 41
normalized circular intensity differentials 319,
331t
nuclear spin 203204, 212213
O
octahedral molecules 233
operators
curl (rot) 397398
del 395, 398400
divergence (div) 397398
591
gradient (grad) 396
Laplacian 400
orders of magnitude 126127
units and 8384
orthogonal matrix 375
oscillating electric dipoles 21, 555561
oscillating electric quadrupoles 562564
oscillating magnetic dipoles 21, 561562
overlap integrals 293, 295
overtones 4344, 237
intensities of 123, 236
P
permittivity, of medium 460
perturbation theory 69, 222, 289, 292, 305
time-dependent 25, 4954, 57
phosphorescence 17
photons 136, 560
emission of scattered 57
incident 7
(C)-trans-pinane 320
beta-pinene, optically active Raman scattering
of 324
Placek approach 64
Placzek-Teller b factors 181, 202, 204, 207,
214215
Placzek invariants 100, 103, 136, 157167,
279, 494
general 120
population factor and 168
Placzek polarizability 144, 155, 185, 308, 323,
502
and optically active scattering 324326
Placzek pure vibrational transition
polarizability 67, 8588
Placzek transition polarizability 6568
Placzek type invariants 490
Placzek vibrational transition polarizability
8788, 325
Placzek-Teller b factors 183
Placzek-Teller factors 160161
for symmetric top rotor 162t
planar molecules, moment of inertia of 186
plane electromagnetic waves 557
plane harmonic electromagnetic waves 52,
524528
Pockels effect 535
592
Poincares sphere 571572
point charges, in molecular systems 450
Poissons equation 438, 447
polar and axial vectors, time even and odd
394395
polar basis vectors, polar coordinates 408
polar coordinates 345, 415
complex numbers and 369370
polar tensors 499
polarizability 54, 457
derivatives 43
ellipsoid 36, 41, 477479
mean 36, 475, 495497
perturbation of 327
principal values of 478
polarizability and dipole moment variations,
comparison of 39f
polarizability tensor components, matrix of 373
polarizability tensor operators 6263
polarizability tensors 22, 32, 39, 84, 154, 272,
422, 456, 473498
ground state 210
invariants of general 119
time dependence of 33
polarizability theory 135
polarization 81, 85
change of 574577
electric 458, 460
of electromagnetic radiation 100
elliptical and circular 566570
left- and right-handed 567, 569, 573
non-linear 469
perpendicular 101
plane of 565566
quasi-monochromatic radiation 570572
scattered radiation 90
states 94, 233
polarization vectors 89
polarized incident radiation 306
polyatomic molecules 3843, 215
pattern of vibrational Raman spectra 125
population factor 168, 203
Porto symbol 95
position vectors 395, 517
potential functions, shapes of 293
Poynting vectors 525, 559, 564
probability tensors 86
Index
prolate symmetric top molecules 186187,
189f
rotational levels of 193f
rotational term values 199f
propagation, direction of 89, 524, 565
propagation vectors 89n, 94
proper rotations 424
pseudo vectors 394
pure rotational transitions 164
intensities for 183
Q
quantum electrodynamics 16
quantum mechanical theory 21
of Rayleigh and Raman scattering 4984
quantum mechanics 2829, 45
incoherent light-scattering phenomena 24
R
radiation 4
as electromagnetic wave 16
linearly polarized 108, 114, 571, 573
out-of-phase 316
quasi-monochromatic 573
Raman
bands 5, 128, 239240, 249
excitation profiles 235f
inactivity 41, 43
lines 5, 125
resonance 12, 239240
state tensors 255262t
Raman activity 36, 3839, 41, 43, 294
conditions for 122
symmetry conditions for 297
vibrational 246
Raman, C.V. 5, 17
Raman intensities, vibrational 128
Raman and Rayleigh scattering, vibrational
133
Raman scattering 23, 64, 66, 97, 110, 117,
166167, 227
continuum 56
continuum resonance 221, 266270
Feynman diagrams for 58f
formulae for intensity of optically active 321
measurement of 112
Index
mechanism of 34
normal 5657, 96, 103, 221
normal electronic/vibronic 7678
normal (non-resonance) rotational 153
normal and resonance electronic/vibronic
289302
normal vibrational 7781
perturbation treatment of 55
Placzek theory of normal 324
pre-resonance 56
properties of normal rotational 64
pure vibrational 64
resonance rotation 63, 276t
resonance rotation and vibration-rotation 161
resonance vibration 301
rotational 49
rotational and vibration-rotation 153213
tensors 5859, 61
theory of optically active 327
types of 49
vibration and vibration resonance 302
vibration-rotation 63, 153
vibrational 31, 45, 49, 86131, 538
vibrational resonance 77, 83, 221270
Raman spectra 5, 43, 113, 170
analysis of 128
distortions in resonance 227
normal electronic 295
presentation of 4748
rotational 183, 188
rotational and rotation-vibration 154
transitions 15
vibration-rotation 169, 181
vibrational wavenumber patterns 122
Raman spectroscopy, applications of rotation
and vibration-rotation 213
Raman transition tensors
optically active 322
vibrational 68
Raman transitions 190f
irreducible tensors and intensities of
resonance rotation 279281
properties of rotational 67
pure rotational 180
symmetric electronic 302
Rayleigh and Raman scattered radiation,
intensity of 9596
593
Rayleigh and Raman scattering 310, 1516,
20, 47, 50, 109110, 456
by chiral molecules 133
by chiral systems 303327
classical theory of 3148
coherence properties of 4445
depolarization ratios 575
differences between 44
energy transfer model 710
example of 4547
generation of 52
history of 1617
mechanisms of 34
optically active 311, 326327
quantum mechanical theory of 4984
theoretical treatment of 16
transition electric dipoles 28
transition polarizability tensors 27
Rayleigh scattering 6061, 109, 117, 164,
169, 170, 225, 238
active 500
classical scattering tensors 35
frequency components 23
frequency dependence of 45
intensity of 134, 310320, 325
mechanism of 34
normal 85
optically active 307308, 310323
tensors 45, 5859
time-dependence for 309
reduced transition polarizability 84
resonance Raman scattering 57, 64, 81, 103,
539
excited state Jahn-Teller effects in
240241
identification of 7577
intensity of 56, 266
rotational and vibration-rotation 271287
time-dependent formulation of 262266
resonance scattering 5
resonant Raman intensity 241
reversal coefficients 109114, 120, 214, 215,
574577
expressions for 107, 115
of scattered radiation 169
tensor decomposition 136
594
rotation 360, 405
and vibration-rotation Raman spectra
selection rules of 167t
degeneracy factors 215
matrix element 155156
and vibration-rotation transitions 217
rotation-vibration interactions 154, 193
rotational constant, vibrational dependence of
197198
rotational degeneracy 172173, 214
rotational invariants, isotropic averages and
487498
rotational matrix, diagonalization of 406
rotational matrix element 155157
rotational partition function 172
rotational quantum numbers, selection rules for
156
rotational resonances 155
rotational transformation 416
rotational transitions 15, 87, 155, 170
rotational and vibrational terms 170172
rovibrational transitions 15
rule of mutual exclusion 41
Rydberg p orbitals 324
S
sandstone samples, Raman spectra of 130f
scalars, vectors and tensors 381382
scatter plane 91, 93f, 95
scattered beam, analyser 103
scattered circular polarization (SCP) 316, 323,
328, 335
intensities for 332
normalized circular intensity differentials for
331t
scattered cross-sections, formulae for 85
scattered intensity 32, 102, 154, 214
angular dependence of 103106
development of a symbol for 9495
formulae for 85
scattered linear polarization 99
scattered linear polarized radiation 316
scattered radiation 92, 95, 103, 108110, 577
angular dependence 94
circular polarization 107320
electric vector of 102
frequency of 13
Index
incident natural radiation 574
induced oscillating electric dipoles 21
intensity 5, 99, 105106, 132, 307
intensity and polarization of 45, 167
linearly polarized 109, 114
polarization of 5, 9798, 319
polarization vector 89
reversal coefficients of 168
spectrum of 5
Stokes parameters for 113116, 317
scattering cross-sections 9597, 103
scattering tensors 61, 311, 314
classical Rayleigh and Raman 33
Schrodinger equation, time-independent 55
second hyper-Raman scattering 3, 89t, 11,
24, 29
second hyper-Rayleigh scattering 3, 89t, 11,
24, 29
second-order electric susceptibility 459
second-order tensors 390
second-rank tensors 417, 427, 431, 462,
473474, 480499
properties of 425426
representation of 419
secondary cartesian axis system 93
secondary right-handed cartesian system 92
selection rules 58, 81, 154, 156157, 170, 276
anharmonicity affects on 123
for fundamental vibrations 3643
for overtones and combinations 4344
parity 205
symmetry 122
vibrational 86, 120123
self-adjoint matrix 376n
simple harmonic approximation 135, 172
singular matrix 375
sinusoidal phenomena, exponential
representation of 34
skew hermitian matrix 376
skew symmetric matrix 375
skew symmetry 425n
solenoidal vector 400
spectra 4, 1416
spherical basis vectors, use of 411
spherical coordinates 92, 371372, 407
spherical harmonics 298
spherical polar coordinates 345
Index
spherical symmetric molecules 168
spherical tensor components 164, 486
spherical top molecules 210211, 215
spherical vector components 273
spin degeneracy 298, 302
spin multiplicity 298
spin-orbit coupling 250, 302
spin-orbit interaction 298
square matrix 374375
square planar MX4 molecules 231
stimulated hyper-Raman spectroscopy 3
stimulated Raman gain/loss spectroscopy 3,
1314
Stokes and anti-Stokes Raman spectrum, pure
rotational 206
Stokes bands 5
Stokes hyper-Raman scattering 10
Stokes law 6
Stokes parameters 86, 114, 116, 134,
318320, 570573
expression for 115, 142t
of scattered radiation 135, 576578
states of polarization 570572
vibrational transitions 132
Stokes Raman
lines 124
spectrum 170, 181
Stokes Raman scattering 7, 34, 47, 118, 133,
171
Stokes scattering 120
Stokes transitions 118119, 214, 323
Stokes vibration-rotation spectrum 192
superconducting solenoids 442
symmetric bending 41
symmetric matrix 374
symmetric stretching 4041
symmetric tensors 145, 164165, 245, 252
symmetric top molecules 156169, 186204,
215, 271
general expression for 272273
oblate 186
pure rotation 187191
rigid 209
rotation transitions in 216
rotational energy of 215
595
selection rules for degenerate Raman bands
200t
Stokes Raman vibration-rotation spectrum
for 193
vibration-rotation 191203
vibration-rotation transitions of 215
wave function of 273
symmetric transition polarizability tensors 99,
101, 132
symmetry 109113
arguments 326
conditions for Raman activity 246247, 291
of electron distribution 36, 40
of electronic states 290
properties 110, 416
Raman activity and depolarization ratios
246270
T
tensor components 308
quadratic product of 355
tensor invariants 499501
equilibrium 184
isotropic averages and 488498
Rayleigh scattering 313
tensors 22, 23, 44, 86, 154, 165, 417431
anti-symmetric 145, 243244, 247, 249,
252, 302
axial 499
decomposition 136, 242
optical activity 499503
reduction of 475477, 481482
tetrahedral molecules 210, 232
third-order electric susceptibility 459
third-order tensor 392
third-rank tensors 418, 423, 425427, 455
time-independent wave functions 5051, 61
transformation, real unitary 404
transformation coefficients 428
transformation matrix, diagonalization of 406
transition dipole moments 224, 291
amplitudes 103, 105
components 291
derivatives 292
transition dipoles 84, 87
transition electric dipole moments 293, 474
transition electric dipole numerators 5658
596
transition electric dipole operators, hermitian
property of 59
transition electric dipole products 76f
transition electric dipoles 2526, 28, 54, 106
components 87
hermitian property of 60
induced 310311
transition electric polarizability 54
transition electric quadrupole moment tensors
28
transition energy 5, 84
transition magnetic dipole moments, induced
49
transition magnetic dipoles 28
induced 311
transition magnetic quadrupole moments,
induced 49
transition moment amplitude
time-dependent 54
time-independent complex 53
transition moments
first-order induced time-dependent 307
frequency-dependent multipole 24
induced 305
transition polarizability 49, 68, 80, 236, 242
general 5458, 67, 126
orders of magnitude for 84
scattered intensities of 127
tensorial nature and symmetry of 58
vibration-rotation 153
transition polarizability components 203204
transition polarizability invariants 99
transition polarizability tensors 28, 8687,
110, 120, 134135, 298, 473, 537540
asymmetry in 292
Born-Oppenheimer approximation and
6164
general 58, 83
invariants for 113
irreducible 154
Rayleigh and Raman scattering 27
symmetry of 121
transition tensors 60, 103, 118, 161
asymmetric 298
components 251
invariants 96, 106, 250
irreducible 158
Index
reduced 84
symmetric 323
symmetry of 292
transition wavenumbers 84, 169, 211
transitions, vibration-rotation 164166, 170
transposed matrix 374
U
uncertainty principle 51
unit vectors 383385, 390, 440, 517
unitary matrix 376
V
vectors 381406, 407416
addition, subtraction and multiplication by a
scalar 385
addition coefficients 544
angular momentum operator 542
change of and effect upon coordinates of a
fixed vector 401404
coupling coefficients 544
cross-product 386388
definition of 382383, 407
differentiation 395401
direct product of 417
dyads 390
effect of symmetry on basis vectors and
404406
field 381
multiplication of two 385390
in n-dimensional space 415416
potential 446
product 386388, 394
rotation of using spherical coordinates
412414
time derivative of 401
transformation upon rotation of axes
definition 393394
triads 418419
triple products of 390393
velocity vector 388, 395
vermilion (mercury (II) sulphide), Raman
spectra of 129f
vibration 4041, 44, 216
vibration interactions 154
vibration transition polarizability 118, 144
vibration-rotation bands, selection rules for 191
Index
vibration-rotation interaction 170, 175176,
206
vibration-rotation Raman band, general pattern
of 177
vibrational degeneracy 97, 120, 134, 214
vibrational frequency 43
vibrational matrix element 156, 158
vibrational modes 43, 196
vibrational overlap integrals 226
vibrational partition function 97, 172
vibrational quantum numbers 97, 117118,
204, 223
vibrational spectra, patterns of 123125
vibrational transition integral 291
vibrational transition polarizability 6667
vibrational transition polarizability tensors
184185, 222224
vibrational transitions 58, 68, 76, 97,
116120
vibrational wave functions 121122
orthogonality of 225, 293
properties of 85, 290
vibrational wavenumbers 86, 127
vibrations, symmetries of 164
vibrators 29, 215
vibronic absorption frequency 62, 75, 77
vibronic coupling 241
vibronic coupling integrals 228, 296
597
vibronic Raman scattering
normal electronic 289292
resonant electronic 292297
vibronic transition polarizability tensors 292
virtual absorption 7, 55
W
wave equations 515524
wave functions 51, 53, 79
adiabatic 66
perturbed 26, 28, 305
real 81, 309310, 323
rotational 166
time-dependent 5051
unperturbed 26
wavelength, of wave in vacuum 518
wavenumber shifts 123125, 134, 171, 190
wavenumber-normalized cross-sections 9697,
126, 135136
wavenumbers 1415, 45, 63, 201, 203
white lead (basic lead carbonate), Raman
spectra of 129f
Wigner 3-j coefficients 299, 340
Wigner 3-j symbols 156, 159, 274275, 277,
485, 548550, 553
Wigner 6-j symbols 277, 550554
Wigner coefficients 544
Wigner-Eckart theorem 255, 299

Long D A The Raman Effect (Wiley, 2002) (610S)

Uploaded by

Copyright:

Available Formats

Long D A The Raman Effect (Wiley, 2002) (610S)

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Long D A The Raman Effect (Wiley, 2002) (610S)

Uploaded by

Copyright:

Available Formats

The Raman Effect: A Unified Treatment of the Theory of Raman Scattering by Molecules. Derek A.

The Raman Effect

The Raman Effect

Copyright 2002 by John Wiley & Sons Ltd,

Library of Congress Cataloguing in Publication Data

Part One Theory

Introduction to Theoretical Treatments of Incoherent Light Scattering

Classical Theory of Rayleigh and Raman Scattering

Quantum Mechanical Theory of Rayleigh and Raman Scattering

Vibrational Raman Scattering

Rotational and VibrationRotation Raman Scattering

Vibrational Resonance Raman Scattering

Rotational and VibrationRotation Resonance Raman Scattering

Normal and Resonance Electronic and Vibronic Raman Scattering

Rayleigh and Raman Scattering by Chiral Systems

Part Two Appendices

A1 The Right-handed Cartesian Axis System and Related Coordinate

A2 The Summation Convention

A4 Isotropic Averages of Products of Direction Cosines

A5 The Euler Angles and the Rotation Operator

A6 Complex Numbers and Quantities

A7 Some Properties of Matrices

A13 The Interaction of a System of Electric Charges with Electric

Electric Dipole in a Molecular System

A14 The Polarizability Tensor

A16 Maxwell Equations in Vacuum and in Media

A17 Monochromatic Plane Harmonic Waves in Vacuum and in

A18 The Transition Polarizability Tensor afi

A19 ClebschGordan Coefficients and Wigner 3-j and 6-j Symbols

A20 Sources of Electromagnetic Radiation

A21 Polarization of Electromagnetic Radiation

developed in comprehensive Appendices to which cross-references are made in the main

I am grateful for permission to reproduce the material listed below.

The Raman Effect

Figures 7.147.18 inclusive from Kiefer, W. (1995). Resonance Raman Spectroscopy

A unified treatment of the theory of Raman scattering by molecules.

The Raman Effect

1.2 SOME BASIC DEFINITIONS

Chapter A17 discusses frequency and related quantities in detail.

Survey of Light-scattering Phenomena

is the magnitude of fi that is M D jfi j. The magnitude of a transition energy is

1.3 RAYLEIGH AND RAMAN SCATTERING

The Raman Effect

Survey of Light-scattering Phenomena

1.3.2 Energy transfer model

Energy level diagram

Table 1.1 (Continued)

Energy level diagram

Survey of Light-scattering Phenomena

The Raman Effect

1.4 HYPER-RAYLEIGH AND HYPER-RAMAN SCATTERING

1.4.2 Energy transfer model

Chapter A17 deals with irradiance.

Survey of Light-scattering Phenomena

1.5 SECOND HYPER-RAYLEIGH AND SECOND HYPER-RAMAN

1.5.2 Energy transfer model

1.6 COHERENT ANTI-STOKES AND COHERENT STOKES

The Raman Effect

Figure 1.2 Diagrammatic representation of CARS, CSRS, SRGS and SRLS.

Ef D Ei C h 1 2 D Ei C h M . Thus the overall process is a Stokes Raman process

where 0 is the value of at the equilibrium configuration, Qk , Ql . . . are normal