THE GEORGE FISHER BAKER

NON-RESIDENT LECTURESHIP
IN CHEMISTRY AT
CORNELL UNIVERSITY

Angular Momentum
Angular Momentum
Understanding Spatial Aspects
in Chemistry and Physics

Richard N. Zare
Department of Chemistry
Stanford University
Stanford, California

A Wiley-lntencience Publication
JohnWiley & Sons
New York Chichester Brisbane Toronto Singapore
Copyright @ 1988 by John Wiley &Sons, Inc.
All rights reserved. Published simultaneously in Canada.
Reproduction or translation of any part of this work
beyond that permitted by Section 107 or 108 of the
1976 United States Copyright Act without the permission
of the copyright owner is unlawful. Requests for
permission or further information should be addressed to
the Permissions Department, John Wiley & Sons, Inc.

Library of Congress Cataloging in Publication Data:

Zare, Richard N.
Angular momentum.

Based on lectures given to the Chemistry Dept, Comell

University, in the autumn of 1980 as part of their Baker
lecture series.
"A Wiley-Interscience Publication"
Includes bibliographies and index.
1. Angular momentum (NuclearPhysics) 1. Title
11. Title: Baker lecture series.
QC793.3.A5Z37 1987 530.1'2 87-16204
ISBN 0-471-85892-7
Printed in the United States of America
 To
Dudley R. Herschbach
 PREFACE

From merry-go-rounds to spinning tops, angular momentum has always fascinated

me, confused me, and at times filled me with dizzying excitement. This is a book
about angular momentum in quantum mechanics and its applicationsto chemistry and
physics. It grew out of twenty lectures given to the Chemistry Department, Cornell
University, in the autumn of 1980 as part of their Baker Lecture series. Thereafter
for many years I taught a graduate course on "Advanced Quantum Mechanics" in
the Chemistry Department, Stanford University, and I distributed this material to my
class in lieu of a suitable textbook.
The material in this text is intended for a one-semester graduate course. The chap-
ters are organized primarily as a way to learn about angular momentum theory, sec-
ondarily as a reference source. A working knowledge of basic quantum mechanics is
presumed, the type of knowledge gained traditionally in every graduate introductory
course on quantum mechanics. It has been my experience that angular momentum
theory often acts as a major roadblock to understanding many contemporary research
problems. Moreover, this topic seems to be difficult to learn only by reading about
it. Indeed, mastery appears to come from more than one exposure to this topic and
from solving real problems of interest. Consequently, I have placed at the end of
each chapter extensive examples and applications of the text. I would urge members
of my class to work on these so-called problem sets and applications individually,
and then collectively. I regard this material as an integral part of the text, a way to
make clear, alive, and concrete the more abstract explication of principles that come
before in each chapter. For someone to read the text and skip the problem sets and
applications is like someone reading a book on how to play the piano without ever
sitting before a keyboard.
Many problems in chemistry and physics involve homogeneous media and iso-
tropic space, for which the charms of angular momentum theory seem remote. How-
ever, as soon as we become concerned about direction in describing some process,
it is necessary to speak of angles and momenta conjugate to them. Thus angular
momentum theory arises naturally in discussing such phenomena as how a beam of
radiation interacts with matter or how a beam of atoms and molecules scatter from
vii
viii PREFACE

other collision targets. Its mastery is essential for a detailed understanding of such
microscopic phenomena. As proficiency increases in this topic, it is possible to disen-
tangle purely geometrical factors from dynamical ones. This separation,embodied in
the Wigner-Eckart theorem, may be regarded as the ultimate goal of angular momen-
tum theory whereby we exploit the full symmetry inherent in some physical process
to analyze it into its essential components.
This is not the first book on angular momentum theory, but it differs from others
in the emphasis placed on making it a learning text for those with a minimum
background in quantum mechanics. It also differs in the choice of examples that
are drawn almost entirely from atomic and molecular phenomena. I believe that it is
not possible to present this material too simply to anyone learning angular momentum
theory for the first time. Consequently, many intermediate steps are left in the text,
which, to the initiated, may appear inelegant, if not annoying. At the same time
this text serves a secondary purpose of being a reference work; the vast majority of
formulas needed to solve any problem in angular momentum theory are contained in
this book.
Angular momentum theory is central to understanding and unifying photon and
particle collision phenomena. It is hoped that what follows serves as a good prepa-
ration for experiencing the pleasures contained in the rich and growing literature on
directionality in chemistry and physics.
RICHARDN. ZARE

Stanford, California
August 1986
 CONTENTS

CHAPTER 1. Angular Momentum Operators and Wave Functions 1

1.1 Definition of Angular Momentum Operators / 1
1.2 Eigenvalues and Matrix Elements of Angular
Momentum Operators / 3
1.3 Angular Momentum Wave Functions 16
1.4 The Vector Model / 12
Notes and References / 18
Problem Set 1 / 20
Application 1 Scattering Theory / 23
Notes and References / 41

CHAPTER 2. Coupling of Two Angular Momentum Vectors

2.1 Clebsch-Gordan Coefficients / 43
2.2 Clebsch-Gordan Coefficients and 3 - j Symbols:
Symmetry Properties and Explicit Values / 48
2.3 Clebsch-Gordan Coefficients and 3 - j Symbols:
Geometric Interpretation / 5 1
Notes and References / 64
Application 2 The Wigner-Witmer Rules / 66
Application 3 The Rotational Energy Levels of a 2X
Free Radical 170

CHAPTER 3. Transformation under Rotation 73

3.1 Angular Momentum Operators as Generators of
Infinitesimal Rotations / 73
 CONTENTS

3.2 Parameterization of Rotations by Euler Angles / 77

3.3 The Direction Cosine Matrix Elements 179
3.4 Representation of the Angular Momentum Operators in
the Space-Fixed and Molecule-Fixed Frames / 81
3.5 The Rotation Matrices / 85
3.6 The Rotation Matrices: Geometrical Interpretation / 91
3.7 The Spherical Harmonic Addition Theorem / 95
3.8 The Clebsch-Gordan Series and Its Inverse / 98
3.9 Integrals Over Products of Rotation Matrices / 100
3.10 The Rotation Matrices as Rigid Body Rotational Wave
Functions / 104
Notes and References / 106
Application 4 Energy Levels of Atoms with ' h o
Valence Electrons / 111
Application 5 Angular Distribution of Rigid Rotor
Axes Following Absorption of Plane Polarized
Light / 117
Notes and References / 119
Application 6 Photofragment Angular Distribution
(Classical Treatment) / 120
Application 7 Introduction to Symmetric Tops / 123
Application 8 Polarized Resonance Fluorescence
and Polarized Raman Scattering: Classical
Expressions / 125
Notes and References / 131
Application 9 Magnetic Depolarization of Resonance
Fluorescence: Zeeman Quantum Beats and the Hanle
Effect / 133
Notes and References / 135
Application 10 Correlation Functions in Molecular
Spectroscopy / 136
Notes and References / 141

CHAPTER 4. Coupling of More Than Two Angular Momentum Vectors 143

4.1 The 6 -j and 9 - j Symbols / 143
4.2 Graphical Methods / 151
Notes and References / 172
Problem Set 2 / 173
CONTENTS

CHAPTER 5. Spherical Tensor Operators

5.1 Definition / 177
5.2 The Wiper-Eckart Theorem / 180
5.3 Spherical Tensor Products / 186
5.4 Tensor Product Matrix Elements / 192
Notes and References / 212
Problem Set 3 / 215
Application 11 The Energy Levels of a Two-Valence-
Electron Atom Revisited / 218
Application 12 Directional Correlations; Breakup of a
Long-Lived Complex / 220
Notes and References / 225
Application 13 Orientation and Alignment / 226
Notes and References / 241
Application 14 Nuclear Quadrupole Interactions / 243
Notes and References / 250

CHAPTER 6. Energy-Level Structure and Wave Functions of a Rigid

Rotor 253
6.1 Moments of Inertia: Classification of Rigid Rotors by
Top 1LpepesI 253
6.2 Classical Motion of a Free Top / 258
6.3 Symmetric and Asymmetric Top Energy Levels / 266
6.4 Nonrigid Behavior: The Van Vleck Transformation / 277
6.5 Line Strength Factors / 283
Notes and References / 291
Problem Set 4 / 294
Application 15 Introduction to Diatomic
Molecules / 297
Notes and References / 3 15
Application 16 Molecular Reorientation in
Liquids / 3 18
Notes and References / 321

APPENDIX. Computer Programs for 3-j, 6-j, and 9-j Symbols 323
Notes and References / 327

INDEX 335
THE GEORGE FISHER BAKER
NON-RESIDENT LECTURESHIP
IN CHEMISTRY AT
CORNELL UNIVERSITY

Angular Momentum
 Chapter 1
ANGULAR MOMENTUM
OPERATORS AND WAVE
FUNCTIONS

1.1 DEFINITION OF ANGULAR MOMENTUM OPERATORS

Let a particle of mass n and velocity v be located at a position r measured from

some origin. Then according to classical rnechanics[l], the particle has a linear
momentum p given by

and an angular momentum t! given by

The quantum transcription[2] of Eq. (1.2) is obtained by replacing p by (h/i) V,

where h = h/2 7r is Planck's constant divided by 2 T,i is the square root of minus
one, and V is the gradient operator whose Cartesian components are

In Eq. (1.3) (and elsewhere) a superscript caret denotes a unit vector. For conve-
nience, we drop the burden of carrying around h by introducing a system of units in
which h = 1. Thus the Cartesian components of p are
2 1 ANGULAR MOMENTUM OPERATORS AND WAVE FUNCTIONS

and those of l are

The commutator [A, B] = AB - BA of two operators A and B plays a central role

in quantum mechanics[2]; the necessary condition for the observables A and B to be
simultaneously measurable is that the corresponding operators A and B commute[3],
that is, [A, B] = 0. From Eq. (1.4) it is readily seen that the position vector of a
particle and its momentum satisfy the basic commutation relations:

with all cyclic permutations. Thus, for example, it is not possible to measure
simultaneously along the same direction the position and linear momentum of a
particle to arbitrary precision.
The commutation relations of the Cartesian components of t are also readily
derived:

Equation (1.7) has the interpretation that quantum states cannot be specified by
any more than one of the labels (eigenvalues) of the three components of angular
momentum. The "good" quantum numbers corresponding to the largest set of
mutually commuting operators represent the maximum information that can be
known about a quantum mechanical system. The measurement of another variable
corresponding to an operator not commuting with this set necessarily introduces
uncertainty into one of the variables already measured. A sharper specification of
the system is, therefore, not possible.
Because of the importance of the commutator, it is natural to define[4] a general
angular momentum operator j as one whose Cartesian components obey the commu-
tation rules

in analogy to Eq. (1.7). This extended definition returns an unexpected dividend. As

we shall see in the next section, it permits the existence of spin-a quantity that has
no classical analogy. We will reserve f for orbital angular momentum and use j for
general angular momentum.
1.2 EIGENVALUES AND MATRIX ELEMENTS

1.2 EIGENVALUES AND MATRIX ELEMENTS OF

ANGULAR MOMENTUM OPERATORS

The square of the total angular momentum is defined as

This operator has the commutation properties that

Hence we can construct states I jm) that are simultaneously eigenfunctionsof j2 and
any one component of j, say, j,; that is,

We proceed to determine the eigenvalues Xi = (jmlj2 I jm) and m = (jml j, Ijm).

The operator j: + ji = j2 - j: is diagonal in the I jm) representation. Moreover,
it has positive definite (nonnegative) eigenvalues

because the expectation value of the square of a Hermitian operator, that is, the square
of a real eigenvalue, is greater than or equal to zero. Hence we conclude that the
value of m is bounded from both above and below in that m2 cannot exceed Xi. This
implies that for a given j there exist minimum and maximum values of m, denoted
by m,i, and m,, ,respectively.
Let us introduce the raising and lowering operators j*, defined by

From Eqs. (1.8) and (1.10) it is readily shown that these operators satisfy the
commutation rules:

Let us examine the behavior of the function ji I jm). We find

4 1 ANGULAR MOMENTUM OPERATORS AND WAVE FUNCTIONS

Thus jf V m ) is an eigenfunction of j2 with the eigenvalue Xi and an eigenfunction

of j, with the eigenvalue m f 1. It follows that jf I j m ) is proportional to the
normalized eigenfunction 1j m f 1); that is

where Cf is a proportionality constant. The ability of the raising and lowering

operators jf to alter m by f1 unit; respectively, while preserving Xi gives them
their names. Note that the j* operators are also referred to in the literature as step-up
and step-down operators, ladder operators, or shifr operators.
Since the values of m are bounded between mmh and m ,, it follows that

By applying j- to Eq. (1.18) and j+ to Eq. (1.19) and by using the identity

we obtain the two equations

Elimination of A, yields

One of these two factors must vanish. Because m, _> mmh,the only solution to
Eq. (1.23) is
m, = -mmh (1.24)
1.2 EIGENVALUES AND MATRIX ELEMENTS 5

Successive values of m differ by unity [see Eq. (1.16)l. Therefore, m, - mmin

is a positive definite integer, which we may denote by 2 j, where j is an integer or
half-integer. Then from m, - m,i, = 2 j and m, + m,i, = 0 we conclude that

andthereare2j+ 1 possiblevaluesofm,m= j , j - 1, . . . , - j + 1,-jforeach

value of j. Substitution of Eq. (1.25) into Eq. (1.21) yields the additional result

We are now in a position to evaluate the proportionality constant C* appearing in

Eq. (1.17). We find

where we have used Eq. (1.20) and the property that the adjoint (complex conjugate
transpose) of j* is jF. (Note that the j* operators are not Hermitian, i.e., not self-
adjoint, although j,, j,, and j, are.) From Eq. (1.27) it is seen that the absolute value
of C* is determined but its phase is arbitrary. We choose C* to be real, that is

This agrees with the standid phase convention[5], namely, that the matrix elements
of j, are real while those of j, are purely imaginary.
In summary, we write down all the matrix elements of the angular momentum
operators in which j2 and j, are diagonal:

(jml j2 lj'm') = j ( j + 1)6jj~6mm) (1.29)

(jml j, I j'm') = rn'6,,16~~t (1.30)

(jml j* lj'm') = [ j( j + 1) - ml(m' f 1 ) ] ' ~ ~ ~1 l 6 ~ , ~(1.31)

~ ~

(jml jz 1j'm') = +[j( j + 1) - m1(m' f 1)] ~6jj~6m,m~*l (1.32)

( mj m )=i j + 1 -( m&1 ) ] j j (1.33)

In Eqs. (1.29)-(1.33) called the Kronecker delta, is defined as having the

properties that Siil = 0 for i # a' and = 1 for i = it. We introduce the practice of
placing some of the most frequently used results in a box.
6 1 ANGULAR MOMENTUM OPERATORS AND WAVE FUNCTIONS

The angular momentum quantum number j can have any of the values 0, 1, i,
$, 2 , ...in units of h. Integral values of j correspond to orbital angular momenta l,
while half-integral values of j are referred to as spin angular momenta. The use of the
commutation rules as the definition of the angular momentum operators puts orbital
and spin angular momenta on the same footing.

1.3 ANGULAR MOMENTUM WAVE FUNCTIONS

So far we have considered the angular momentum eigenfunctions Ijm) in an abstract

vector space of 2 j + 1 dimensions. For integral j, denoted by l,it is often useful to
work in an explicit coordinate representation for!A and its states [em)by introducing
the spherical polar coordinates

x = rsin 0cosd

y = r sin 0 sin 4

z = rcostl

which have the inverse relations

Then Eq. (1.5) becomes

1.3 ANGULAR MOMENTUM WAVE FUNCTIONS 7

This result is readily obtained using the nine partial derivatives in which the
appropriate Cartesian coordinates are held constant

-a r- - sin 0 cos 4, dr
- = sin 8sin 4,
ar = COS e
-
ax ay az

a4 = --sin 4
- -a$- --cos 4 a4- - 0
-
ax rsinfl' ay rsin8' dz
and applying the chain rule for differentiation. For example,

8 cos 19 sin 8
r
-+ -)I
cos 4 8
86' rsin 0 84

a + cot 0 cos 4 -

as advertised, and so on. From Eq. (1.36) it follows that

and
e2 = - -- 1 a2 + - l - a (sin
Lin2 e am2 sm e ae
oh)]
In this representation the eigenvalue problem
8 1 ANGULAR MOMENTUM OPERATORS AND WAVE FUNCTIONS

yields two partial differential equations whose solutions are the spherical harmonic
functions[2]
Yem(e, 4) /em) (1.42)

The spherical harmonics may be written as the product of two functions, one that
depends only on the polar angle 0 and the other on the azimuthal angle 4

Yem(o,4) = @em(0)@m(4)

The am($) satisfy the differential equation

and have the explicit form

These functions are normalized so that

The @em(0) satisfy the differential equation

and have the explicit form for m 2 0

@em(@=
(-1)-
2y! -
[
2 t + I (t-m)!
2 ( t + m)!
]+ cane)-

These functions are normalized so that

For many purposes it is useful to relate the Oem(8) to the so-called associated
Legendre functions

Pem(cos6) = sin 0 -]
[d(cts e) Pe(cOS0)
1.3 ANGULAR MOMENTUM WAVE FUNCTIONS 9

where

are the ordinary Legendre polynomials. Comparison of Eq. (1.48) with Eqs. (1.50)
and (1.5 1) shows that we may make the identification

[2e+ 1 ( e - m ) ! ] ;
eem(e)= ( - I ) ~ - PT( cos 8) (1.52)
2 (e+ m)!

for m 2 0 . For negative m we choose to define

so that for m < 0 is still given by Eq. (1.48) or (1.52) but with Iml in
place of m and the factor ( - 1) omitted. This choice of phase is consistent with the
previously made choice of sign concerning the matrix elements of 4, and 4,[5].
For the special case of m = 0

Because P r ( 1) vanishes except for m = 0 , it follows that

Finally, we also note that

For reference purposes we list the first few spherical harmonics:

 1 ANGULAR MOMENTUM OPERATORS AND WAVE FUNCTIONS

The spherical harmonics constitute an orthonormal set over the unit sphere

In particular for the special case m = m' = 0, we obtain the orthonomality relations
for the Legendre polynomials

iT sin B d6 Pe(cos 6) Pu(cos 6) = -

2
2e+ i6"
by substituting Eq. (1.54) into Eq. (1.58) and carrying out the integration over d4.
Equation (1.58) allows us to identify Y&Yemd12 with the probability that the position
vector r points into the solid angle element dl2 = sin BdBd4 when the expectation
value (average value) of l, is m. Note that Y&Ye,, is independent of 4. This leads
naturally to a geometric interpretation of l as a vector making a constant projection m
on the z axis when the system is in the state lem). We discuss the implicationsof this
statement more fully in the next section.
It is not possible to express the basis vectors ljm) for half-integral values of j
and m in terms of single-valued continuous functions on a unit sphere, as we just
have done for /em). However, it is possible to build up the Vm) from the spin
eigenfunctions for j = $, denoted by

and

Using Eqs. (1.29)- (1.33) we can easily show that the angular momentum operators
j,, j,, j, have the 2 x 2 matrix representation
1.3 ANGULAR MOMENTUM WAVE FUNCTIONS

where the CJ,, cry, cr, are called Pauli spin matrices. It follows that

Here a, is often called spin up and P spin down for j,a, = ( + $ ) aand j,P = ( - $) P.
The eigenfunctionsa and /I are an example of what are called spinors because of their
special transformation properties under rotation, which are discussed in Section 3.5.
Equations (1.62) and (1.63) are one particular representation of the angular mo-
mentum operators. Symbolically, the operators j+, j-, and j, may also be written as
differential operators in this spin space

from which it is readily verified in terms of these spinor differential operators

j-a = P, j-/I = 0

Combining Eqs. (1.63) and (1.64), we may express j2 in the notation of spinor
differential operators as

from which it is easily checked that j2a, = $a, and j2P = $P.
12 1 ANGULAR MOMENTUM OPERATORS AND WAVE FUNCTIONS

Using this notation, we show that an arbitrary eigenfunction Ij m ) of j2 and j, may

be constructed in terms of a and p by

(a)i+m(P)l-m
lim) =
[ ( j+ m ) ! ( j- m ) ! ] 4

This is proved at once by applying Eqs. (1.64) and (1.66) to (1.67), from which it is
found that

It follows that Eq. (1.67) provides a representation of the angular momentum eigen-
functions Ij m ) in terms of the spinor differential operators. Moreover, this result is
valid for both integral and half-integral j.

1.4 THE VECTOR MODEL

The angular momentum vector j can never point exactly along the z axis. The

*
maximum value of (j,) = m is when na takes on the value j while the length of
the vector j is given by (j .j) = (j2) 4 = [j( j + I ) ] ). This result is consistent with
the fact that there must be an uncertainty in the values of j, and j,. However, as j
becomes large, the eigenvalue of j2, which may be written as j2( 1+ 1/ j ) ,approaches
j 2 , allowing us to make the correspondence between the quantum number j and the
classical angular momentum for integral j. Indeed, the presence of the 1/j term is
a quantum mechanical effect that reflects our inability to measure simultaneously all
three components of j and hence specify precisely the direction of j.
Let us attempt to quantify this matter. The spread in the measurements of an
observable A corresponding to the Hermitian operator A is conveniently described
in terms of the variance of A. defined as

It is customary to call the positive square root of the variance of A , A A , the

uncertainty in A. In a representation that diagonalizes A, ( A A ) vanishes
~ and A
can be determined with no uncertainty, that is, to arbitrary precision in principle.
1.4 THE VECTOR MODEL

FIGURE 1.1 Vector model for the state

V m ) , which is represented by a vector j
precessing about the axis of quantization
(the z axis) making a constant projec-
tion m on it.

Let us consider the sum of the variances of j, and j, in the Vm) representation.
According to Eq. (1.69), we have ( Aj,)2 = (jz) and ( Aj,)2 = ( j i ) since (j,) and
(j,) both vanish [see Eqs. (1.32) and (1.33)]. But (j2) = ((1';) + ( j i ) + (j!); that is,

Thus the sum of the variances ( A j,)2 + ( Aji)2 is constant for a given value of m.
Moreover, the value of this sum reaches a minimum for JmJ= j, that is, when the
angular momentum vector points as nearly as possible along +z or -2.
We are thus led to a picture, called the vector model[6], in which the eigen-
state V m ) is represented by an angular momentum vector j of length [j ( j + I)] ) that
precesses about the z axis (the axis of quantization) with a constant projection m (see
Figure 1.1). Thus j moves in some uniform but unobservable manner on the surface
of a cone whose apex half-angle 8 satisfies the relation

cos 8 = m/[ j( j + I)]) (1.7 1)

In this picture the motion of j must be uniform so that j spends as much time pointing
along + x as along -x or +y as along -y, causing both (j,) and (j,) to vanish; in other
words, the projection of j along the x and y axes averages to zero. Hence a classical
orbit is ascribed to the state Vm) for integral j, that is, to the state [em). Of course,
the uncertainty principle prevents us from taking this picture too literally. However,
the vector model does suggest how to regard the state llm) in the correspondence
limit [6]. Because large values of the angular momentum are
commonplace in molecular problems, the vector model can often give insight into
the interpretation of angular momentum theory.
To explore further the implications of the vector model and the interpretation
of Jem) when the orbital angular momentum t becomes large, we develop an
14 1 ANGULAR MOMENTUM OPERATORS AND WAVE FUNCTIONS

asymptotic approximation[7] to the Legendre polynomials. Equation (1.47) may be

rewritten for m = 0 as

The substitution
xe( 8)
Pe(c0s 8) = ---
(sin 8) *
leads to the differential equation for Xe

which does not contain the first derivative of Xe with respect to 8 and is similar to a
one-dimensional Schrijdingerequation. For large Q the term (Q+ $)2 is much larger
a
than csc2 8 everywhere except for angles very close to 8 = 0 or 8 = T . This
suggests that we approximate Eq.(1.74) by the differential equation

whose solution is
xe(8) = Ae sin [(L + g)8 + a] (1.76)
where Ae and a are constants[8]. Thus for large Q and for 8 >> Q-' and ( T - 8) >>
Q-' ,we obtain

Pe( cos 8) -+ Ae
i)
sin [(L + 8 + a]
(sin 8) 4
and we may also write under the same conditions[9]

G ( e , 41)-, ( Q+ g T
' Ae
sin [(Q+$)8+ a]
(sin 8)
(1.78)

Thus

Iyeo(8, 4) l2 - (Q+ i)
---
2~ A:
sin2 [(Q+ i ) 8 + a]
sin 8
(1.79)

For large Q the factor sin2 [(Q+ $)8 + a] oscillates very rapidly and can be
i.
replaced by its average value of Then if we insist that the integral of IYeo(8, 4) l2
over dR be unity, we find that
1.4 THE VECTOR MODEL

(c) 2

Band of Area

FIGURE 1.2 Classical orbits corresponding to (a) IYm(8,4) 1' and ( b ) (&,,,(8,4) IZ for
large e. Note that e and r are perpendicular to one another. The region between 8 and 0 + d0
&fines a band shown in (c).

This result holds for large l and for all 8 except very close to 8 = 0 and 8 = n.
We may arrive at the same result by use of the vector model. For large l the particle
performs a circular orbit about t (see Figure 1.1); for m = 0, e is perpendicular to
the z axis. A typical orbit is shown in Figure 1.2a Here 8 is the angle measured
about e; 4 is the angle measured about z (in the x y plane). The probability of finding
the particle between 8 and 8 + d0 is uniform. But 4!, itself, may have any azimuthal
orientation about z , because once we specify the values of t2 and t,, we cannot
locate the position of 4! in the xy plane. Hence the chance of finding e between 4
and 4 + d4 is also uniform, and we must take this into account. The probability
of finding the particle between 8 and 0 + d9 is simply dO/n since we choose 8 to
range from 0 to n. Because 4 is not specified, the region between 8 and 8 + dB
defines a band on the unit sphere as shown in Figure 1.2~.Then the probability d 8 / n
must be spread over this band, which is bounded by two spherical segments, one
16 1 ANGULAR MOMENTUM OPERATORS AND WAVE FUNCTIONS

of height h = cos 8 and the other of height h = cos(8 + dB) on the unit sphere.
Recall that the area of a spherical segment of radius r and height h is 2 nrh. The area
of the band is the difference between the areas of the two spherical segments, that
is, 2 n [COS 8 - cos(8 + do)] = 2 n sin Ode. Thus the probability density function is
given by the probability (d8/n) per unit area ( 2 7r sin Odd), that is, by (2 n2 sin $)-I,
in agreement with Eq. (1.80). Another way to visualize this result is to consider the
density of intersections of the longitudes and latitudes on a globe (sphere). These
intersections crowd together at the poles and spread apart at the equator; that is, the
density of intersections is inversely proportional to sin 8.
Consider the situation for m # 0, as shown in Figure 1.2b. Here a is the angle
between t! and the z axis, 8 the angle between the z axis and the particle's position
vector r, #J the azimuthal angle about the z axis and 7 the angle measured about e.
Once again, the probability of finding the particle is uniform in the angle 7 while f!
may have any azimuthal orientation about z as long as a is constant. The probability
that the particle is between 7 and 7 + dy is dy/,rr since again 7 is chosen to range
from 0 to n; moreover, this probability is spread over the zone 2 ,rr sin 8 dB (see
Figure 1.2~).Hence the probability density of finding the particle at some point with
angle 8 is (dy/,rr)/2 ,rr sin 8 dB, that is

In the m = 0 case, 7 = 8 and Eq. (1.81) reduces to Eq. (1.80), as expected.

Actually, 7 is the dihedral angle[10] between the planes containing z and f2 and
containing e and r. We can relate 7 to the angles 8 and a by

Thus
sin 8d8 = sin a sin ydy (1.83)

from which it follows that

1 - 1
Ifim( B,#J)I2 = n2 sin sin
2 ,rr2 sin a ( 1 - cos2 8/ sin2 a ) )

This expression holds for sin2 a > cos2 19. The region where sin2 a < cos2 8 cor-
responds to either I9 < $ - a or I9 > + a and is classically forbidden. In Figure 1.3
we present a pictorial summary of this behavior. As compared with the semiclassical
limit, the quanta1 result for large L has rapid oscillations in the classically allowed
region and dies exponentially in the classically forbidden region[6]. Moreover, the
1.4 THE VECTOR MODEL

8 (degrees)
FIGURE 1.3 Probability density of (Gm(O,4) I2as a function of 0 for selected values of t
and m. In each plot the dashed curve is the semiclassical probability density given by
1/[2a2(sin2 a - cos2 0) 1/2],where cos2 a = m2/ [t(L + I)]. Each curve is normalized
&'" &"
so that IGm(O,4) l2 sin 0 dB &$ = 1.

quanta1 result rounds off the sharp singularitiespredicted by this semiclassical treat-
ment. Figure 1.3 suggeststhe value and deficiencies of the vector model. Apart from
rapid oscillations, the probability density IYe,(O, 4) I* behaves like that of a classi-
cal particle uniformly moving in a circular orbit for large 4. Moreover, as shown in
Figure 1.3, the vector model limit is rapidly approached as L increases in magnitude.
18 1 ANGULAR MOMENTUM OPERATORS AND WAVE FUNCTIONS

NOTES[ll] AND REFERENCES

1. Two of my favorite introductory texts about classical mechanics are H. Goldstein, Clas-
sical Mechanics (Addison-Wesley, Reading, Massachusetts, 1950) and L. D. Landau
and E. M. Lifschitz, Mechanics (Pergamon Press, Oxford, 1960).
2. See, for example, L. Pauling and E. B. Wilson, Jr., Introduction to Quantum
Mechanics (McGraw-Hill, New York, 1935); D. Bohm, Quantum Theory (Prentice
Hall, Englewood Cliffs, NJ, 1951); and L. D. Landau and E. M. Lifschitz, Quantum
Mechanics (Pergamon Press, New York, 1974).
3. If A and B are two Hermitian operators that do not commute, then the obsewables A
and B cannot be measured simultaneously. The degree to which an inevitable lack of
precision is introduced is expressed by the inequality

called the Heisenberg uncertaintyprinciple[2]. Once again the commutator has a special
importance.
4. Three standard texts on angular momentum theory are indispensable to the serious
student of this topic: M. E. Rose, Elementary Theory of Angular Momentum (Wiley,
New York, 1957); A. R. Edrnonds, Angular Momentum in Quantum Mechanics
(Princeton University Press, Princeton, NJ, 1957); and D. M. Brink and G. R. Satchler,
Angular Momentum (Clarendon Press, Oxford, 1962). The preceding represents a
blend of the above.
5. E. U. Condon and G. H. Shortley, Theory of Atomic Spectra (Cambridge University
Press, 1935; reprinted in paperback form, 1963). E. U. Condon once told me with
some disappointment that he thought that this work was most often cited in the literature
for setting this phase convention! In my first encounter with phase conventions I took
the attitude, "Why should I bother because after all, the phase is arbitrary?" Another
example of a phase conventionis driving a motor vehicle on the right- or left-hand side of
the road. As long as you and everyone else are consistent in this choice, it certainly does
not matter, but inconsistency can be detrimental to your health! This is the first of many
instances where care must be exercised in the choice of phase, a most vexing aspect of
angular momentum theory. Finally, a warning should be made to the uninitiated that
phase conventions must be checked in going from one literature reference to another
just as one checks driving habits in going from one country to another.
6. The classical correspondence of angular momentum operators and related quantities is
explored in detail by P. J. Bnissaard and H. A. Tolhoek, Physica, 23,955 (1957). The
vector model appears to have been first introduced by A. Sornmerfeld,Ann. Physik. 51,
l(1916).
7. We follow closely here L. D. Landau and E. M. Lifschitz, Quantum Mechanics
(Pergamon Press, London, 1958), pp. 166-168. Please note that cos B in Eq. (1.72)
and what follows is not the same cos 8 as in Eq.(1.71) and Figure 1.1 but refers instead
to the polar angle between the z axis and the particle vector r.
NOTES AND REFERENCES 19

This is the differential equation satisfied by the zero-order Bessel function .To with
argument (L + f ) 0, that is,

for 0 << 1. It is well known that the asymptotic expansion of the zero-order Bessel
function for x >> 1 is

Hence we have

for 6'(C+ 4) >> 1. Comparison with Eq. (1.77) permits the identification

and a = x/4. Hence for 0 >> L-' or n - 8 >> L-'

2 isin[(C++)~+n/4]
P ~ ( C O6')S = -
[.v+ +I] (sin 0) 4

9. The condition (L + 4)' >> cscZ8 implies that 2(L + ))I sin 81 >> 1, which for
small 0 may be restated as Be >> 1 or ( n - 0)L >> 1.
10. Let the three line segments AO, BO, and CO intersect at the common point 0. Denote
the included angle LAOB by 4AOB.LAOC by tPAOC, and LBOC by tPBOC. Then
according to spherical trigonometry, the dihedral angle $,qBC, defined as the angle
between the AOB and BOC planes, is related to the q5 values by

11. Notes are intended to offer further information, often of a peripheral nature, in a manner
that does not interrupt the flow of the main text; they are not meant to be skipped. See,
for example, A. Held and P. Yodzis, General Relativify and Gravitation, 13,873 (1981).
20 1 ANGULAR MOMENTUM OPERATORS AND WAVE FUNCTIONS

PROBLEM SET 1

1. Derive Eq. (1.20).

2. Prove Eq. (1.56).
3. In the I j m ) representation in which j2 and j, are diagonal, the matrix elements
of an arbitrary operator 0 diagonal in j2 are given by
Ommt = ( j m l 0 Ijm')

Thus each such operator may be represented by a ( 2j + 1) x ( 2 j + 1) matrix

where the rows are labeled by m and the columns by m'. Such matrices are
called representations.
A. For the j = 2 case, write down explicitly the 5 x 5 matrices for the operators
jz,~y,jz,j+,~-,j2

B. By carrying out the indicated matrix operations, show that

iziy- j y j z = i j z
C. Find the specific matrix element
( j = 2 , m = O l j x j y j x I j =2 , m = 1 )

4. Given that (j,) = ( j m l j, I j m ) = m, and that m ranges from - j to +j

in unit steps, we explore here an alternative procedure for identifying ( j 2 ) =
( j m l j2 I j m ) with j( j + 1). This is accomplished by equating the expectation
value of an operator 0 with its spatial average, defined by

Because the choice of coordinates is arbitrary and space is isotropic,

Perform the spatial average and show that ( j 2 ) = j( j + 1).

Sums of the form

are frequently encountered where k is an integer but j may be integral or half-

integral. We discuss here the general evaluation of Eq. (1.87). Of course we
PROBLEM SET 1 21

need not bother with Szk+ ( j ) since this sum vanishes when m ranges from -j
to +j in integral steps. We begin by examining a related problem, namely, the
sums
n

The k = 0 case is particularly simple, involving just counting:

However, we may recast Eq. (1.89)into a particularly provocative form as

where enumeration of this series, that is, [ 1 - 01 + [ 2 - 11 + . . .+ [ ( n+ 1) - n] ,

shows that the only surviving term is n + 1 . This suggests a general scheme for
the evaluation of Sk+1 ( n) in terms of the known Sk(n) for k less than k + 1 by
consideringthe sum Cy=o [ ( i + 1)k+l -ik+lI , which must equal (n+ 1)k + l . For
example, we illustrate this procedure by deriving S1( n ) with our knowledge of
So(n):

C [(i + 1)' - i2] = ( n + I)* = C [ ( i 2+ 2i + 1) - PI

i=O i=O

from which it follows that

In a similar manner it is shown that

22 1 ANGULAR MOMENTUM OPERATORS AND WAVE FUNCTIONS

For integral j we rewrite Eq. (1.87) as

and we obtain with the help of Eqs. (1.89) and (1.92H1.95) the following:

For half-integral j, Eq. (137) may be reexpressed as

and explicit evaluation of Eq. (1.100) with the help of Eqs. (1.89) and (1.92)-
(1.95) yields the same expressions as shown in Eqs. (1.97)-(1.99); that is,
Eqs. (1.97H1.99) are valid for both integral and half-integral j. An alternative
proof of the validity of Eqs. (1.97)-(1.99) for half-integral j is to recognize that
APPLICATION 1 SCAiTERlNG THEORY 23

APPLICATION 1
SCATTERING THEORY

Angular momentum plays an important role in describing and understanding collision

processes in addition to the dynamical behavior of bound systems. The following is
intended as a cursory introduction to the beautiful topic of elastic scattering as well as
a good review of many aspects of Chapter 1. First a classical treatment is presented,
then a quantum one; finally, their connection is briefly outlined.

Classical Treatment

The quantum corrections to the bulk behavior of gases, such as transport or equilib-
rium properties, are normally rather unimportant except for light gases at low tem-
peratures[l, 21. In contrast, scattering experiments are a much more delicate tool
for observing two-body interactions. As we shall see later, quantum mechanics does
play an essential role in many scattering effects. Nevertheless, classical mechanics is
capable of illuminating the outline of this subject, is more intuitive, and thus provides
a useful conceptual framework.
In the center-of-mass frame, a collision may be pictured as the interaction of a
mass point p with a central force potential V ( r ) ,chosen to be located at the origin.
The position vector r describes the location of the mass point with respect to the
origin. Then the angular momentum about the origin is given by

and the time rate of change of L is

for the force pi: is always directed along r if the potential V ( r ) is only a function
of the magnitude of r. Hence L is a vector of constant length, that is, a conserved
quantity during the course of the collision. Since r and r must both be perpendicular
to the fixed direction of L in space, the trajectory of the mass point is confined to
a plane perpendicular to L. This argument does not appear to hold if L = 0, but
then the motion must be along a straight line through the origin as L = 0 requires r
and r to be parallel. Thus central force motion may always be regarded as motion in
a plane. We arbitrarily choose the z axis along L so that the trajectory is confined to
the xy plane.
A typical trajectory is pictured in Figure 1. For elastic scattering, the initial and
final asymptotic motions differ only in the direction of the velocity v, which has been
rotated through the angle X, called the deflection angle. The asymptotic speed v and
 1 ANGULAR MOMENTUM OPERATORS AND WAVE FUNCTIONS

FIGURE 1 Elastic scattering trajectory.

impact parameter b (which is the distance of closest approach if V( r ) = 0 ) are related

to the total energy E and the magnitude of the angular momentum L by

and
L = p1r x rj = pub

Note that the trajectory is symmetric about the point of closest approach (r = r,,
8 = 8,). We choose t = 0 at this point.
Conservation of angular momentum provides three independent constants of the
motion, one for each Cartesian component of L. Two of these suffice to express the
direction of L, which is constant during the trajectory, and the third determines the
magnitude. An additional constant of the motion, the total energy E, is provided by
conservation of energy E = T + V, where T = . r is the kinetic energy and
V = V( r ) is the potential energy.

A. Show that
L = ,ur2e

and
E = +,U(f2 + r2e2) + V(r)
Hint : Introduce polar coordinates

y = r sin 8
APPLICATION 1 SCATTERING THEORY

to describe the motion in the scattering plane and evaluate

L = p(xy - yi)

By eliminating 9 in Eqs. (5) and (6) we obtain an equation for the radial motion

Equation (7)describes the one-dimensional motion of a particle of mass ,u with total

energy E in an effective potential

composed of the "true" potential V ( r ) and the centrifugal potential L2/2,ur2.In

molecular scattering, the influence of the centrifugal potential is often dominant.
The reason is that large values of L, corresponding to large impact parameters, often
contribute most to the scattering process. The effective radial force is given by

We observe that the centrifugal contribution is always repulsive, whereas that from
V ( r ) is usually attractive for large r but repulsive for small r.
The complete solution for the scattering motion can be obtained from Eqs. (5)
and (7) by integrating the differential equations

and taking into account the initial starting conditions of r and 8. Hence in the
center-of-mass frame a classical trajectory is specified by six parameters, the three
Cartesian position vectors and three Cartesian velocity vectors of the representative
mass point p, or alternatively, the direction and magnitude of L, the value of the
total energy E and the initial values of r and 8. However, we are not interested in
26 1 ANGULAR MOMENTUM OPERATORS AND WAVE FUNCTIONS

the complete timedependent solution but instead in the one observable of elastic
scattering, namely, the deflection angle X. From Eqs. (10) and (11) we find

where the collision starts at t = -w where 8 = 0 and r = w. Note that in Eq. (12)
the square root in Eq. (10) is taken with a negative sign since drld8 is negative.
With respect to the angle of closest approach, 8,,

Hence the angle of deflection is given by

where the radial distance of closest approach is determined from the condition that
drlde = 0 or equivalently .i- = 0 . From Eq. (7) we have

which may be restated as

B. Derive Eq. (16) from Eq. (15).

Once the potential function V(r) is specified, the deflection angle x may be
calculated by using Eq. (14). Moreover, if V(r,) < 0 (attractive potential at the
turning point of the radial motion), then rc < b, whereas if V (rc) > 0 (repulsive
potential at the turning point), r, > b.
Consider a beam of particles incident on some scattering center. Collisions occur
with all possible impact parameters (angular momenta), giving rise to a corresponding
APPLICATION 1 SCATTERING THEORY 27

distribution in angles of scattering, which is described by a diflerential cross section.

Let the intensity of the incident beam be characterized in terms of the flux 10,
where lois the number of particles crossing a unit area normal to the beam direction
per unit time. The differential cross section I(.X) is defined so that I ( X)cUZ is the
number of particles per unit time scattered into a solid angle element dC2 divided by
the incident flux. Note that I(x) has the dimensions of area per steradian.
Because of the spherical symmetry of the force field, the deflection pattern is
axially symmetric about the incident beam direction. Hence I(x) depends only
on X. Accordingly, the solid angle element dC2 may be taken between the cones
defined by x and x + dx, that is, dl2 = 2 n sin xdx. Even if the force field were
nonspherical, the averages over all possible impact parameters and all orientations of
the target would cause any azimuthal dependence to vanish unless one of the reactants
were initially polarized.
For a given initial velocity v , the fraction of the incident flux with impact parameter
between b and b + db is 2 nb db; these particles undergo deflections between x and
x + dx if dxldb > 0 or between x and x - dx if dxldb < 0. This gives the relation
I(x) = 2 nb db, from which it follows that

b
I('' = sin x ldxldbl

where x is related to b through Eq. (14). Hence a knowledge of ~ ( b for) a given v

serves to determine the differential cross section, I(x). This is related to the total
cross section a by

a = 2 11 loI( X) sin xdx (18)

where we disregard the possibility of divergence. Here a is a measure of the

probability for removal by scattering from the incident beam (i.e., attenuation) and a
has the dimensions of area.

C. Consider scattering from an impenetrable sphere of radius a so that

V(r) = oo, r <a

Show that the classical differential cross section is independent of the deflection
angle and that the classical total cross section is just the area of a circle (bull's
eye) of radius a. Hint: Begin by proving that x = 2 arccos (bla) for b 2 a,
x = 0 forb > a.

The deflection angle often resembles the behavior shown in Figure 2 as a function
of impact parameter. For large impact parameters x is small and negative (net
 1 ANGULAR MOMENTUM OPERATORS AND WAVE FUNCTIONS

FIGURE 2 Deflection angle as a function of the

impact parameter (generic form).

attraction), and as b decreases, the deflection becomes more negative (increasing

attraction). The deflection then either approaches negative infinity, corresponding
to classical orbiting, or reaches a finite minimum. Classical orbiting occurs when
the impact parameter equals the maximum in the centrifugal barrier and the initial
velocity is such that the (radial) velocity vanishes at the maximum in the centrifugal
barrier. For still smaller impact parameters, the effect of the short range repulsive part
of the potential causes the deflection angle to increase, passing through zero (where
the net attraction and repulsion cancels) and finally to approach 180 as b --+ 0. As
Eq. (17)shows, the differential cross section becomes singular when either x = 0•‹,
called forward glory scattering, or d x l d b = 0 at the minimum, called rainbow
scattering. The rainbow deflection angle X , will depend, of course, on the velocity v .
The calculation of the differential cross section is considerably simplified if we
consider only those collisions at large impact parameters since they contribute only
to scattering at small deflection angles, that is, in the forward direction. Let the initial
momentum of the particles be along the x axis. Denote by p' the momentum of the
particle after scattering. Then, p' sin x = p:, and for small deflection angles, we take
sin x x and write
' momentum transfer
X"- PY =
P' momentum

Since the time derivative of the momentum is the force, the momentum transferred
perpendicular to the initial direction is given by integrating the force component Fi
over the trajectory. But

where we use the facts that r 2 = s2 + Y2 and y N b. Hence

APPLICATION 1 SCATTERING THEORY

where x = ( 2 E/p)'12t and x varies from -co to cx, as r varies from cx, to b and
back.

D. Use Eq. (21) to examine the small-angle elastic scattering from the potential
V(r) = Csr-" for s > 0. Show that

from which it is concluded that XEis a function of b alone.

Complete this study by evaluating I(x), using Eq. (17) to prove that

for small values of X. Hence a log-log plot of the center-of-mass differential

scattering cross section I(x) as a function of the scattering angle x at fixed en-
ergy should give a straight line with slope -(2 + 21s) from which the value of
the s can be determinedf2-51. For the usual case of neutral closed-shell systems,
s = 6 (van der Waals long-range attractive potential) and I(x) is proportional
to E-1 l3X-7 13.

Quantum Treatment

The quantum formulation of scattering in a central force field may be found in many
texts[5-111 and is not developed in detail here. Briefly, the scattering is determined
by the asymptotic form of the wave function
30 1 ANGULAR MOMENTUM OPERATORS AND WAVE FUNCTIONS

where A is a normalization constant and

is the magnitude of the initial propagation vector k that is directed along x = 0.

In Eq. (24) the term A exp( ikz) represents a plane wave incident on the scattering
center and the tern A f (x) exp( ikr) /r represents an outgoing spherical wave. Note
that the amplitude of the scattered wave is inversely proportional to r since the radial
flux must decrease as the inverse square of the distance. If the incoming and scattered
fluxes of particles are separated from each other, by collimating slits, for example,
the differential cross section is related to the scattering amplitude f (x) by

Thus the asymptotic form of the wave function determines the differential scatter-
ing cross section but cannot be found without solving the wave equation throughout
all space. This may be carried out by the method of partial waves in which the general
solution is represented as an infinite sum of Legendre polynomials

The boundary condition that Re(r) remains finite at r = 0 determines the asymptotic
form of the solution, up to the normalization constant.
In the absence of a potential V (r ) = 0, the wave function is simply A exp( ikz) ,
which can be cast into the form of Eq. (27) for large r as
w
eikz- r-+w ~ ( 2 + e1) exp
e=o
(F) sin ( kr - h / 2 )
kr Pe( X)

Equation (28) has an appealing physical interpretation, namely, the incident plane
wave is equivalent to a superpositionof an infinite number of incoming and outgoing
spherical waves in which each term in Eq. (28) corresponds to an orbital angular
momentum of magnitude

about the scattering center. Classically, this angular momentum would correspond to
an impact parameter
APPLICATION 1 SCATTERING THEORY 31

Thus we can picture the incident beam as divided into cylindrical zones such that the
4th zone contains particles with impact parameters between 4X and (4 + 1) X.

E. In quantum mechanics only integral values of 4 are allowed, yet we view b

as being continuous. How will this affect our results when treating chemical
systems?

Provided V(r) falls off more rapidly than 1/ r for large r , the general solution
[Eq. (27)] can be shown to have the asymptotic form

where is called the phase shift and must be a real number. Comparing Eq. (31)
with the solution for V(r) = 0 , namely, Eq. (28), we conclude that the effect of
the scattering potential is to introduce a change of phase in the asymptotic form of
the radial wave functions. Moreover, the phase shift occurs only in the outgoing
part of the asymptotic wave function. This is indicated schematically in Figure 3.
A repulsive potential causes a decrease in the relative velocity of the particles for
small r so that the wavelength is increased. Thus the scattered wave is "pushed
out" relative to that for V = 0 , and consequently the phase shift qe is negative. An
attractive potential "pulls in" the radial wave function and produces a positive phase
shift. Comparing 4 = 0 to l # 0 , we also see that the repulsive centrifugal potential
contributes a negative phase shift of -ln/2.
Further analysis shows that the scattering amplitude is given by

which may be derived by comparing Eqs. (24), (28), and (31). Then the differential
cross section has the fonn

Equation (33) shows that interference between the terms with different values of l
plays an important role in determining the differential cross section. In terms of
 1 ANGULAR MOMENTUM OPERATORS AND WAVE FUNCTIONS

(a) V>O, Repulsive Potential

(b) V< 0 , Attractive Potential

FIGURE 3 Form of the radial wave function (solid curve) versus separation distance for (a) a
repulsive potential (V(r) > 0 ) and (b) an attractive potential (V( r) < 0). For comparison,
the radial wave function (dashed curve) for no potential (V(r) = 0 ) is also shown.

our previous picture, the outgoing partial waves scattered from the various zones
of incident impact parameters are superimposed with weighting factors that depend
on the phase shift associated with each zone. Whenever the phase shift is zero or
an integral multiple of n, the corresponding partial wave does not contribute to the
scattering, since it has been "pushed out" or "pulled in" an integral number of half
wavelengths so that the nodes and extrema in its asymptotic form match those of the
incident partial wave, that is, the V ( r )= 0 solution.
The total cross section is obtained by integrating Eq. (33) over all directions:

Cross terms in the expansion of Eq. (33), which represent the interference of different
partial waves, cancel on integration because of the orthogonality of the Legendre
polynomials. Equation (34) shows that each partial wave contributes separately
to a in proportion to sin2 qg and its statistical weight (2L + 1). The coefficient 47r
represents the integral over the solid angle of scattering while h2 = (XI2 n) = 1/ k2
is associated with the squared wavelength of the incident particles. From another
viewpoint, however, the physical interpretation of Eq. (34) is somewhat strange. The
flux of particles incident with impact parameters within the tth zone is proportional to
APPLICATION 1 SCATTERING THEORY 33

the cross sectional areaof the zone, namely, 7rh2[ ( l + 1)2-12] = 7rX2 ( 2 l + 1). This
disagrees with Eq. (33) by a factor of 4. Actually, this factor is the result of quantum
mechanical diffraction effects that are not properly included in our simple physical
picture. Nevertheless, we did find the correct form in which a sin2 weighting factor
appears to express the effectivenessof the potential in scattering the lth partial wave.
The uncertainty principle provides a clarification concerning the convergence of
Eq. (18) for the total cross section. In a classical treatment, a diverges unless the
potential vanishes outside some finite radius. If the potential extends to infinity, then
even very large values of the impact parameter b produce a small deflection and thus
contribute to the scattering cross section. However, if the potential is sufficiently
weak at large r, the collisions with large impact parameters produce such slight
deflections that the scattering angle x is smaller than the directional uncertainty
required by the uncertainty principle. For such collisions, the particle cannot be
regarded as scattered, and thus the uncertainty principle, in effect, causes a cutoff in
the impact parameter b that can contribute to the scattering process, making the total
cross section finite. This is yet another example where quantum mechanics rounds
off a classical singularity.
Let us make these considerations more quantitative. There is a largest value of b,
denoted by b,, for which x is just large enough to be observable. Let Ap; be the
uncertainty in the transverse component of the momentum. Then according to the
uncertainty principle, b,Ap; = h or Ap; = hlb,. Thus it follows from Eq. (19) that
the deflection angle X, associated with b, must be

However, we also found accordingto Eq. (22) that for a potential of the form V (r ) =
C,r-', xc is related to the impact parameter b, by

Combining Eqs. (35) and (36), we have

where v is the initial velocity. It follows that the cross section cannot exceed xbf and
that its velocity dependence is v-~/(~-').For s = 6 , a(v) is proportional to v-~/'.
Actually, an undulatory velocity dependence of a ( v) is often observed superim-
posed on the general trend predicted above[2-51. This is brought about by the quan-
tum interference arising from trajectories at very large impact parameters (b 2 b,)
and glory trajectories b 21 b, since both contribute to x 5 x,. Nevertheless, a plot
of In a(v) versus In v can be used to reveal the power law behavior of the long- and
short-range parts of the potential and even to obtain an estimate of C,.
 1 ANGULAR MOMENTUM OPERATORS AND WAVE FUNCTIONS

FIGURE 4 Reprinted with permission from D. Beck and H. J. Loesch, Z. Phys., 195, 444
(1966).

To illustrate this procedure, we present in Figure 4 the velocity dependence of the

total cross section when a velocity-selected beam of K crosses a thermal beam of Cs,
Hg, Xe, Kr, or Ar [see D. Beck and H. J. Loesch, Z. Phys., 195,444(1966)l. The data,
which have been corrected for the velocity distribution of the target gas, show a slope
close to -2 1 5 , the value expected for an s = 6 long-range attractivepotential. Glory
scattering undulations in the total cross section are also apparent. The amplitudes of
the glory undulations contain information on the width of the potential well. Total and
differential cross section measurements have been widely used to obtain information
about intramolecularpotential energy functions[2-51.

Transition from Quantum to Classical Scattering

The phase shift, which so naturally arises in a quantum description of the scatter-
ing process, may also be defined semiclassically by comparing the number of wave-
APPLICATION 1 SCATTERING THEORY 35

lengths contained in two different paths, the actual trajectory, and the path that would
be followed if the scattering potential were "switched off":

Here R is the radius of a sphere whose center coincides with the scattering center,
and

is the "local" de Broglie wavelength associated with the radial motion. Thus the
semiclassical phase shift is given by

This formula can be put in a more familiar form by introducing k, = pi-/h and
rewriting Eq. (38) as

The last integral is readily evaluated:

=kR- -
kbn
2
Hence

$' = lim [ l r ( k, - k ) d r + k( R - 7,) - k ( R - b7r/2)

R-tw !

which is the standard form for the semiclassical phase shift as given in many quantum
mechanics texts[5-111.
36 1 ANGULAR MOMENTUM OPERATORS AND WAVE FUNCTIONS

The deflection angle x may also be expressed in a form similar to that for GCby
writing

Equation (44) may seem a peculiar way to express X , and it might be wondered why
bother. After all, the last integral in Eq. (44) is

so that Eq. (44) reduces to Eq. (14). However, Eq. (44)allows us to deduce a simple
and useful connection between the classical deflection angle x and the rate of change
of the semiclassical phase shift with angular momentum, dr$c/de. Recall that the
general rule for differentiating a definite integral is

(46)
By differenting Eq. (40) with respect to l and using Eq. (46) and the relation
(db/de)E= 1l k , we find that

Next let us examine the differential cross section I ( x ) . The mathematical identity

is readily verified. If we exclude the singular point x = 0, which contributes

negligibly to the total flux, we can rewrite Eq. (32) as
APPLICATION 1 SCATTERING THEORY 37

To obtain the classical limit, we consider the situation where the potential varies
slowly over distances comparable to the de Broglie wavelength and that many e
values contribute to the scattering at a given angle so that the major contribution
arises from large e values. Then for x # 0 ,

where

F. Derive Eq. (50).Hint: Reread Note [8] of Chapter 1.

In Eq. (50)the exponential factors are rapidly oscillating functions of 4, since the
phases +* are large. Thus the majority of terms cancel and this sum is determined
mainly from the range o f t values for which either 4' or 4- is an extremum. From
Eq. (51) this implies that Eq. (50) is essentially determined only by phase shifts
satisfying the relation

where the plus sign is for &+/de = 0 and the minus sign for &-/de = 0 . Compari-
son of Eq. (52)with Eq. (47)demonstratesthat only these phase shifts q correspond-
ing to the impact parameter b contribute to the differential cross section I ( x ) in the
classical limit. Conversely, the condition for classical scattering at a given deflection
angle x is that the values of L be large for which Eq. (52)applies.
We conclude by examining one of the most striking phenomena in scattering,
namely, resonance scattering,in which the total cross section shows an abrupt change
as a function of energy. As shown in Eq. (34),each partial wave e contributes to the
total cross section a term

Generally the phase shifts qe are slowly varying functions of the energy. A resonance
occurs for some partial wave e when qe changes rapidly over a small energy range,
and we write

where ngis the background phase shift and q, is the resonance phase shift.
What makes a resonance occur? To understand this we turn our attention to
Eq. (31),the asymptotic form of the wave function. For a given l the ratio Se( k )
38 1 ANGULAR MOMENTUM OPERATORS AND WAVE FUNCTIONS

of the outgoing wave amplitude to the incoming wave amplitude at energy E =

f i 2 k 2 / 2 is
~
St( k) = e2*(k) (55)

up to a constant factor. Here we have emphasized the dependence of the phase shift
on energy by writing qe = qe( k) . The wave number k is a real variable. However, by
the process of analytic continuation,Q( k) can be regarded as a function of a complex
variable k. Suppose there exists an imaginary value of k, that is

where )E is real and positive definite such that

Of course, this implies that the energy is negative, and

that is, the asymptotic behavior of the wave function will go as exp( - ~ r +h / 2 ) / K T ,
and there is no outgoing wave ampitude, only an exponentially dying form. This is
exactly the condition of a bound state. Thus the zeros of Secorrespond to bound
states. Conversely, when k = - i ~ ,St will grow without bound, and it follows that the
poles (singularities in the complex plane) of St correspond to scattering resonances.
Close to a resonance

The simplest way of causing Seto grow without bound near Eo is to make it have a
simple pole, that is, to set

where the meaning of r 12 will become apparent but for the present may be assumed
to be a constant independent of energy. Then Eq. (59) becomes
APPLICATION 1 SCATFERING THEORY

FIGURE 5 Line profile of an isolated scatteringresonance for two different background phase
shifts, ng= 0 and ng= ?r/10. Note that when ng# 0 it is possible for the cross section to
vanish at a particular energy (the so-called Cooper minimum).

and Se( k ) has a pole at E = Eo - i ( T / 2 ) .

When mgcan be neglected, substitution of Eq. (60) into Eq. (53) yields the result

where we have made use of the identity

Equation (62) is called the Breit-Wigner formula[l2] for an isolated scattering

resonance. It describes a bell-shaped curve with a half-width at half-maximum
of r / 2 . This suggests that the occurrence of a resonance at Eo in the partial
40 1 ANGULAR MOMENTUM OPERATORS AND WAVE FUNCTIONS

cross section is associated with the formation of a virtual state (metastable bound
state) with energy Eo and width T . Moreover, in analogy to a bound state, the
time dependence of this virtual state (scattering resonance) is given by exp[-i x
(Eo - iT/2)t/A] so that it is characterized by a lifetime r = A/T.
In the more general case, qb, cannot be ignored and there is interference between
the background and resonance phase shifts. Figure 5 shows plots of sin2 qe(E ) as a
function of E for qg= 0 .O and qbg = T / 10, assuming that r is independent of E
over the resonance.

G. Show that the partial cross section is then given by

where

is the energy difference measured in half-widths and

q = cot qg

is called the line profile index.

It has been shown by Fano[l3] and Fano and Cooper[l4] that absorption lines in
photoionization may have profiles as illustrated above. Again this arises from inter-
ference between the continuum states produced by nonresonant photon absorption
and the continuum state produced by the decay of a resonance.
One of the commonest situations that cause the appearance of resonances is an
effective potential made up of an attractive part at small distances and a repulsive
centrifugal barrier at long distances; as shown in Figure 6. For energies below the
maximum in the centrifugal barrier, there would be bound states inside the attractive
part of the potential if tunneling could be ignored. However, the presence of quantum
mechanical tunneling permits particles "trapped" inside the attractive part of the
potential to escape to infinity, and the tunneling rate depends on the height and
thickness of the barrier. Conversely, particles incident on the potential at energy
close to the virtual state energy are able to penetrate inside the attractive barrier. This
behavior explains why resonances generally become narrower as l increases. Larger l
values cause bigger centrifugal barriers, thus suppressing tunneling.
NOTESANDREFERENCES

FIGURE 6 Plot of the potential energy including the centrifugal potential versus separation
distance.

NOTES AND REFERENCES

1. J. 0.Hirschfelder, R. F. Curtiss, and R. B. Bird, Molecular Theory of Gases andLiquids,

Wiley, New York, 1954.
2. G. C. Maitland, M. Rigby, E. B. Smith, and W. A. Wakeharn, Intramolecular Forces,
Clarendon Press, Oxford, 1981.
3. R. B. Bernstein, Chemical Dynamics via Molecular Beam and Laser Techniques,
Clarendon Press, Oxford, 1982,Chapter 5;Adv. Chem. Phys., 10,75(1966).
4. J. P. Toennies, "Molecular Beam Scattering Experiments on Elastic, Inelastic, and
Reactive Collision," in Physical Chemistry, Vol. VIA,H. Eyring, D. Henderson, W. Jost,
eds., Academic Press, New York, 1974,Chapter 5.
5. M. S.Child, Molecular Collision Theory, Academic Press, New York, 1974.
6. H. S. W. Mott and N. F. Massey, The Theory of Atomic Collisions, 3rd ed., Clarendon
Press, Oxford, 1965.
7. R. G.Newton, Scattering Theory of Waves and Particles, McGraw-Hill, New York,
1966.
8. L. I. Schiff, Quantum Mechanics, 3rd ed., McGraw-Hill, New York, 1968.
9. L. D.Landau and E. M. Lifshitz, Quantum Mechanics, Addison-Wesley, Reading, MA,
1958.
10. J. R. Taylor, Scattering Theory, Wiley, New York, 1972.
11. U. Fano and A. R. P. Rau, Atomic Collisions and Spectra, Academic Press, Orlando,
1986.
12. G.Breit and E. Wigner, Phys. Rev., 49,519(1936).
13. U.Fano, Phys. Rev., 124, 1866 (1961).
14. U. Fano and J. W. Cooper, Phys. Rev., 137, A1364 (1965);Rev. Mod. Phys., 40,441
(1968).
 Chapter 2
COUPLING OF TWO
ANGULAR MOMENTUM
VECTORS

2.1 CLEBSCH-GORDAN COEFFICIENTS

In classical mechanics, the total angular momentum j of a system composed of two

parts, one with angular momenta jl and the other with j2, is given by the vector
addition of j, and j2 :

Actually, it is a simple matter to show that the sum of two angular momenta is
also an angular momentum in the quantum mechanical sense in that it satisfies the
commutation rules of Eq. (1.8 ). For example

where the middle two commutators vanish because angular momenta in different
spaces commute.
There are two useful points of view in describing such a compound system. One
complete set of commuting angular momentum operators is j: , jlz,ji , and j2,.
The states ml ,jz m2)r Ijl ml)V 2 m2)are simultaneous eigenfunctionsof these
operators:
 2 COUPLING OF TWO ANGULAR MOMENTUM VECTORS

Here the I j l ml ,j 2 m2) states are called the uncoupled representation and span a
space of dimension ( 2 jl + 1)(2j2 + 1).
It is easily verified that another complete set of commuting angular momentum
operators is j:, j:, j2 = ( j l + j2)2, and j, = jl, + jzZ.The states Ijm),
which are simultaneously eigenfunctions of these operators, are called the coupled
representation and span a space of dimension 2 j + 1 for each j value. It is also
common to write 1j l j 2 jm) or I( j l j2) jm) for I jm), but we prefer to use in general
the more compact notation. The relations corresponding to Eq. (2.3) are

Note that each set of commuting angular momentum operators contains the same
number of observables. Hence these two descriptions are equivalent and the two
representations are connected by a unitary transformation[l]

or by the inverse transformation

where the elements of the unitary transformation, called Clebsch4ordun coeffi-

cients, are chosen to be real. We make the formal identification

The goal of the remainder of this section as well as the next two sections is to
elucidate the nature of the Clebsch-Gordan coefficients, which are also referred to in
the literature as vector coupling coefficients,vector addition coeflcients, and Wigner
coefficients.
2.1 CLEBSCH-GORDANCOEFFICIENTS

TABLE 2.1 Matrix of Clebsch-Gordan Coefficientsa

(3,lI (3,Ol (2,11 .., ( - 2 - (-3,OI (-3,-11

14 4 ) (3,444) 0 o
14 3 ) 0 (3,014 3 ) (2,114 3 )
13 3 ) 0 (3,013 3 ) (2,113 3 )

"The ( j l ml ,j2 m2 1j m ) symbol is abbreviated here as (ml ,mz Vm). The case jl = 3, j2 =
1 is illustrated. In general, C is a ( 2j l + 1)( 2j2 + 1) x ( 2jl + 19( 2 j2 + 1) unitary (orthogonal)
matrix of (real) elements. Equations (2.8) and (2.9) express the relation CCT = I where
CT = C-' . Most of the elements vanish because the triangle condition m = ml + m2 is not
satisfied. By grouping elements with the same m value, C is put into block diagonal form, as
above.

The orthonormalityof Vm)and Ij l m l j2 , rnz ) leads to the orthogonality relations

and

where we have chosen to use the bracket notation [see Eq. (2.7)] for the Clebsch-
Gordan coefficients[2]. These orthogonality relations express the unitary nature of
the matrix of Clebsch-Gordan coefficients(see Table 2. I), namely, the scalar product
of any two column vectors or row vectors of this matrix vanishes and the scalar
product of any vector with itself is unity.
The Clebsch-Gordan coefficientsvanish unless the so-called triangle condition is
satisfied, namely

and

that is, when coupling the angular momentum states V l m l ) and V2mz ) the magnetic
quantum numbers m l and m 2 add algebraically while the angular momenta j1 and j2
add vectorially.
46 2 COUPLING OF TWO ANGULAR MOMENTUM VECTORS

The derivation of Eq. (2.10) is straightforward. We apply j, = jl, + j2, to

Eq. (2.5):

By equating the coefficients of like terms of I jl ml ,j2 m2) we obtain the condition

It follows that either m = ml + m2 or ( j l ml ,j2 m2 1j m ) vanishes. As a consequence

of Eq. (2.10), the sum over ml and m2 in Eq. (2.5) is constrained by the condition
ml + m2 = m so that one of the two indices is redundant. Similarly, the sum over m
in Eqs. (2.6) and (2.9) is also superfluous.
The derivation of Eq. (2.11) is more involved and often ignored in elementary
texts. However, the ideas presented in this proof are useful in counting the states
of atoms and diatomic molecules. We know that for a given j , - j < <
m j and
m, = j. Since m = ml + m2, the maximum value of m for all j is m, = jl + j 2 ,
.
that is, the maximum value of ml + m z This must also be the maximum value of j ,
that is,

for otherwise there would be values of m larger than m., Consider how the cou-
pled representation I j m ) is related to the uncoupled representation I jl ml ,jz m2)
(see Table 2.2). For the coupled states with m = m-, there is only one such
State I,j = jl + jz , mmx = jl + j2 ), and it gives rise to only one uncoupled state
ljl jl ,j2 j2). The next smaller m value is jl + j2 - 1, and it occurs only in
the coupled states I jl + j z , jl + jz - 1 ) and V 1 + j2 - 1 ,jl + jz - 1). Combina-
tions of these two coupled states give rise to the uncoupled states ljl jl - 1 , j2 j 2 )
and ljl jl ,j2 j2 - 1). Similarly, linear combinations of the three coupled states
withm = jl + j2 -2,namely, I j l + j : ! , j l + j z - 2 ) . Ijl + jz - l , j l + j2 - 2 ) ,
and + j2 - 2 , jl + j2 - 2 ) , give the three uncoupled states Ijl j l , j2 j2 - 2 ) ,
l j l j l - 1 , j z j z - l ) , a n d l j l j l - 2 , j2j2). Thu~eachtimemisreducedbyoneunit,
one of the coupled states that occurs belongs to a j value of one less unit, and the m
value of this state is at its maximum value. The lower limit jmi, is reached when
the number of coupled states (of which there are 2 j + 1 for each j value) has been
matched against all the uncoupled states [of which there are ( 2jl + 1 ) ( 2jz + I ) ] ,
that is,
2.1 CLEBSCH-GORDAN COEFFICIENTS 47

TABLE 2.2 Relation between Coupled and Uncoupled Representations (jl 2 j 2 )

Coupled Representation Uncoupled Representation

m = m, -2

m = m, -4

To evaluate Eq. (2.15) we use the following identities concerning the sums of
integers or half-integers: For integral j

The same result is readily shown to hold also for half-integral j:

 2 COUPLING OF TWO ANGULAR MOMENTUM VECTORS

We also need the sum

which also is valid for either integral or half-integral j. Equation (2.15) becomes

Substitution of jl + j2 for jmm yields

Since jmin 2 0 , it follows that

Equation (2.21), together with Eq. (2.14), establishes Eq. (2.11) for integral or half-
integral j [3].

2.2 CLEBSCH-GORDAN COEFFICIENTS AND 3-j SYMBOLS:

SYMMETRY PROPERTIES AND EXPLICIT VALUES

Recursion relations between the Clebsch-Gordan coefficients may be derived by

,
applying the operator identity j* = jl + j2* to both sides of Eq. (2.5). With the use
of Eqs. (1.17) and (1.28) the left-hand side of Eq. (2.5) becomes
2.2 C-G COEFFICIENTS AND 3 - j SYMBOLS: SYMMETRIES AND VALUES 49

By equating like coefficientsof Ij l ml ,j 2 m2) in the preceding two expansions [in

Eqs. (2.22) and (2.23)], we obtain the relation

The left-hand side of Eq. (2.24) vanishes for m = j if we take the upper sign; this
permits us to determine the various (jl ml ,j2m2 Ijj) subject to the orthonormality
conditions given in Eqs. (2.8) and (2.9) and subjectto an overall phase. The latter may
be fixed by the convention that (jljl ,jzj - jl Vj) is always real and positive[4]. The
lower sign in Eq. (2.24) then relates (jl ml ,j2m2 Ijj - 1) to the (jl mi, j2mi Ij j )
so that one can work out all the Clebsch-Gordan coefficientsby starting with m = j.
Equation (2.24), which is based on the phase convention Eq. (1.28), implies that all
the Clebsch-Gordan coefficients are real.
In this manner explicit expressions for the Clebsch-Gordan coefficients may be
derived. In particular, after much algebra Racah[4] showed that

x x ( - ~ ) ~ / [ v ! ( j +l j 2 - j 3 - v)!(jl - ml - v)!(j2 + mz - v)!

Y

where s = jl + j p + j3 and the index v ranges over all integral values for which the
factorial arguments are nonnegative[5].
From Eq. (2.25) the following symmetry relations are readily verified:
 2 COUPLING OF TWO ANGULAR MOMENTUM VECTORS

The relations given in Eq. (2.26) take on their most symmetric form if we rewrite the
Clebsch-Gordan coefficients as Wiper 3 - j symbols, defined by

Note particularly the minus sign appearing in the Clebsch-Gordan coefficient

in Eq. (2.27) and the minus sign appearing in the 3 - j symbol in Eq. (2.28). The
3 - j symbols have the property that an even permutation of the columns leaves the
numerical value unchanged

while an odd permutation is equivalent to multiplication by (- l)j1+j2*j3

2.3 C-G COEFFICIENTS AND 3 - j SYMBOLS: GEOMETRIC INTERPRETATION 51

as is the replacement of the bottom row by the negative of all its arguments:

The orthogonality properties of the 3 -j symbols are not as convenient as that of

the Clebsch-Gordan coefficients:

and

Nevertheless, the 3 -j symbols are extensively used.

The monograph, the 3-j and 6 - j symbols, by Rotenberg et al.[6], permits the
numerical evaluation of all 3 -j coefficients in steps of up to any argument equal
to 8. Attention is also called to the existence of efficient algorithms for the exact[7]
and semiclassical[8] evaluation of 3-j symbols. These are particularly useful for
large quantum number arguments. In Tables 2.4 and 2.5 we list for reference purposes
algebraic expressionsfor the Clebsch-Gordan coefficients and 3 -j symbols when one
i,
of the angular momenta is 0, 1, or 2. We will have reason to make frequent
use of these tables[9] since they give alegebraic expressions for almost all Clebsch-
Gordan coefficients commonly encountered. However, often the numerical value is
what matters in many practical problems, and for this purpose we list in the Appendix
a simple (unoptimized)FORTRAN program for evaluating 3 -j symbols.

2.3 CLEBSCH-GORDAN COEFFICIENTS AND 3-j SYMBOLS:

GEOMETRIC INTERPRETATION

We look to the vector model as a means of providing an interpretation of the Clebsch-

Gordan coefficients and 3-j symbols. According to this picture the coupled state
Vm) is represented by two vectors j 1 and j2 precessing in phase about their resul-
tant j , which, in turn, precesses about the z axis (Figure 2.1). The precession of j ,
and j , about j introduces an indeterminancy in their z components, ml and m2,
although at every instant m = ml + m2 is fixed. The Clebsch-Gordan coeffi-
cient ( j lml ,j2m2Ijm) thus represents the probability amplitude that the coupled
state V m ) will be found having its component parts jl and j2 making the projec-
tion ml and m2 for jlz and j2 t , respectively. The corresponding 3 -j symbol is the
probability amplitude divided by the factor ( 2 j + 1) ' I 2 .
In terms of the vector model, the uncoupled state )jl ml ,j2 m z ) is represented
by two vectors j l and j2 precessing independently about the z axis and making the
 2 COUPLING OF TWO ANGULAR MOMENTUM VECTORS

FIGURE 2.1 Vector model representation of the coupled state Ijm).

projections ml and m;!,respectively (Figure 2.2). The Clebsch-Gordan coefficient

( j l ml ,j2 mz 1j m ) can thus be interpreted as giving the probability amplitude that
the uncoupled state Vl ml ,jz m2 at any instant couples together to form a resultant
'I
state 1j m ) of length [j ( j + l)] 2 . The squares of the Clebsch-Gordan coefficients
represent probabilities and as such are proper fractions whose values range from 0
to 1.

FIGURE 2.2 Vector model representation of the uncoupled state 1jl ml ,j2
2.3 C-G COEFFICIENTS AND 3-1 SYMBOLS: GEOMETRIC INTERPRETATION 53

FIGURE 2.3 Alternative vector model representation of the uncoupled state J j lml ,j 2 m2).

It is instructive to use the vector model to approximate the value of

( j lml ,j2 m2)j m ) 2 . Let us return to Figure 2.2. Because of the cylindrical sym-
metry of the diagram as a whole, we may fix the orientation of j , arbitrarily and
considerj2 to precess uniformly about the z axis, as shown in Figure 2.3. We denote
the polar angles of j , and j2 by ( 81 , and ( 82 , +2), respectively. Let us denote
by 7 the included angle between jl and j2. Then

where - 4 2 is the dihedral angle between the j l z and j2 z planes. It is precisely

this angle that changes uniformly as j i precesses about the z axis.
Let us differentiate Eq. (2.34) with respect to time to obtain the rate of change of
the length of j. From the law of cosines

so that
 2 COUPLING OF TWO ANGULAR MOMENTUM VECTORS

However, as Figure 2.3 shows, the projections of the jl, j2,j vectors in the xy plane
form a triangle whose sides are s l = U1 I sin 91, s2 = Ij2I sin 82, and s = sin 8 and
whose angle opposite UI sin 8 is the dihedral angle $I - $2. The area of this triangle
is given by

Comparison with Eq. (2.37) shows that

We normalize the rate by setting d( $ 1 - $2) l d t = 2 T , that is, one revolution per
unit time. Then the probability density P ( j ) that the resultant of j, and j2is a vector
of length is given by the dwell time, that is, by 2(dUl/dt)-' where the factor of
two arises because U I occurs twice as the dihedral angle $1 - $2 runs from 0 to 2 T :

Hence, according to thevectormodel[lO], ( j l ml ,j 2 m2 Ijm)2 is proportional to UI/s

and inversely proportional to twice the area of the (j, ,j2,j) triangle projected onto
the xy plane. This may be expressed in terms of the square of the corresponding 3 -j
symbol as

where has been approximated by j + for large j values. Equation (2.41) gives a
geometric interpretation of the square of the 3- j symbol in the limit of large angular
momenta.
2.3 C-G COEFFICIENTS AND 3 - j SYMBOLS: GEOMETRIC INTERPRETATION 55

Let us make Eq. (2.41) quantitative by referring again to Figure 2.3. From the law
of cosines

=[I-( UI2 sin2 82-DlIj,I sin

1' sin2 8, - Ij2l2
81 1j2I sin 82
sin2 82 )] (2.42)

Substitution of Eq. (2.42) into Eq. (2.38) yields for the area the expression

where

so that Eq. (2.41) becomes

Note that when ( a + b + c)' - 2(a2 + b2 + c2) is negative, the 3 - j symbol (and
corresponding Clebsch-Gordan coefficient) is classically forbidden and the classical
probability is zero.
Table 2.3 is a numerical test of the validity of Eq. (2.45). The behavior is typical.
Equation (2.45) does not approximate well the squares of individual 3 - j symbols,
but it does provide an approximation for the average taken over a few neighboring
values[ll]. The reason for this is that the 3 - j symbols and corresponding Clebsch-
Gordan coefficients exhibit the usual quantum behavior, namely, rapid oscillations
in the region classically allowed by the vector model and exponential decrease
in the classically forbidden region. Nevertheless, the vector model does provide
a good insight into the nature of 3 - j symbols and Clebsch-Gordan coefficients.
Regarded in this manner, these hostile-appearing symbols bristling with six different
arguments may even turn out to be friendly companions in this journey through
angular momentum land.
56 2 COUPLING OF TWO ANGULAR MOMENTUM VECTORS

TABLE 2.3

Square of Exact Vector Model

3 - j Symbol Value Estimate
2.3 C-G COEFFICIENTS AND 3 - j SYMBOLS: GEOMETRIC INTERPRETATION 57

TABLE 2.4 Algebraic Expressions for Some Commonly Occuning Clebsch-Gordan

Coefficients ( j lml ,jz m2 Vm)'
 2 COUPLING OF -TWO ANGULAR MOMENTUM VECTORS

TABLE 2.4 (continued)

(jlm+f,f-iljl+im)=

(jl m + 8 , ; -+1j1 1
- T m)
(j1+3rn+f) jl-m++
2 j 1 ( 2jl + 1 ) ( 2jl + 3) I'
2.3 C-G COEFFICIENTS AND 3-3' SYMBOLS: GEOMETRIC INTERPRETATION
TABLE 2.4 (continued)
60 2 COUPLING OF TWO ANGULAR MOMENTUM VECTORS

TABLE 2.4 (continued)

2.3 C-G COEFFICIENTS AND 3 - j SYMBOLS: GEOMETRIC INTERPRETATION 61

TABLE 2.4 (continued)

--

"Symmetry relations [see Equation (2.26)] may be used in conjunction with this table to
evaluate all nonvanishing Clebsch-Gordan coefficientswith one angularmomentum argument
i,
equal t o o , 1, o r 2 .
62 2 COUPLING OF TWO ANGULAR MOMENTUM VECTORS

TABLE 2.5 Algebraic Expressions for Some Commonly Occuning 3 - j Symbols

where A5(X) = [ X ( X - 1)(X - 2)(X - 3)]-)

2.3 C-G COEFFICIENTS AND 3 - j SYMBOLS: GEOMETRIC INTERPRETATION 63

TABLE 2.5 (continued)

where A 2 ( X ) = [ X ( X - 1)(X - 2)(X - 3)(X -4)]-)

"Symmetry relations [see Equations (2.29)-(2.31)] may be used in conjunction with this
table to evaluate all nonvanishing 3 - j symbols with one angular momentum argument equal
too, $, 1, $,or2.
64 2 COUPLING OF TWO ANGULAR MOMENTUM VECTORS

NOTES AND REFERENCES

1. In the language of group theory, the set of functions PI m l ) span a (2 j l + 1) x (2 j l + 1)

representation V" of j: and jl,, while the lj2m2)span a ( 2 j2 + 1) x ( 2 j 2 + 1)
Vl
representation D'Z of j; and jz,. Then the direct product of the two sets, ml ,j2 m 2 ) ,
spans the direct product representation dl @ dz, which is in general a reducible
representation. The group theory content of Eq. (2.6) is that V" @ D'2 can be
decomposed into the irreducible representations V',where j ranges from j l + j2 to
V1 - j2 1 in unit steps:

2. There are numerous notations for the Clebsch-Gordan coefficients. Because m = ml +

m2 (as will be proved in the text), it is common to drop m and write C(jl j2 j ; mlmz),
as advocated by M. E. Rose, Elementary Theory of Angular Momentum (Wiley, New
York, 1957). We prefer the bracket notation ( j l ml ,j2 m2 bm) because it makes
very explicit the coupling (&coupling) of the angular momentum arguments involved.
A. R. Edmonds, Angular Momentum in Quantum Mechanics (Princeton University
Press, Princeton, NJ, 1974) uses the notation ( j l ml j2 m2 1jl j2 j m ) ; D. M. Brink
and G. R. Satchler, Angular Momentum (Oxford University Press, Oxford, 1979)
employ ( j lj2 ml m2 1j m ) and sometimesjust (ml m2 Vm);while L. C. Biedenharn and
J. D. Louck, Angular Momentum in Quantum Physics (Addison-Wesley, Reading, MA,
1981) use ~ $ , ' t n ' ~ For
, , , . a list of other notations, see Edmonds, p. 52, and Biedenharn
and Louck, pp. 150-152.
3. Please note that this treatment takes no account of the Pauli exclusion principle and
hence cannot be used for equivalent electrons (and other fermions).
4. See G. Racah, Phys. Rev., 62,438 (1942). This is equivalent to the phase convention
adopted by E. U. Condon and G. H. Shortley, Theory of Atomic Spectra (Cambridge
University Press, 1935).
5. An explicit formula for the Clebsch-Gordan coefficients was first derived by E. P. Wigner
in 1931 (see E. P. Wigner, Group Theory (Academic Press, New York, 1959),pp. 188-
191. A different but equivalent expression was obtained by B. L. Van der Waerden,
Der Gruppentheoretische Methode in der Quantenmechanik (Springer, Berlin, 1932),
pp. 68-71. Racah's derivation (see reference [3]) is the first not involving group theo-
retic arguments. Other derivations have been given by S. D. Majumdar, Progr. Theor.
Phys., 20,798 (1958), J. Schwinger in L. C. Biedenharn and H. Van Dam,Quantum
Theory of Angular Momentum (Academic Press, New York, 1965), pp. 241-246, and
R. T. Sharp, Am. J. Phys., 28, 116 (1960).
6. M. Rotenberg, R. Bivins, N. Metropolis, and J. K. Wooten, Jr., the 3 - j and 6 - j
symbols (The Technology Press, MIT, Cambridge, MA, 1959). See also Lundolt-
Bornstein, Numerical Data and Functional Relationships in Science and ~ e c h n o l o ~ ~ ,
K.-H. Hellwege, ed., (Springer-Verlag,Berlin, 1968) Group I, Vol. 3, "Numerical Tables
for Angular Correlation Computations in a-,/3-, and 7- Spectroscopy: 3 - j , 6 - j , 9 - j
r
Symbols, F- and -Coefficients."
7. K. Schulten and R. G. Gordon, J. Math. Phys., 16, 1961 (1975); Computer Phys.
Commun., ll,269 (1976).
8. K. Schulten and R. G. Gordon, J. Math. Phys., 16, 1971 (1975).
NOTES AND REFERENCES 65

9. Tables of Clebsch-Gordan coefficients or 3 - j symbols in algebraic form are given

by Edmonds, pp. 126-127 for angular momentum arguments and 2 , by R. Saito
and M. Morita, Progr. Theor. Phys., W , 540 (1955) for 5, and by D. L. Falkoff,
G. S. Colladay, and R. E. Sells, Can. J. Phys., 30,253 (1952) for 3. This information is
also collected and displayed in L. C. Biedenham and J. D. b u c k , Angular Momentum
in Quantum Physics (Addison-Wesley,Reading, M A , 1981),pp. 635-643.
10. This result was first obtained by E. P. Wigner, Group Theory (Academic Press, New
York, 1959), pp. 351-353.
11. For further investigationsof the correspondencebetween the quantum behavior for large
angular momenta and the classical limit, see P. J. Bmssaard and H. A. Tolhoek, Physica,
23, 955 (1957); G. Ponzano and T. Regge, in Spectroscopic and Group Theoretic
Methods in Physics (North-Holland, Amsterdam, 1968), pp. 1-58; W. H. Miller, Adv.
Chem. Phys., 25, 69 (1974); and K. Schulten and R. G. Gordon, J. Math. Phys., 16,
1961 (1975).
66 2 COUPLING OF TWO ANGULAR MOMENTUM VECTORS

APPLICATION 2
THE WIGNER-WITMER RULES

E. Wigner and E. E. Witmer, 2. Physik, 51, 859 (1928) were the first to derive
the number of different electronic states (terms) of a diatomic molecule formed by
bringing together two atoms 1 and 2 having orbital angular momenta L1 and L2
and spin angular momenta S1 and S2. This matter is of interest when describing
the nature of the compound state when two atoms collide or the nature of the atomic
fragments when a diatomic molecule dissociates.
An atomic term is written as 2 S + ' ~where
, 2 S + 1 is the multiplicity expressing
the degeneracy associated with the total spin S (which has 2 S + 1 values of Ms)
and the total orbital anguIar momentum L takes the values 0,1,2, ...and is denoted
by the capital Latin letters S, P, D, ....
For a (nonrotating) diatomic molecule, the total orbital angular momentum of
the electrons is not a good quantum number since the molecule does not possess
spherical symmetry and the orbital angular momentum operator does not commute
with the total molecular Harniltonian. However, the projection of the orbital angular
momentum on the internuclear axis is conserved since the diatomic molecule has
axial symmetry about a line passing through the two nuclei. The electronic terms of
a diatomic molecule are classified according to the value of this projection. The
absolute value of the projected orbital angular momentum along the axis of the
molecule is denoted by A . It takes on the values O,1,2, .... Terms with different A
are denoted by the capital Greek letters Z , n, A , ... corresponding to the capital
.
Latin letters S, P, D, ... for atomic terms with L = 0,1,2,. .. To this, the multiplicity
is affixed on the left as a superscript,that is, 2 S + 1 just
~ , as for the atomic term 2 S + 1 ~ .
The electronic wave function for a many-electron atom can be written as the
product of one-electron atomic orbitals, or more properly, as the antisymmetrized
linear combination of such products. Recall that the sum of angular momenta is also
an angular momentum. Hence the atomic term 2 S + 1 ~has a spin part that transforms
as ISMs) and a space part that transforms as I L ML) r YLML.
Let us consider first the case where the molecule consists of two different atoms.
Then A = I ML, + Mr, 1. Let L, and L< stand for the greater and lesser value of the
( L 1 , L2) pair, respectively. Let S and S< have a similar meaning.

A. Show that there are

( 2 S < + 1 ) ( 2 L < + 1)

C terms (called C states) of different multiplicity.

B. Show that there are a total of

APPLICATION 2 THE WIGNER-WITMER RULES 67

electronic terms where A ranges from 0 to L 1 + L2 and S ranges from IS1 - S 2 I

to S1+ S2, that is, from S , - S< to S, + S<.

In addition to rotation through any angle about the internuclear axis, a diatomic
molecule also has reflection symmetry in any plane that contains this axis. On
reflection the sense of the precession of L about the z axis (taken to coincide with
the internuclear axis) is reversed and M changes sign. Diatomic molecules with
A # 0 are doubly degenerate in the absence of rotation. To each value of the
energy there are two states that differ in the direction of the projection of the orbital
angular momentum on the internuclear axis. Hence the reflection symmetry leads
to no additional classification for nonrotating A # 0 states. However, for Z states
(A = 0). the state of the molecule is not changed on reflection, so that Z states are
nondegenerate. The wave function of a I: state can be multiplied by only a constant
phase factor p. Since double reflection in the same plane restores the system to its
original condition, p2 = 1,and it follows that p = f1. 2: states whose wave functions
are unaltered by reflection are called C+ states; those that change sign are C - states.
Recall that for each multiplicity there are (2 L< + 1) C states. The problem we next
confront is how many of these are C+ states and how many are C - states.
We may use the ILl MLl,L2 ML,) uncoupled representation as basis functions,
where MLl + ML, = 0. For M = IML,I = I ML, I # 0 , the basis functions can be
written as symmetric and antisymmetric products under reflection:

Clearly there are L< reflection functions of the form IC'; M ) and L, of )C-; M).
The energies of these nondegenerate C states are found by solving a set of
( 2 L< + 1) secular equations. The use of the reflection basis set IC*; M ) factors
the secular determinant into two blocks. Depending on the reflection symmetry of
1210 , L2 0 ) ,this will lead to a L, + 1 by L< + 1 secular determinant for the energies
of the C states of one reflection symmetry and a L< by L, secular determinant for
the other. Hence we need only establish how 1L 1 0 , L20 ) behaves under reflection to
complete this problem.
Consider a coordinate system with its origin at the center of atom 1 and its z axis
along the internuclear axis, as shown in Figure 1. It is an easy matter to convince
oneself that a reflection in the xz plane, o(xz), is equivalent to an inversion through
the origin, i, followed by a 180" rotation about the y axis, C2(y). This may be
established by sketching what happens to some representative points. Then

Under inversion, an atomic wave function remains unchanged or changes sign

depending on whether the parity p of the atomic term is even or odd, respectively.
 2 COUPLING OF TWO ANGULAR MOMENTUM VECTORS

FIGURE 1 Coordinate system for a diatomic

x molecule.

Thus the wave function is multiplied by the factor p whose value is determined as
follows.

C. Under inversion x --t -x, y -t -y, and z -r -2. Show that q5 -+ T + q5 and
l9-tT-6'.
D. Use the preceding result to prove that the spherical harmonics under inversion
obey the relation
iU,m(0,4) = (-l)eye,(@,4)
from which it is concluded that the parity of the state of a particle with angular
momentum t is independent of m. Hint: Show that under this transformation
6' -+ T - 8,andcos0 --t - c o s ~ ' .
E. For an atomic system, present arguments to show that
p = (-1)Ci'

where liis the orbital angular momentum quantum number of the ith electron
and the sum runs over all electrons.

From this equation we have

4x2) ILlO,L20) = P I P ~ ~ ~ILlO,L20)
(Y)

But the rotation by 180 about the y axis causes the transformation x --t -x, y --t y,
Z --t -2.

Hint: Use the fact that [LO) is proportional to PL(cos 6') to derive the above.
APPLICATION 2 THE WIGNER-WITMER RULES 69

Thus IL10 , LzO) is multiplied by the factor pip2 ( - I ) ~ L +on ~ zreflection, and
depending on whether this factor is +1 or - 1 there will be one more C+ or C - term,
respectively, for each multiplicity.
Consider now a molecule consisting of the same atoms (the so-called homonuclear
diatomic). There now exists a new symmetry, inversion of the electronic wave func-
tion through the midpoint of the internuclear axis. Those wave functions remaining
unchanged are called gerade, while those changing sign are called ungerade states.
A subscript g or u is attached to the right of the term symbol. If the two atoms are in
different electronic states (one excited and the other not, or one neutral and the other
ionized), the total number of possible terms is doubled, since each term can be either
gerade or ungerade. If both atoms are in the same electronic state, the total number
of states is the same as for a molecule composed from different atoms.
We must then sort out which states belong to gerade and ungerade symmetry. The
sorting process is involved [see P. Pechukas and R. N. Zare, Am. J. Phys., 40, 1687
(1972)l. The results are as follows:

If A is odd, the number of 2 S + 1 ~ gstates equals the number of 2S+1~,, states.

If A is even, the number of S + i ~ gstates is one greater or lesser than the number
of 2 S + i ~ , ,states. The "extra" 2 S + i ~state is g or u as S is even or odd.
If S is even, all I:+states are g and all C - states are u.
If S is odd, all Z + states are u and all Z - states are g.

Note that the g, u designations refer to the electronic wave function. Thus, for
example, this classification is used to describe the electronic states of 14N2 as well
as 1 4 ~ 1 5 ~ .

G. The ground state of the hydrogen atom is

H: 1 s 2~

and that of the oxygen atom is

0: ( 1 ~ ) ~ ( 2 ~ )3~~ ( 2 ~ ) ~

Find all the electronic states of Hz, OH, and Oz that can be built up by bringing
together their respective ground-state atoms.
70 2 COUPLING OF TWO ANGULAR MOMENTUM VECTORS

APPLICATION 3
THE ROTATIONAL ENERGY LEVELS OF A 2Z FREE RADICAL

To a first approximation we may imagine a diatomic molecule as a three-dimensional

rigid rotor. The Hamiltonian H for the total system would be simply

where Horepresents all those terms that fix the origin of the vibrational state v , B, is
the rotational constant of the molecule in the vibrational state v , and N2 is the square
of the rotational angular momentum. The wave functions that diagonalize H may be
written as
I ~ N M N=) Iv) INMN) = 10) Y N M ~4)( ~ , (2)

where Iv) is the vibrational wave function and YNMN

is a spherical harmonic. We
take

where uo is the band origin and

Hence the energy levels are given by

Actually, this is a fine starting point for a 'C diatomic molecule. In a 'Z molecule,
however, we must include the presence of nonzero electronic spin S, arising from
an unpaired electron. We add to the Hamiltonian a phenomenological term 7,N . S
to represent the effective interaction between N and S, where 7, is called the spin-
rotation constant for the molecule in the vibrational state v . We also include (to
first order) the nonrigidity of the molecule by introducing into the Hamiltonian
the vibration-rotation term -DUN4,where D, is called the centrifugal distortion
constant for the molecule in the vibrational state v . The term - D , N ~ expresses, in
part, the fact that in a nonrigid molecule the moment of inertia changes with the rate
of rotation. Thus the total Hamiltonian becomes

A. Use the coupled representation I N S J M ) to calculate explicit expressions in

terms of the rotational quantum number N for the energies of the rotational
i.
levels J = N + and J = N - Hence show that levels with the same value
APPLICATION 3 THE ROTATIONAL ENERGY LEVELS OF A ZZFREE RADICAL 71

of N are split (doubled) by spin-rotation splitting by the amount 7,( N + , 4)

which grows as N increases. Hint: Construct the 2 x 2 secular determinant,
using the IN N + $M ) IN iN - ; M ) basis functions. Use the relation
2 N . S = J~ - N~ - s2to evaluate the matrix elements of N . S.
B. Repeat part A but this time use the uncoupled representation IN M N , SMs).
Hint: Work specifically with IN 6i ) 1
M ~ , and N M N + 1, - $) Use the i
identity

to evaluate the matrix elements of N S appearing in the 2 x 2 secular determinant

with these basis functions.

Consider next the Zeeman effect in 21: molecules. Because 2~ molecules have an
unpaired electron spin, they are paramagnetic. Application of a magnetic field H
splits the M degeneracy of each J level (the Zeeman effect). The energy of
interaction between the magnetic dipole ps associated with the electron spin S and
the external field is

Each unit of spin momentum produces slightly more than two electronic Bohr
magnetons, that is,

where gs = 2.002 and po is the electronic Bohr magneton (- 1.4 MHz/G). We

choose the field direction to be along the z axis. Hence we write

If the field strength is weak, first-order perturbation theory may be used to calculate
the energy of each magnetic sublevel M of the level J :

C. Evaluate A E,( M) by rewriting 1N S J M ) in terms of the IN MN , S M s ) and

then calculating the matrix elements of X,. Show that

where the upper sign is for J = N + i and the lower sign is for J =N - $.
72 2 COUPLING OF TWO ANGULAR MOMENTUM VECTORS

At larger field strengths, it is necessary to add 'FI, to the molecular Hamiltonian

and evaluate the resulting new 2 x 2 secular determinant for its roots (eigenenergies).

D. Work in the coupled representation to obtain an explicit expression for the

energies of the magnetic sublevels of a 2Cmolecule in a magnetic field H. Show
that for small field strengths these expressions approach the results of first-order
perturbation theory in which the energies of the magnetic sublevels vary linearly
with H.

As the field strength increases, the energies of the magnetic sublevelsdeviate from
linearity in the field strength (onset of the molecular Paschen-Back effect) and at
still larger field strength the unpaired electron of the free radical decouples from the
molecule and acts as a free electron.

E. Plot a graph showing the variation of the 2Z energy levels with magnetic field
strength for the lowest eight levels (N = 0 , l ) . Use B, = 1 cm-', D, =
1x cm-', and 7 , = 1 x cm-'.
 Chapter 3
TRANSFORMATION
UNDER ROTATION

3.1 ANGULAR MOMENTUM OPERATORS AS GENERATORS

OF INFINITESIMAL ROTATIONS

We come to a matter both simple and profound. The symmetriesof a physical system
and its conservation laws are related to one another. To bring out this concept, let us
look at the action of a linear transformation U on the eigenvectorsI$j) of the system,
which changes its representation into a new set of basis functions I*.):

where the matrix representing ut is obtained from the matrix representing U by

taking its transpose complex conjugate. Equation (3.2) shows that in the {($'I)
vector space the lengths of the vectors and the angles between any two of them will
be preserved (will be the same as in the (($1) vector space) if

Transformationswhose matrices satisfy this condition are said to be unitary, for ut U

equals the unit matrix I (the identity matrix).
Equation (3.1) shows how the representationof a vector changes under the unitary
transformation U. It is convenient to know how to relate the matrices representing
the same operator Q in the two different representations. In the "old" representation
74 3 TRANSFORMATION UNDER ROTATION

Introducing the identity Ut u = I, we may rewrite Eq. (3.4) as

Thus a unitary transformation U transforms an operator Q into

Suppose that the operator U defining a unitary transformation approaches the identity
matrix I as the change E in some variable caused by the transformation becomes
smaller and smaller. Then U may be represented as

where E is an infinitesimal real quantity and the operator S is called the generator
of the infinitesimal transformation[l, 21. For the condition Ut u = I to obtain, it is
necessary that
st = S (3.8)

Operators that satisfy Eq. (3.8) are said to be Hermitian.

Two successive unitary transformations are still a unitary transformation. We
apply n successive infinitesimal transformations, each having a change 8 = ~ / n ,
so that to the same order in E
U = [ I + iSSIn (3.9)

We take the limit as n increases:

U = lim [ I + iSSIn = lim

6-tO WOO

In Eq. (3.10) the exponential of an operator means no more than its power series
expansion

Thus Eq. (3.7) is seen to be just the leading two terms of Eq. (3.10).
The operators U and S commute; thus
3.1 GENERATORS OF INFINITESIMAL ROTATIONS 75

which can be readily demonstrated by replacing exp( i&S)by its power series expan-
sion in S. The eigenfunctions of S are also eigenfunctions of U, and the eigenvalues
of S are unchanged (conserved) by the transformation U. Hence the invariance of
the system under the unitary transformation U implies a conservation law for the
eigenvalues of the Hermitian operator S, which is the generator of the infinitesimal
transformation[3].
We illustrate this result with three important examples. However, before we start
we must distinguish carefully between active and passive transformations[4]. In the
active sense, one transforms the physical system from one position to another, in the
passive sense, one transforms the coordinate frame. Both descriptions, of course,
are equivalent, but they lead to confusion unless it is made explicit what is being
transformed and what is remaining fixed. As we shall see, transformation of the
physical system in one "direction" is equivalent to transformation of the coordinate
frame by an equal amount in the opposite "direction."
First, let us consider the transformation T(t), which canies the physical system
from time to to time to + t; hence the time coordinate from to to t' = to - t:

The operator T(t) is called the time displacement operator. As t -+ 0 , T(t) t I.

For an infinitesimal change in time we can replace It') by its Taylor's series expansion

In Eq. (3.14) we denote the name of the time variable and the value of that variable
by the same symbol t. This lazy but traditional notation may cause some confusion
at first; please note in the power series expansion of exp( -td/at) that t is a constant
(the amount of the time change) and commutes with d/dt. Comparison of Eq. (3.14)
with Eq. (3.10) shows that

where 31 = ialdt (in units of tt = 1) is recognized as the Hamiltonian of the system

(for those systems whose Harniltonians do not depend explicitly on time). Hence the
symmetry of time invariance of the system implies the conservation of energy (the
eigenvalues of 'H).
In particular, if I$(to)) is an eigenstate of 7f with the eigenvalue Eo, then time
evolution of the eigenstate (not the time variable) through a time interval t causes
the eigenstate to transform as I$'(to)) = T(t) I$(to)) = ($(to - t)). Fort = to it
follows that I$'(t)) = l$(O)). Thus I$'(to)) has the same value at time t = to that
76 3 TRANSFORMATION UNDER ROTATION

($(to 1) has at time to = 0. Consequently, I$'(t)) can be obtained from J$(t)) by

letting I$(t)) develop backward in time from t to 0 :

Traditionally U(t) = T(-t) = T(t) is called the time evolution operator or time
development operator, which operates on the eigenstates of the system to cause them
to advance in time. Comparison of U( t) with T(t) again shows the expected relation
between active and passive transformations; a change in the active sense is equivalent
to an equal and opposite change in the passive sense.
Next, consider the one-dimensional displacement operator D( z) , which carries
the system from zo to zo + z; hence the position coordinate 20 into z' = zo - z:

Again D(0) = I, and for an infinitesimal displacement

We identify

where p, = -id/& (in units of ta = 1) is the z component of the linear momentum

operator. Thus p, is the generator of an infinitesimal displacement along z, and the
invariance of the system to such a change implies that the z component of the linear
momentum is conserved.
As our last example, we consider a rotational transformation. We define apositive
rotation as one that carries a right-handed screw in the positive direction (forward)
along its axis. Any rotation %(a) can be specified by giving three parameters, two
to fix its rotation axis ii and one to fix its rotation angle a about ii. For convenience,
let ii be chosen to point along the z axis. The transformation R,(+) then carries the
system from 40 to 4o + 4 and hence the angle 4o to 4' = 4o - 4 in the coordinate
frame:

By the same line of argument as before

3.2 PARAMETERIZATIONOF ROTATIONSBY EULER ANGLES n
and we make the identification

where we have replaced L, by J, since the algebraic properties of the opera-

tors L,, L,, L, are determined by their commutation relations, which are the same
for the more general operators J,, J,, J,. Thus the z component of the angular mo-
mentum operator is the generator of an infinitesimal rotation about z.
The choice of the z axis was for convenience. For an arbitrary rotation about 6 by
an angle a, we obtain

where
J . 6 = -i- 8 (3.24)
aa
If the system is invariant under rotation about the axis 6, then the component of the
angular momentum along ii is conserved.

3.2 PARAMETERIZATION OF ROTATIONS BY EULER ANGLES

As we have remarked, any finite rotation may be uniquely specified by giving the
three components of the vector aii, where a is the angle of rotation and 6 is a unit
vector pointing along the axis of rotation. For many purposes, however, an equivalent
but more useful description is achieved by introducing the three Euler angles 4, 8,
and x shown in Figure 3.1. Here the angles and 8 are familiar spherical polar
coordinates, where 9 measures the angle from the Z axis to the z axis and #IJ measures
the angle from the X axis to the projection of the z axis on the XY plane. The angle x
measures the angle from the line of nodes N, defined to be the intersection of the X Y
and xy planes, to the y axis. Thus x is an azimuthal angle about the z axis just as I$ is
an azimuthal angle about the Z axis. Note that the line of nodes ON is perpendicular
to both the z and Z axes.
The Euler angles 4, 8, and x should be regarded as defining a prescription
whereby the F = X Y Z frame (the space-fixed frame) may be made to coincide
with the g = xyz frame (the body-fixed frame) by three successive finite rotations
(Figure 3.2):

1. A counterclockwiserotation 4 about 2,the vertical axis. This carries the Y axis

into the line of nodes N.
2. A counterclockwiserotation 8 about the line of nodes N. This carries the Z axis
into the z axis (the figure axis of the body).
3. A counterclockwiserotation x about z , the figure axis. This carries the line of
nodes N into the y axis.
 3 TRANSFORMATION UNDER ROTATION

FIGURE 3.1 Euler angles 4,8,and x relating the space-fixed F = XYZ and molecule-fixed
g = xyz frames.

Note that all rotations are positive. This definition of the Euler angles 4, 0,
and X,which are also frequently written as a, /3, and 7 ,agrees with that of Rose[S],
Edmonds[6], and Brink and Satchler[7]. According to Eq. (3.23), we may express
the above prescription as

where the form of R is chosen to be an active rotation of the physical system[4,8].

For many purposes Eq. (3.25) is rather awkward because it is a mixed expression
involving angular momentum operators referred to both the F and g coordinate
frames. We now show that R ( 4,0, X) is equivalent to first a rotation x about the
Z axis, then a rotation 0 about the Y axis, and finally a rotation 4 about the same
Z axis:

The proof of this rather surprising result makes extensive use of Eq. (3.6),
namely, that a unitary transformation U carries an operator Q into UQU-'. Thus
exp(-i0 J N ) is the transform of exp( -iOJy) under the rotation exp( -i+ Jz), which
carried the Y axis into the line of nodes N. Hence
3.3 THE DIRECTION COSINE MATRIX ELEMENTS

Original Frame Final Frame

FIGURE 3.2 Transformation of the F = XYZ frame to the g = xyz frame having a common
origin 0 by three successiverotations, first Rz( b), then R N ( ~and
) , finally R,(x).

Similarly, exp( -ix J,) is the transform of exp( -ix Jz) under the previous rotation
exp( -iO JN),which carried Z into z:

Finally, we can always write the identity

since Jz commutes with itself. If one likes, Eq. (3.29) may be regarded as expressing
that exp( -ix Jz) is also the transform of the first rotation by the angle 4 about the
Z axis. When Eqs. (3.27)-(3.29) are substituted into Eq. (3.25), Eq. (3.26) results.
Equation (3.26) shows that the three Euler angle rotations may all be carried out in
the same coordinate frame if the order of the rotations is reversed!

3.3 THE DIRECTION COSINE MATRIX ELEMENTS

Let r be an arbitrary vector having the Cartesian components r~ = ( r x ,ry , rZ) in

the space-fixed (laboratory) frame and the Cartesian components r, = ( r,, r,, r,) in
the molecule-fixed (body) frame. These two equivalent descriptions of the vector r
80 3 TRANSFORMATION UNDER ROTATION

are related by a unitary transformation that is conveniently parameterized by the

Euler angles $,8, and X:

and

where is called the direction cosine matrix. The unitary transformation 0 is

conveniently expressed by the product of three Euler angle rotations

where $ is a rotation about the original Z axis, 8 is rotation about the new y axis that
coincides with the line of nodes N, and x is rotation about the final z axis. Equa-
tion (3.32) brings the F frame in coincidence with the g frame, and not surprisingly,
Eq. (3.32) is identical to Eq. (3.25).
Let the original coordinate frame be denoted by X Y Z , the coordinate frame
after the rotation Rz(b) by x'y'z', the coordinate frame after the rotation RN(8)
by xN y ' I z I' , and the final coordinate frame after the rotation R,(x) by xyz (see
Figure 3.2). Specifically,

I'[ z'
=Rz($) [c]
Z
= (
cos $ sin $
i n cos$
0 0
:) [i]
1

(3.34)
sin 8 0 cos 8

and

In Eq. (3.34), note that x" = x'cos 8 - z'sin 8, and so forth, because the "new" x"
coordinate is rotated out of the "old" x'z' first quadrant, as shown in Figure 3.2.
Equation (3.3 1) then takes the form
3.4 SPACE-FIXED AND MOLECULE-FIXED ANGULAR MOMENTUM OPERATORS 81

= (- cos x sin x
si; x cos x
0
9(
1
cos 8 0 -sin 8
0 1
sin 8 0 cos 8

cos 4
x ( s i n
0
sin 4 0

0
0)
C O S ~
1
I:[
where c and s represent cos and sin respectively. Because the <PgF are the real
elements of a unitary transformation, @g2
= QFg. Then Eq. (3.30) may be rewritten
as

~4 c o C X - S ~ s X - c $ c e S X - S ~C X ~4 se
~4 c e C X + Cs x~ - s 4 c o S X + Cc X ~ ~4 se
se SX ce
(3.37)
The elements appearing in the 3 x 3 matrices of Eqs. (3.36) and (3.37) are simply
the cosines of the angles between the various pairs of axes (see Figures 3.1 and 3.2),
that is, the direction cosines OgFand QFgt, respectively. They have the property that

where the indices F g on QFg denote the row and column, respectively. Equa-
tion (3.38) is nothing more than the familiar orthonormality conditions for the el-
ements of a real unitary matrix.

3.4 REPRESENTATION OF THE ANGULAR MOMENTUM OPERATORS

IN THE SPACE-FIXED AND MOLECULE-FIXED FRAMES

Previously we found that the projection of the angular momentum J on an arbitrary

spatial axis ii is the operator that generates an infinitesimal rotation about ii, that
is, J - ii = -ia/aa,, where a, is the angle associated with a rotation about 8. By
examining Figure 3.2 it is seen that the unit vectors iig, fie, and fi, associated with
82 3 TRANSFORMATION UNDER ROTATION

the three Euler angles can be expressed in terms of the Cartesian unit vectors fix, iiy ,
and iiz in the space-fixed frame by

fie = -fix sin 4 + fiy cos 4

ii, = fix sin 9 cos 4 + aysin 0 sin 4 + iizcos 8

When Eq. (3.24) is applied to Eq. (3.39), there results

a = -sin ~ J +xcos $Jy

-2-
ae
a
4- = sin Ocos 4Jx + sin esin 4Jy + cos 0J.z
ax
which may be solved algebraically to yield

~2=--.-
a2
do2
a 1
cote---(2L+--
30 sin2e 134~
a2
ax2
2 cos 0- a2
848~
) (3,411

It is readily verified that Jx, Jy , and Jz satisfy the normal commutation rules, that
is, [ Jx , Jy] = iJZ,and so on. These are the angular momentum operators in the
Euler angle representation referred to the space-fixed frame.
To find the expression analogous to Eq. (3.41) in the molecule-fixed frame, we
start once again with Eq. (3.39) but use the direction cosine transformation
3.4 SPACE-FIXED AND MOLECULE-FIXEDANGULAR MOMENTUM OPERATORS 83

to replace the unit vectors in the space-fixed frame by the unit vectors in the molecule-
fixed frame:

fie = fi, sin x + fi, cos x

A A

nx = nz (3.43)

Proceeding in the same manner as in Eqs. (3.40) and (3.41), we find

-2
. a
- = - sin 8 cos x J, + sin 8 sin x J, + cos 8Jz
34
a
4- = sin xJ, + cosxJ,
a0

which may be solved algebraically to yield

a
[
J , = i s i n x cote----
ax
a -1
sin e a4
-icosx-
ae

If we examine the commutation relations of J,, J,, and Jz, we find, for example,

and so forth. The commutation relations of J,, J,, and Jzare anomalous[9-1 l] in
that they differ from the normal ones by a change in the sign of i.
This result is more than a mere curiosity and can lead to irreconcilablephase errors
if simply ignored. To apply the angular momentum apparatus to molecular problems,
84 3 TRANSFORMATION UNDER ROTATION

the molecular angular momentum operators must satisfy three conditions: (1) be
defined with respect to a common center, (2) have the same axes of quantization, and
(3) obey a consistent set of commutation rules. The last condition poses a dilemma.
There are many ways to resolve this paradox[ll-181, but certainly one of the
simplest is to set the matrix elements of the raising and lowering operator of the
total angular momentum referred to the body-fixed frame J* = J , f iJ, so that it
behaves as J , = J x iJy in the space-fixed frame[l3-151. Let IJSZ) be the basis
set functions that diagonalize J~ and J,:

Then we insist that

The rationale for this choice is as follows. The space-fixed F = XYZ and
molecule-fixed g = xyz frames rotate about one another, having only the common
origin 0 fixed. Thus an observer in the space-fixed frame sees the molecule-fixed
frame rotate in a sense opposite to that of an observer in the molecule-fixed frame
looking at the space-fixed frame. This is exactly the same as time reversal in classical
mechanics, where the sign of the time variablet is changed to -t. This operation does
not affect the value of any coordinate q, but time reversal does change the sign of the
velocity dqldt and hence the signs of the linear and angular momenta. In particular,
the change in the sign of the angular momentum corresponds to a change in the sense
of rotation, that is, from clockwise to counterclockwise or vice versa.
What is the nature of the time reversal transformation in quantum mechanics? It
is insufficient to change only the time variable t into -t for then the time-dependent
Schrodinger equation, H$ = ( -h/i) d$/at, would not be invariant under time
reversal for a time-independent Hamiltonian. In quantum mechanics, momenta are
represented by operators of the form (h/i)d/aqthat do not explicitly contain the
time variable. However, in contrast to position coordinates, they do contain the pure
imaginary number i. This suggests a formalism in which time reversal corresponds
not only to substituting -t for t but also taking the complex conjugate[19]. Let 8
denote the time reversal operator. Then in this formalism

and in particular, 0 ~ ~ 8 =- -
l J , and 8 ~ ~ 3 =- -l J,. Note that the commutation
rule for the components of J changes sign when the sign of each component is
reversed. This suggests that all we need to do in referring J in the space-fixed frame
to the molecule-fixed frame is apply the time reversal operation to it, causing J* to
behave as J,. Then the commutation rules for J referred to the molecule-fixed frame
are the same as those for the internal angular momentum operators (e.g., L and S)
3.5 THE ROTATION MATRICES 85

defined in this frame[20]. Alternatively, we can refer molecule-fixed operators to

the space-fixed coordinate system, thereby removing the anomalous commutation
properties. This approach is discussed in more detail in Chapter 5. In conclusion, the
so-called anomalous commutation relations of J,, J,, and J, are no anomaly, but a
necessary consequence of time reversal symmetry.

3.5 THE ROTATION MATRICES

We return to the general problem of the transformation properties of angular momen-

tum under rotation. First we note that an arbitrary rotation cannot change the value
of J since J2 commutes with the rotation operator [defined in Eq. (3.23)]:

Thus when a rotation acts on the eigenstate IJ M ) of J2 and J z , it can only trans-
form [ J M )into a linear combination of other M values:

where the expansion coefficients

are the elements of a ( 2 J + 1) x ( 2 J + 1) unitary matrix for R, called the rotation

matrix[2 11.
Substitution of Eq. (3.26) reduces Eq. (3.53) to the form

In deriving Eq. (3.54) we made use of the identity

86 3 TRANSFORMATION UNDER ROTATION

A central problem to the development of the properties of the rotation matrices

is the evaluation of d L l M ( 8 ) .Equation (3.55) shows that the matrix d J ( 8 ) can be
calculated from the matrix representation of Jy . AS Wigner[211 first demonstrated,
the d L I M ( e )can be expressed as a finite polynomial in arguments of the half-
angle 812 :

where the sum over v is for all integers for which the factorial arguments are
nonnegative. One way in which this result may be derived is as follows.
As shown in Eq. (1.67), an arbitrary eigenfunction I J M ) of J2 and J, can be
constructed in terms of the spin functions a and /3 by

( a )J + M ( ~ J-M
)
IJM) =
[ ( J + M ) ! ( J - Ad)!]+

Let us use a' and p' to denote the rotated spin functions. Then Eq. (3.52) may be
recast as
(,I) J + M ( p l ) J-M

[ ( J +M ) ! ( J - M ) ! ] ;

= x
MI
D L l M (R)
(a)J+M1( p) J-M'
[ ( J + M')!(J -M I ) ! ] +
(3.59)

In order to evaluate D & , ~ ( Rwe) must find the values of a' and P' under the
rotation R. We use Eq. (3.55) to write

where we have replaced Jy by i o y [see Eq. (1.62)].In the second line of Eq. (3.60),
( J M'I and I J M) are row and column vectors. Expressing the exponential of a matrix
by its power series expansion we find that
3.5 THE ROTATION MATRICES

cos 912 - sin 912

sin 8/2 cos 812

Thus the rotation matrix elements for J = $ are

1
where the d L t M ( 9 )is simply the M'M element of the matrix shown in Eq. (3.61).
For the special case 4 = 0 , x = 0,the transformation of the spin functions is

a'
cos 'I2
= (sin 812
- sin 'I2
cos 9/2
) (i ) = (COS
sin 912
'I2 ) = 9 + p sin -
cos -
2
9 (3.63)
2

and

" = (sin
cos 912
9/2
- sin 912
cos 912
) (; ) ( -cossin 812
= 912 ) 9
= p c o s -2- a s i n
e
-
2
(3.64)
On substituting Eqs. (3.63) and (3.64)into Eq. (3.59) and setting 4 = 0 and x = 0,
we obtain the expression

J+M-I'
[ ( J + M)!(J- M)!]-$
( J + M)!
z ! ( J + M - i)! (a cos f) (p sin f) ]
[ C
i
(J -M)!
j.(J - M -j ) !
I
(' cos f ) '-*(-I (-a sin f) '1

where we have made use of the binomial expansion

( a + b)" = C n ! / [ k ! ( n - k ) ! ]arkbk
k
88 3 TRANSFORMATION UNDER ROTATION

We equate like powers of a and P by setting M' = M + j - i, that is, j = MI - M + i.

Hence

and Eq. (3.66) is readily rearranged to the form of Eq. (3.57) by replacing the index i
with the index u.
For reference purposes we list in Table 3.1 the matrix elements of d L I M e( ) for
5,
J = 0 , *,1, and 2. Explicit expressions for d L t M( 8 ) for J = 3 , 4 , 5 , and 6 may
be found elsewhere[22].
The dLtM satisfy a number of useful symmetry relations. Equation (3.57) is
invariant under the substitution M -t -MI and M' -t -M :

Moreover, the replacement of 8 by -8 in Eq. (3.57) yields

However, rotation by -6J is the inverse of rotation by 8, that is, d L j M ( - 8 ) =

[ dLt ( 8 ) 1- I . The d L I Mare real and are elements of a unitary transformation,
~pe~ificdly, [ d L t M ( 8 ) ] - l = [ d L I M ( 8 ) l t= d L M l ( 8 ) .Thus Eq. (3.68) may ds0
be written

Combination of the preceding results gives

Examination of Eq. (3.57) also shows that

3.5 THE ROTATION MATRICES 89

TABLE 3.1 Algebraic Expressions for the dL,,(0) for J = 0 , $, 1, $, and 2 .

d! ,(o)=-di
2 1-T -is; ( o ) = - s i n ( + )
C.J=l
d;,(e) =dL,,-,(e) =cos2($) =+(~+cos~)
d l =d l = sin2 (); = + ( I - cos 0 )
( 0 ) = dll,o(e) =- , ( ) =-do() = &sin 8
&(e) =C O S ~

D.J=$
dttw = d f t-i(e) = cos3 (+)
dji(e) = dt
-+-T
,(8) = -djb(e) = -di ,( 8 )
4-1 = -\/5cos2 (f) sin (i)

dt-i( e ) = d f ;+(el ,
= d Tf - T ( 8 ) = di,
-TT
,( 8 ) = \/Scos (f) sin2 (4)
d!-+(e) = - d i t i ( e ) = -sin3 (i)
d f T, ( 8 ) = d f t + ( 8 ) = c o s ( ; ) [ ~ c o )(;s ~ -21
90 3 TRANSFORMATION UNDER ROTATION

Because successive rotations correspond to multiplication of the appropriaterotation

matrices, we find

and

In particular when 8 = T, Eq. (3.74) yields the result

Hence for integral J a rotation by 8 = 2 T is the same as a rotation by 8 = 0 ,

in agreement with our physical intuition for bodies in three-dimensional space.
However, for half-integral J, there is a change of sign in going from 8 = 0 to 8 = 2 n,
and a rotation by 8 = 4 T is required to return the system to its initial state. Thus in
the range (O,2 T) the rotation matrices for spin angular momenta are double-valued,
the two values differing by a phase factor of - 1. This is another manifestation of
the nonintuitive behavior of spin angular momenta. Provided care is taken to make a
consistent phase choice throughout a calculation, this ambiguity, however, does not
affect observable quantities.
Finally, because the rotation matrices are unitary, they satisfy the sum rules

and
3.6 THE ROTATION MATRICES: GEOMETRICAL INTERPRETATION

where the argument R represents the Euler angles $,0, and X, and * is the
complex conjugate.

3.6 THE ROTATION MATRICES: GEOMETRICAL INTERPRETATION

We can readily interpret Eq. (3.52) once again in terms of the vector mode1[23].
The state IJ M ) is represented by a vector of length - 4 making a fixed
projection M on the Z axis. The indeterminancy in the values of Jx and Jy is
represented in the vector model by making J precess uniformly about the Z axis.
Suppose that the rotation R ($,6, X) transforms the system from the "old" X Y Z
frame to the "new" frame X'Y'Z' with a common origin 0. Then the OZ' axis will
be inclined at an angle 0 with respect to the OZ axis, and the probability for finding J
making the projection M' on the OZ' axis is just

Thus DL.~ ( 4,6, x) represents the probability amplitudethat an angular momentum

vector J making a projection M on the OZ axis will be found making a projection M'
on the 02' axis that is related to the original frame by the rotation R ($,6, x). The
rotation R includes the two Euler angles 4 and x representing azimuthal rotation
about the old and new 2 axes, respectively. Equation (3.54) shows that these angles
are associated with the phase factors e-iM&and e-*M1x,which can be seen from
Eq. (3.79) to play no role in the probability interpretation described above. On
the basis of the vector model, we should expect to find the spread in the values
of M' between the two dotted projection lines shown in Figure 3.3. However, the
vector model is not an exact description. The effect of quantum indeterminancy
of Jx and Jy is to allow values of M' to fall outside these limits. The probability
0) ] can be shown to die exponentially in these classically forbidden regions
and to oscillate in the classically allowed region[23]. The number of oscillations
increases with J , and, as expected, the envelope of oscillations approaches the
classical limit predicted by the vector model.
This geometric interpretation of the rotation matrices allows us to assess at once
without detailed calculation the relative absolute magnitudes of the dLtM(0)for
any pair of M, M' for a given J as a function of the angle 0. An instructive
example concerns molecular beam magnetic resonance[24-261. A schematic of the
experimental setup is shown in Figure 3.4. If an atom or molecule with a magnetic
moment p = gpoJ (where po is the electronic Bohr magneton, J the total angular
momentum, and g a proportionality constant called the g factor) passes through an
inhomogeneous magnetic field whose gradlent is perpendicular to the direction of
motion, the particle experiences a force proportional to the component of p projected
along the direction of the gradient. Let us choose the Z axis to be along the direction
of the magnetic field gradient. Then a beam will be split into 2 J + 1 components
on emerging from the inhomogeneous field, each corresponding to the expectation
value of Jz, that is, M. Thus in the classic Stern-Gerlach experiment, a spin
 3 TRANSFORMATION UNDER ROTATION

FIGURE 3.3 Vector model picture for the rotation of the state function 1J M ) from Z to 2'
by the angle 0.

system on passing through the inhomogeneous field is deflected along the trajectories
symmetrically displaced above and below the initial direction and produces two
symmetric spots when the particles impinge on a screen perpendicular to their initial
direction of motion. Each spot corresponds to particles in the state Ii i) I i).
or -
In the experiment shown in Figure 3.4, the first Stem-Gerlach magnet, labeled A, is
used as a state selector to prepare a beam of (noninteracting)particles in a particular
state I J M ) ; all other particles with M' f M are removed by beam stops. The
particles next pass through the so-called interaction region having a homogeneous
magnetic field along 2. In the interaction region external perturbations may be
applied, such as a radiofrequency (rf) field to induce changes in M. The particles
pass through a second inhomogeneous magnetic field, labeled B, which acts as
an M state analyzer and then are detected (counted) by various means, often by the
use of an electron impact ionization mass spectrometer. In the interaction region
as the homogeneous magnetic field strength is changed for a fixed rf frequency, or
alternatively as the rf frequency is swept for a fixed magnetic field strength, a different
count rate is observed when the energy difference between neighboring M levels
matches the energy of the rf photon (the resonance condition). In this manner an
rf spectrum of the beam species is obtained by using a molecular beam magnetic
resonance spectrometer[24-271.
Let us assume at first that the beam is unaffected in its journey through the
interaction region, that is, that no external perturbations are present in the C region.
We wish to know the transmission probability when the B magnet is inclined at
angle 8 with respect to the A magnet (double Stem-Gerlach experiment). Let the
detector be set to intercept only those particles in the state IJM'). We wish to
determine how many particles are incident on the detector, that is, what fraction of the
3.6 THE ROTATION MATRICES: GEOMETRICAL INTERPRETATION

Beam Source

8_
L@ Collimator

Particle
Detector

FIGURE 3.4 Schematic of a molecular beam magnetic resonance spectrometer. Two

Stem-Gerlach (inhomogenwus) magnets A and B are separated by an interaction region C
in which some external perturbation such as a radiofrequency (rf) field may be used to in-
duce transitions. Magnet A and its beam stops serve to prepare the state IJ M ) of a beam of
noninteracting particles; magnet B, beam stops, and detector serve to analyze particles in the
state IJ M ' ) . For simplicity, a J = 1 system is pictured.

particles in the state I J M ) find themselves later on passing through magnet B in the
state I J MI). Clearly, the answer to this question is given simply by I (JM IJ MI) 1' =
[ &IM( 9) 12 . Thus knowledgeof the rotation matrices is equivalent to understanding
the outcome of a double Stern-Gerlach experiment. Moreover, the vector model
picture for the rotation matrices allows us to make qualitative predictions.
Let us consider some special cases in the double Stem-Gerlach experiment. When
9 = 0•‹,Eq. (3.71) shows that [dLtM(0)12 = SMtM;all particles appear at the
same M value. This agrees with the vector model and is a consequence of the fact
94 3 TRANSFORMATION UNDER ROTATION

that IJ M ) is an eigenstate of Jz and a "remeasurement" of Jz cannot alter this state.

As 8 begins to increase, other M' # M states begin to appear, but those with the
same sign as M are the more populated (for M # 0). At 8 = 90•‹,many M' states
are present but the population in I J M ' ) is the same as that in IJ -M ' ) . Again this
result follows at once from the vector model or may be shown to be a consequence of
Eq. (3.75) with 8 set equal to 7r/2. As 8 becomes larger than 90•‹,M' states having
a sign opposite to M become the more populated (for M # 0). At 0 = 180•‹,the
vector model predicts that no particles with M' = M are transmitted for M # 0 ;
that is, the polarizer-analyzer pair of Stem-Gerlach magnets are "crossed." The
same result is seen to be a consequence of Eqs. (3.72) and (3.73), which show
that [ d L t M ( f r ) l 2 = SMt-M; all particles appear at the value of M' of the same
magnitude but opposite sign when the direction of the inhomogeneous magnetic field
is reversed.
Let us suppose that the perturbation in the interaction region C has the effect of
only altering the populations of the individual IJ M ) states. Then we represent the
new state IJ Mc) in terms of the IJ M ) basis functions as

where the fM coefficients are real and have the property that

The detector signal is then proportional to

For example, with the polarizer and analyzer magnets crossed, a fraction f& of the
original beam will be transmitted and the detector will respond only to transitions
in the interaction region in which M is changed into Mc = -M . In particular, let
us imagine that the rf field was replaced by another Stem-Gerlach magnet whose
gradient direction is at an angle Oc with respect to magnet A. Then

and

Equation (3.84) shows that particles will pass through the crossed polarizer-analyzer
combinationprovided the gradient direction of the intermediate Stern-Gerlach mag-
net is parallel to neither magnet A nor magnet B[28]. Again this conclusion is an-
ticipated on the basis of the geometric interpretation of the rotation matrices. Thus
3.7 THE SPHERICAL HARMONIC ADDITION THEOREM 95

magnetic resonance spectroscopy and a triple Stem-Gerlach experiment have many

similarities to each other.

3.7 THE SPHERICAL HARMONIC ADDITION THEOREM

An extremely useful relation between the coordinates of particles i and j is expressed

by the spherical harmonic addition theorem:

where ( di ,bi) and ( dj , mj) are the spherical polar coordinates of i and j ,respectively,
and Oij is the angle between the position vectors to these two particles (Figure 3.5).
In particular, for the case L = 1, Eq. (3.85) reduces to the familiar expression

cos Oij = cos Oi cos Oj + sin Oi sin Oj COS(),- dj) (3.86)

We may derive Eq. (3.85) by specializing Eq. (3.52) to the case of integral J,
which we denote by the orbital angular momentum L. Then in the spherical polar
coordinate representation

FIGURE 3.5 Spherical coordinates of particles i and j.

96 3 TRANSFORMATION UNDER ROTATION

where the rotation R ($,8, X) causes the primed and unprimed frames to coincide.
Let us consider the quantity

under the rotational transformation R:

where we have used Eqs. (3.78) and (3.87). Thus we conclude that Q is invariant
to rotation, that is, is a scalar quantity. We can select therefore any set of axes in
Figure 3.5. We choose Bi = 0,4i arbitrary, Oj = eij, and 4j arbitrary. From the
properties of spherical harmonics [see Eq. ( I S ) ]

and

which completes the proof of the spherical harmonic addition theorem.

3.7 THE SPHERICAL HARMONIC ADDITION THEOREM

We specialize Eq. (3.87) to the case where M = 0.We find

Reference to Eq. (3.85) yields the identification

which is the same as

In particular,
D& ( 4 , e . x ) = P L ( c o ~ @
Equation (3.93) may also be written

( L - M)!
= (-I) [(L + M)!]' P?( cos 0) e-iMd

where we have used Eq. (3.93) and the relation [see Eq. (1.56)]

YiM= (-llMyL -M
As we saw in Eq. (3.86), the L = 1 case has great practical importance. In
particular, the rotation matrix for L = 1 occurs so often that it deserves some special
consideration. For L = 1, Eq. (3.87) becomes
98 3 TRANSFORMATION UNDER ROTATION

The spherical harmonics Y I Mof order one have the explicit form in Cartesian
components

*
The factor r-' ( 3/47r) appearing in Eq. (3.100) is invariant under rotation so that
the transformation properties of YlM ( e , $ ) are determined by the linear combination
of Cartesian components indicated in square brackets. This suggests expressing an
arbitrary vector r in this so-called spherical basis. For the space-fixed frame we
define r; by

and for the molecule-fixed frame by

The two descriptions are related by

Note that Eq. (3.103) differs from Eq. (3.99) in the form of the rotational transfor-
mation. Comparing Eq. (3.103) to Eq. (3.30), we conclude that D z M 1( 4 , 8 ,X ) and
QFS ( 4 , 8 ,X) are related by a unitary transformation, which is the transformation
from a spherical basis to a Cartesian basis[29]. Thus the rotation matrix elements for
L = 1 are no more or less familiar than the direction cosine matrix elements; either
can be written as a linear combination of the other.

3.8 THE CLEBSCH-GORDAN SERIES AND ITS INVERSE

The general properties of the rotation matrices may be further developed by consid-
ering the connection between the uncoupled IJ1M I ) IJ2 M 2 ) and coupled IJ3 M 3 )
3.8 THE CLEBSCH-GORDAN SERIES AND ITS INVERSE

FIGURE 3.6 Vector model picture of the Clebsch-Gordan series. The angular momenta J1
and J2 precess independently, making the projections MI and Mz on the axis.

representationsunder a rotational transformation. Applying a rotation R to both sides

of Eq. (2.6) and using Eq. (3.52),we obtain the relation

We multiply both sides of Eq.(3.104) by (J1 Mi I ( J z M i I and obtain

which is called the Clebsch4ordan series. If desired, this can be given a sim-
ple probability interpretation (see Figure 3.6) in terms of the vector model pic-
tures shown in Figures 1.1, 2.1, 2.2, and 2.3. Suppose that the two angular mo-
menta J1 and J2 independently precess about the Z axis with projections Mi
100 3 TRANSFORMATION UNDER ROTATION

and M 2 . Consider a rotation through angular coordinates R. The probability am-

plitude that the projections of J1 and J2 in the new frame are Mi and Mi is given
by D;,, ( R) D;,, ( R ) . Let J 1 and J2 couple to form J3. The probability
1 1
amplitude that a p'dcular value of J3 is formed in the old coordinate frame is
( J 1M I ,J2 M2 IJ3 M 3 ) , the probability amplitude that the same value of J3 results
in the new coordinate frame is ( J , M i , J2 M i IJ3 M i ) , and the probability amplitude
that J3 having the projection M I + M2 = M3 in the old frame makes the projection
Mi + M i = Mi in the new frame is D$, ( R) . Combining these probabilities and
3 3
summing over all possible J values that can be the resultant of J1 and J 2 , we have
& ( J 1M I ,J2 M2 IJ3M 3 ) ( J 1M i , J2 Mi IJ3 M i ) D$,3 3 ( R ). This may be equated
to D$ ( R) D$ M2( R) to obtain the Clebsch-Gordan series. Thus we see that in
spite of the formidable appearance of Eq. (3.105)it has the geometrical interpreta-
tion of relating the way the uncoupled and coupled representations transform under
rotation.
An alternative expression is obtained by applying a rotation by R to both sides of
Eq. (2.5).Using the same procedure, we find

which is called the inverse Clebsch-Gordanseries.

3.9 INTEGRALS OVER PRODUCTS OF ROTATION MATRICES

Equations (3.105)and (3.106)allow us to evaluate the integral over two rotation ma-
trices having the same arguments where the solid angle element dl2 is @ sin 9 d8 d ~ :
3.9 INTEGRALS OVER PRODUCTS OF ROTATION MATRICES 101

The integration over the angles 4 and x are readily carried out:

The integration over the angle 8 becomes

Jo sin B ~ 4
O (8 ) = sin BOY P J (~cos 8 )
I0

where we have used the orthogonality properties of the Legendre polynomials [see
Eq. (1.59)].Thus Eq. (3.107) becomes

= 8 ~ ~ ( - 1 ) ~( J; l-- M
~ i1, J2Mll00)

x ( J l - M f ,J z M f l 0 0 ) (3.110)

However, according to Eq. (2.28) and with the help of Table 2.5,

= ( - I ) ~ ' ~ ~+' I () -~ ~J~ ~J , J , (3.111)

and similarly

( J 1- M i , ~ ~ ~ i 1=0( -0 1)) ~ 1 ' ~ ; (+2l ~) - ~) 6 J 1 J 2 (3.112)

Hence we conclude that[30]

102 3 TRANSFORMATION UNDER ROTATION

and a normalized rotation matrix is [ ( 2 J + 1) /8 ?r2] D L , ~R)( . In Eq. (3.113)

the phase is unity since ( - 1)2(Jl-Ml) = 1 regardless of whether J1 is integral or
half-integral.
With the aid of the Clebsch-Gordan series and Eq. (3.113), we evaluatethe integral
over a triple product of rotation matrices having the same angular arguments:

This integral occurs in various applications. We shall encounter it again in the

calculation of rotational line strengths. A special case of Eq. (3.114), for example, is
the integral of three spherical harmonics

where here dQ = sin 0 df3 d+. Equation (3.115) may be derived from Eq. (3.1 14)
by setting to zero the indices MI, M2 , and M3 and using Eq. (3.93). This integral
is needed for the calculation of the atomic structure multiplet splittings in Russell-
Saunders coupling[31]. If L2 = 1, this expression also occurs in the calculation of
matrix elements for electric dipole radiation between angular momentum states for
spinless particles.
The use of 3 - j symbols in place of Clebsch-Gordan coefficients allows us to
rewrite the results presented in this section. In particular, it is readily shown that
3.9 INTEGRALS OVER PRODUCTS OF ROTATION MATRICES 103

and that
104 3 TRANSFORMATION UNDER ROTATION

3.10 THE ROTATION MATRICES AS RIGID BODY ROTATIONAL

WAVE FUNCTIONS

Let $J M ($,e, X) be the rotational wave function of a rigid body so that $J M is an

eigenfunction of the Hamiltonian H with the eigenvalue E J Mthe , square of the total
angular momentum operator J2 with an eigenvalue J ( J + I), and the Z component
of the total angular momentum JZ with an eigenvalue M. We imagine that a rotation
R1 = ($1,81, X I )is applied to the system. Then $J M will be transformed according
to

where the value of the rotated wave function $iM = R1$ j M is the same as the
original wave function $J M at point P' that is carried into P by the rotation R1 (see
discussion on active vs. passive transformations):

Here we are using the Euler angles in two senses: (1) as coordinates to describe the
rigid-body wave functions $J M and $J M I relative to the space-fixed frame, and (2) as
the arguments of the rotation R1 = R1 ($1, 81, x l ) in the rotation matrix DLlM.
We wish to set R1 = R, that is, $1 = 4, 81 = 8, and = x in Eq. (3.121). The
effect of this is to bring point P' to the new 2' = z axis so that $' = 0 , 0 1= 0 , and
X' = 0. Then Eq. (3.121) reads

We multiply both sides of this equation by D ; , , ~($,e, X) and sum over the in-
dex M. From the unitarity of the rotation matrices we find
3.10 ROTATION MATRICES AS RIGID BODY ROTATIONAL WAVE FUNCTIONS 105

Writing M for M" and K for M , we obtain

Thus the general wave function for an asymmetric top is a linear combination of
rotation matrices.
Let us suppose that the rigid body has an axis of symmetry, which we may take to
lie along the z axis. Then an arbitrary rotation about this axis leaves the wave function
$J~ (4,8, X) unaltered, except for a possible phase factor, that is, the probability
I$jMI2 is unchanged[32]. Hence we require +jM (4,e1X) to be the same up to
a phase factor for a rotation about z through the angle X. Considering the form
of the rotation matrices, DC~ ( 4 , e, X) = eiMddhK(e)eiKx, this condition can be
fulfilled only if $jM(+,e,X) is proportional to one particular rotation matrix, say,
DC~ ( + , e lX) in the sum over K shown in Eq. (3.124). If we also insist that the wave
function is normalized [see Eq. (3.113)], then the wave function of a symmetric top
is given by

For the special case that the rigid body is a rigid linear rotator, K = 0, that is, J is
perpendicular to the z axis (the molecular axis), the symmetric top wave function
I J O M ) reduces to the spherical harmonic wave function I J M ) [see Eq. (3.94)],
provided we no longer integrate over the angle X.
The I J K M ) satisfy the differential equations [see Eqs. (3.41) and (3.431:

J z I J K M ) = -i-
a IJKM) = M IJKM)
84

Jz I J K M ) = -i- a
ax
IJKM)= K I J K M )
106 3 TRANSFORMATION UNDER ROTATION

Thus we identify M with the projection of J on the space-fixed Z axis and K with
the projection of J on the body-fixed z axis (the figure axis). From the geometric
interpretation of the rotation matrices, it is, of course, clear that they should be
proportional to the symmetric top wave functions. Historically, the symmetric top
wave functions were laboriously derived by solving the differential equations shown
in Eq. (3.126) subject to suitable boundary conditions for the wave functions, rather
than developing their identification with the rotation matrices[33-351. It should be
noted that Eq. (3.126) also applies to systems with half-integral angular momenta,
although the I J K M ) lose their meaning as wave functions of a classical symmetric
top.
The symmetric top wave functions are simultaneous eigenfunctionsof Jz and J,.
By rewriting Eq. (3.125) as

where the dependence on the variables 4, 8, and )t is explicit, it is easily seen that
the simultaneous specifications of the projections K and M cause the correspond-
ing eigenvectors of the states to transform as (JKI and ( J M ) , that is, reversed
from each other (contragradiently). The transformation to a scheme of description
in which ( J K I transforms as I J K ) is attained by means of the so-called conjugation
operator K, which is equivalent to the quantum mechanical operation of time rever-
sa1[19]. Under time reversal, as we have mentioned, the angular momentum J also
changes sign:

that is, Jx, Jy,and Jz are transformed into -J x , -Jy, and -Jz, respectively, so that
the components of the angular momentum operator obey anomalous commutation
rules. Thus the anomalous commutation rules of J referred to the rotating frame is a
consequence of the simultaneous specification of the eigenvalues of JZand J,. The
internal angular momenta L and S have normal commutation rules in the molecular
rotating frame essentially because these momenta involve the motion and spins of
the electrons for fixed nuclei, so that no question concerning molecular rotation can
enter.

NOTES AND REFERENCES

1. H.Goldstein, Classical Mechanics (Addison-Wesley, Cambridge, MA, 1953).

2. L.D.Landau and E. M. Lifshitz, Mechanics (Pergarnon Press, Oxford, 1969).
3. E. Merzbacher, Quantum Mechanics (Wiley, New York, 1961).
4. A. A. Wolf, Am. J. Phys., 37,531 (1969).
5. M. E.Rose, Elementary Theory of Angular Momentum (Wiley, New York, 1957).
6. A. R. Edrnonds, Angular Momentum in Quantum Mechanics (Princeton University
Press, 1957), 3rd printing with corrections, 1974.
NOTES AND REFERENCES 107

7. D. M. Brink and G. R. Satchler,Angular Momentum (Oxford University Press, 1962),

2nd edition with corrections 1979.
8. M. Bouten, Physica, 42, 572 (1969) points out a major inconsistency in the way
Edmonds[b] treats active and passive rotations. Unfortunately, Bouten also finds that
the treatment of Rose[S] has this confusion, but two errors are made that cancel each
other. We treat rotations here in the same manner as Brink and Satchler[7].
9. 0. Klein, Z. Physik, 58,730 (1929).
10. H. B. G. Casimir, Rotation of a Rigid Body in Quantum Mechanics (J. H. Woltjers, The
Hague, 1931).
11. J.H.VanVleck,Rev. Mod. Phys.,23,213 (1951).
12. R. F. Curl, Jr. and J. L. Kinsey, J. Chem. Phys., 35,1758 (1961).
13. K. F. Freed, J. Chem. Phys., 45,4214 (1966).
14. J. T. Hougen, "The Calculation of Rotational Energy Levels and Rotational Line
Intensities in Diatomic Molecules," National Bureau of Standards Monograph 115,
U.S. Government Printing Office, Washington DC,June 1970; J. Chem. Phys., 36,519
(1962).
15. I. Kovaks, Rotational Structure in, the Spectra of Diatomic Molecules (AkadBmiai
Kiad6, Budapest, 1969).
16. A. Camngton, D. H. Levy, and T. A. Miller,Adv. Chem. Phys., 18, 149 (1970).
17. I. C. Bowater, J. M. Brown, and A. Carrington, Proc. Roy. Soc., A333.265 (1973).
18. J. M. Brown and B. J. Howard, Molec. Phys., 31, 1517 (1976).
19. See E. P. Wtgner, Group Theory (Academic Press, New York, 1959),Chapter 26, "Time
Inversion." Actually only under special circumstances is the time reversal operator 0
simply the same as the conjugation operator K. For example, one must disregard spin
and work in the coordinate (not momentum) representation. See also A. S. Davydov,
Quantum Mechanics (Addison-Wesley, Reading, MA, 1965),pp. 43 1437.
20. This method is not the only means of remedying the inconsistency of the commutation
rules of J and P in the molecule-fixed frame, where P i s an intemal angular momentum
operator. An alternative procedure, first suggested by Van Vleck[ll] is to "reverse"
all internal angular momentum operators, that is, to replace by -P, where the
s u p e d p t bar is omitted to denote the reversed operator. One must also relabel the
basis set functions lPp) that diagonalize Fzby IP - p). It is then readily verified that
the reversed
-- internal
-angular
- momenta also satisfy anomalous commutation rules, for
example, P,Pu - PUP,= P,Pu - PUP,= iFz= -iP,. If we also agree to change
the sign of ieverywhere it explicitly occurs, such as in the use of raising and lowering
operators, the behavior of the angular momentum operators in the molecule-fixed frame
is the same as in the space-fixed frame. To quote Van Vleck[ll]: "The sign of i is never
important as long as one is consistent in usage, and aconsistent anomaly is effectively

*
-
no anomaly at all." This is the procedure also advocated by Freed[l3], who introduced
the symbol to describe this procedure. It is apparent that Eqs. (3.47), (3.48),
and (3.49) result. Moreover, we see that, for example,

(S'Z'(SZ ISZ) (S' - C ' ( S 2 1s- C ) = S ( S + l)SS'SSZ'Z

108 3 TRANSFORMATION UNDER ROTATION

(SJZJls* ISZ) J=L(SJ- 2'1s, 1s - 2 )

This causes off-diagonal matrix elements of J .Pto differ in sign. While both procedures
give the same energies when the molecular Hamiltonian is diagonalized, the reversed
angular momentum method implies a different phase convention for the resulting
molecular eigenfunctions.
21. E. P. Wigner,Group Theory (Academic Press, New York, 1959); U. Fano and G. Racah,
Irreducible Tensorial Sets (Academic Press, New York, 1959). In the language of
group theory, each D forms an irreducible representation (Darstellung) of the three-
dimensional rotation group of dimension ( 2 J + 1).
22. H. A. Buckrnaster, Can. J. Phys., 42,386 (1964); 44,2525 (1966).
23. P. J. Brussaard and H. A. Tolhoek, Physica, 23,955 (1957).
24. N. E Ramsey, Molecular Beams (Oxford University Press, 1963).
25. E. Majorana,Nuovo Cimento, 9,43 (1932).
26. E Bloch and I. I. Rabi, Rev. Mod. Phys., 17,237 (1945).
27. For the electric field analog, see J. C. Zom and T. C. English, Adv. Atomic Molec. Phys.,
9,244 (1973).
28. Not everyone can set up a triple Stem-Gerlachexperiment, but the results can be readily
appreciated by the analogy with linear polarizers for a light beam. When two polarizers
are crossed, no light is transmitted, but if a third linear polarizer is inserted between the
pair of crossed polarizers, light is transmitted provided the middle linear polarizer is
parallel to neither of its neighboring linear polarizers. See T. S. Carlton, Am. J. Phys.,
42,408 (1974); J. Chem. Ed., 52,322 (1975).
29. The explicit form of this unitary transformation is given in reference [5], pp. 62-67.
30. Equation (3.113) is the great orthogonality theorem for the continuous irreducible
representations of the three-dimensional rotation group; see reference [21]. It serves
as a jumping off point to look at finite point g r o u p A o s e symmetry operations acting
on an object of finite dimensions in which at least one point of the object remains fixed.
Each symmetry operation R leaves the object unchanged, and hence each symmetry
operation corresponds to a unitary transformation. In some suitable orthonormal basis
set the symmetry operation R will be represented by a unitary matrix whose elements
satisfy the relation T '(R):, = T '( R)&. The set of square matrices of all the symmetry
operations in the group is called a representation of the group. The set of orthonormal
functions ql, q2,...,& with respect to which these matrices are defined is called
the basis of the representation. Here the number k gives the dimensionality of the
representation. A representation is said to be irreducible if it cannot be decomposed
NOTES AND REFERENCES 109

into a sum of representationsof smaller dimensions each of which satisfy the symmetry
properties of the group. The analog of Eq. (3.1 13) for a finite group of order h (i.e.,
with h symmetry elements) is

where the summation index R runs over all h symmetry operations in the group and ti
is the dimensionality of the ith irreducible representation Ti.
For many purposes what is more important than the group representations is the
corresponding group representation characters, defined as the sum of the diagonal
elements (trace) of the matrix representation:

The reason for this is that equivalent representations, which are related to each other
by a unitary transformation, have the same character since the trace of a matrix is
unaffected by a unitary transformation. By summing both sides of Eq. (3.129) over
the indices p , v, a,and p, we obtain the result

r
where we denote the character of the irreducible representation '(R) of the symmetry
operation R by x(')( R). Equatio~(3.131) states that the sum of the squared moduli of
the characters of an irreducible representation equals the order of the group.
Let X(R) be the charactersof some reducible representationof dimension d, and let di)
'
be the number of times the irreducible representation with dimension di)appears in
the reducible representation. Then

and it follows that

We can solve for di)by multiplyingboth sidesof Eq. (3.133) by x('') ( R) 'and summing
over all symmetry operations in the group:

Recall that representationscorresponding to physically equivalent symmetry operations

have the same character. These are said to belong to the same class. Thus, instead of
110 3 TRANSFORMATION UNDER ROTATION

summing over all symmetry operations in a group, we need only sum over the different
classes weighted by the number of symmetry operations belonging to that class. Hence
Eq. (3.134) may be rewritten as

where h, is the number of elements in the cth class and the sum runs over all classes of
the group. Equation (3.134) or its equivalent form, Eq. (3.135), serves as the common
starting point in applicationsof the theory of finite groups to many problems of interest,
such as the crystal field splittings of atomic ions [see H. Bethe, Ann. Physik, 3, 133
(1929); J. S. Griffith, The Theory of Transition-Metal Ions (Cambridge University
Press, 1964); S. B. Piepho and P. N. Schatz, Group Theory in Spectroscopy (Wiley-
Interscience, New York, 1983)l and the symmetry classification of vibrational normal
modes [see E. B. Wilson, Jr., J. C. Decius, and P. C. Cross, Molecular Vibrations
(McGraw-Hill, New York, 1955); and S. Califano, Vibrational States (Wiley, New
York, 1976)l. Tables of the characters of irreducible representations of finite point
groups may be found elsewhere; see E A. Cotton, Chemical Applications of Group
Theory (Wiley-Interscience, New York, 1963, 1971), M. Tinkham, Group Theory and
Quantum Mechanics (McGraw-Hill, New York, 1964), and P. W. Atkins, M. S. Child,
and C. S. G. Phillips, Tables for Group Theory (Oxford University Press, 1970, 1982
with corrections).
31. E. U. Condon and G. H. Shortley, Theory of Atomic Spectra (Cambridge University
Press, 1935).
32. Prototypical examples of molecular symmetric tops are the monohalogenated and
trihalogenated methanes, CH3X and CHX3. These molecules have a three-fold axis
of rotational symmetry. It might be wondered why their rotational wave functions are
unaltered by an arbitrary rotation about the symmetry axis. The answer to this question
can be found in terms of the rotational symmetry of the moment of inertia ellipsoid of
the molecule, as discussed in Section 6.1.
33. E Reiche and H. Rademacher, Z. Phys., 39,444 (1926); H . Rademacher and E Reiche,
2.Phys., 41,453 (1927).
34. R. L. Kronig and I. I. Rabi, Phys. Rev., 29,262 (1927).
35. D. M. Dennison, Rev. Mod. Phys., 3,310 (1931).
APPLICATION 4 ENERGY LEVELS OF ATOMS WITH TWO VALENCE ELECTRONS 111

APPLICATION 4
ENERGY LEVELS OF ATOMS WITH TWO VALENCE ELECTRONS

The Central Field Approximation

The nonrelativistic Hamiltonian for an n-electron atom is

where Ze is the charge on the nucleus, e is the charge on the electron, r, is the distance
between the i th electron and the nucleus (taken as the origin), and rjj = Iri - rj 1 is
the distance between the ith and j th electrons. In the central field approximation the
preceding Hamiltonian is replaced by an effective Hamiltonian of the form

This allows each electron to be expressed as a spin-orbital

where R.,g is the radial wave function, Yemis a spherical harmonic, and (sm,)is
(I
the spin function, which may be written a, for spin up ;$)) or /3 for spin down
((i -+)). Because of the Pauli exclusion principle, the total wave function of the
central field Hamiltonian is the antisymmetrized product of the spin-orbitals

which is conveniently written as a Slater determinant

n m ( 1) 4se2rnk m, ( 1) ... $n,,e,,mhm, ( 1)

Q n , ~ , r n m,,
~, (2) $We2mt2 m,, (2) .. . $n,,e,,mt,m, (2)
(5)

&,e, mf, m., ( n) $s&mt2 ms2 ( n)

. ' '
4 n , , ~ m t ,m,
, ( n)

We see that if two e xtrons have the same spatial and spin coordinates, that is, occupy
the same orbital, the determinant has two identical columns and hence vanishes.
Moreover, if any two electrons are interchanged, this corresponds to the interchange
of any two rows of the Slater determinant and hence I$)changes sign.
A specification of an electronic configuration gives the d.values of all the orbitals
but it does not specify I$)completely since there are me and m, quantum numbers
112 3 TRANSFORMATION UNDER ROTATION

as well. In the central field approximation all the possible 2S+1 L terms arising from
the same configuration are degenerate, that is, have the same energy. The form of ]$J)
suggests that the eigenvalues of ti and si are all observables. This implies that the
total angular momentum of the atom, expressed as L and S, may also be observable.
This conclusion is certainly valid for the central field Hamiltonian, but for it to hold
for the nonrelativistic Harniltonian, it is necessary to show that the noncentral part,
namely, the Coulomb repulsion terms, commute with L' ,L Z , S' ,and S z
Consider the i th and j th electrons.

A. Show that la and Qj, do not commute with the nonrelativistic Hamiltonian by
finding explicitly the nonzero values of the commutators

and

B. Use the preceding results to show that the sum ti, + tj, does commute with
e2 / r i j , that is, that

Hence it follows that Lz commutes with 3t. By symmetry, Lx and L y also com-
mute with %, from which it is concluded that L' = L; + L$ + L; commutes with %.
Since the Coulomb repulsion terms contain no spin coordinates, S2 and S z also com-
mute with %. Finally, the inversion operator leaves terms such as e2/Tij unchanged.
It follows that the Coulomb repulsion terms have nonzero matrix elements only be-
tween states of the same L , S , and parity.

Construction of Antisymmetrized Two-Electron Wave Functions

in a Central Field

Let us restrict our attention to atoms having two valence electrons; we regard the
closed shells as simply effectively screening the nuclear charge. They play no role
in determining the relative energies of terms, although they do, of course, make the
dominant contribution of the absolute energy of a configuration and its terms.
APPLICATION 4 ENERGY LEVELS OF ATOMS WITH TWO VALENCE ELECTRONS 113

Consider two particles 1 and 2 having orbital angular momenta el and t2 with
projections me, and me,. We may write the coupled representation in either of two
ways:

If the two particles are electrons, we must insist that the Pauli exclusion principle
is satisfied, specifically, that the total wave function (both space and spin part) be
antisymmetric with respect to the interchange of the two indistinguishable electrons.
If the two electrons are equivalent, then nl el G m& ,so that

where ( - 1) el+e2 = 1 when el = 12. Similarly

For a two-electron wave function, we may always write it as a product of a space and
a spin part. For two equivalent electrons

Therefore, for the total wave function to be antisymmetric, it is necessary that

( - I ) ~ + ~ - '= -1,thatis,that L + Sbeeven.

D. Determine the terms that arise from the ( ns) configuration, the ( np) config-
uration, and the ( nd) configuration.
114 3 TRANSFORMATION UNDER ROTATION

If the two electrons are inequivalent, then this implies that nl el is not identical to
m&. The spatial part of the wave function may be written as
1
- []el(l ) e 2 ( w ~k~(--)el+ez-L
) I&( ~ ) , I ( ~ ) L M L ) ]

This is symmetric in the coordinates 1 and 2 for the plus sign and antisymmetric for
the minus sign. Thus the entire wave function will be antisymmetric if we multiply
by an antisymmetric spin function when we take the plus sign and a symmetric spin
function when we take the minus sign. The antisymmetric spin combination is

and the symmetric combinations are

E. Show that s2acts on the antisymmetric spin combination to give S ( S + 1) = 0

and the symmetric combination to give S ( S + 1) = 2. Combining these results,
we conclude that the plus sign is for a singlet state ( 'L) and the minus sign is for
a triplet state ( 3 ~ ) .

Calculation of the Term Separations Arising from a Single Configuration

To find the relative term energies, we need only consider the Coulomb repulsion term
e2/r12 and calculate its nonzero matrix elements between the states of the same L
and S (and parity, but of course they all have the same parity) of a configuration. We
use the well-known expansion

where 0 is the angle between rl and r;?and r, is the lesser and r the greater of rl
1'
andr2. TheLegendrepolynomial P ~ ( C OO) S = [ 4 ~ / ( 2 k +1)]TYkO(0,q5)may be
expressed as a sum over the products of spherical harmonics by using the spherical
harmonic addition theorem. Hence we obtain
APPLICATION 4 ENERGY LEVELS OF ATOMS WITH TWO VALENCE ELECTRONS 115

A general matrix element in the coupled representation has the form

(LMI e2/r12 ILM)
because different L or different M cannot be connected by the Coulomb repulsion
term. If we expand these states in an uncoupled representation, we obtain one or
more terms of the form

where

Here the labels a, b, c, and d refer to electrons in different spin-orbitals (most general
case).
It is customary to define the Slater-Condon parameters as

Provided we consider matrix elements of e2 /rl2 within the same configuration, the
radial integrals reduce to only those of the type F kand G ~ .

F. Work out the relative term energies for the terms arising from the p2 configura-
tion. Show that
116 3 TRANSFORMATION UNDER ROTATION

Actual observations do not agree well with this result, showing that the theory
is approximate. Much better agreement is obtained by considering the interaction
between terms of the same L, S, and parity arising from different configurations.
This effect is called configuration interaction.
The angular momentum machinery mastered so far allows this application to be
worked out in a rather straightforward manner but with much algebra. In Appli-
cation 11 we revisit this topic to show how spherical tensor techniques allow this
problem to be solved almost effortlessly.
APPLICATION 5 ANGULAR DISTRIBUTION OF RIGID ROTOR AXES 117

APPLICATION 5
ANGULAR DISTRIBUTION OF RIGID ROTOR AXES
FOLLOWING ABSORPTION OF PLANE POLARIZED LIGHT

The wave function of a rigid rotor is characterized by the total angular momentum J
and its projection M on the axis of quantization, that is, by I J M ) . It has the explicit
form
I J M ) = YJM(e,4> (1)

where 0 and 4 are the polar and azimuthal angles, repectively. Thus the probability
PJM(O,4) of finding the rotor axis pointing into the solid angle element dn =
sin 8 dB d$ when the rotor is in the state ( J M ) is given by

which is seen to be independent of the azimuthal angle 4. Note that this is normalized:

The probability amplitude of finding the rigid rotor in the state IJ M ) following
the dipole absorption of plane polarized radiation ( jph= 1 , m,h = 0) is proportional
to the Wigner coefficient (Clebsch-Gordan coefficient) (J1'M",10 IJ M ) , where
I J" M") is the initial state of the rotor. Thus the distribution of rigid rotor axes for
this particular M state is

(which has not been normalized). The MI1states are assumed to be equally populated
with no phase relations (meaning random phase relations) among them. Thus the total
probability is obtained by simply summing over all initial MI1 states, giving

The remainder of this application concerns the evaluation of Eq. (4).

A. Use the Clebsch4ordan series or its inverse to show that

118 3 TRANSFORMATION UNDER ROTATION

B. Make the identification

jl = J I 1 mI -
-Mu kl = K l 1
j2=1 mz=O k2=K-K"
j3=J mg=M k3=K

Then Eq. (5) becomes

Square both sides of Eq. (6)and sum over MI1. Use this to prove that

where the sum over K" is from K - 1 to K + 1 . Thus Eq. (7)has the explicit
value

C. Use the fact that '1 I2 to express

for K = 0 is equal to [ 4 ~ / ( 2J+ l ) ]lYJM
PJ( 8) in the form
P J ( 8 ) = asin2 8+ bcos2 8 (9)

Find the coefficients a and b as an explicit function of J" for the two types of
transitions, a P branch in which A J = J - J" = - 1 and an R branch in which
A J = J-JU=+1.

D. P j ( 8 ) can be recast into the familiar form

NOTES AND REFERENCES

where

and

Show that do -t for either branch as J -+ oo.

dois called the alignmentparameter. It ranges in value from +2 for a pure cos2 8
distribution to - 1 for a pure sin2 8 distribution. For do = 0, the distribution is
isotropic. This problem illustrates the use of plane polarized light to prepare aligned
molecules that may then be used as projectiles or targets in scattering experiments[l].
It is a special case of the production of anisotropically distributed targets by beam
excitationE21.

NOTES AND REFERENCES

1. R. N. Zare, Berichte der Bunsen-Gesellschaft, 86,422 (1982).

2. R. Bersohn and S. H . Lin, Adv. Chem. Phys., 41,67 (1969).
120 3 TRANSFORMATION UNDER ROTATION

APPLICATION 6
PHOTOFRAGMENT ANGULAR DISTRIBUTION
(CLASSICAL TREATMENT)

Here we consider the form of the spatial anisotropy of the photodissociation frag-
ments when a molecule is dissociated by a beam of linearly polarized light. We re-
strict ourselves to a one-photon electric dipole transition. Let ($, 8, X) denote the
Euler angles that describe the molecule-fixed axes relative to the space-fixed axes.
We choose the Z axis of the space-fixed frame to lie along the electric vector E of the
light beam and the z axis of the molecule-fixed frame to be along the direction of the
transition dipole moment p in the molecule. Then the probability Pdissof making a
dissociative transition is proportional to Ip . El2. For a given molecular orientation
(4, 0, X) we write

where Pdisshas been normalized to unity:

Let f (Om,4,) denote the final recoil distribution of the fragments of interest in
the molecular frame. Here ( Om, )6, are the polar and azimuthal angles about the
z axis (the direction of p). For some cases f ( Om, )4
, may take a particularly simple
form, namely,

An example is that of a diatomic molecule when the dissociation occurs sufficiently

promptly that the fragments have the same final direction as their initial orientation
when the dissociative transition occurred (the axial recoil approximation). Then
8; = 0 if p lies along the internuclear axis (a parallel-type transition) and 8; = n/2
if p lies in a plane at right angles to the internuclear axis (a perpendicular-type
transition). In the general case, f ( Om, 4,) is a more complicated distribution whose
exact form depends on (1) the relative orientation of the bond that cleaves with respect
to p, (2) the internal motions of the fragment (molecular rotation, bending, torsion,
etc.), and (3) the time elapsed between absorption and fragmentation.
APPLICATION 6 PHOTOFRAGMENTANGULAR DISTRIBUTION (CLASSICAL) 121

Whatever its form, f (Om,4,) can always be expanded in the complete set of
spherical harmonics

where the expansion coefficients are given by

Let I ( & , 4,) denote the angular distribution of the fragment of interest in the
lab frame where (O,, 4,) are the polar and azimuthal angles about the Z axis (the
direction of E). A particular molecular orientation ($,O, X) makes the contribution
Pdiss ( $,Or X)f ( Om, 4,) to I ( 8, ,4,), and the complete fragment angular distribution
is obtained by integrating over all molecular orientations (assumed to be randomly
distributed):

A. Show that the fragment angular distribution has the form

where

is the total cross section and

= 2 (P2(cos Om)) (9)

is the asymmetry parameter. Here (. - .) denotes an average over the recoil
distribution f ( Om,4,). Note that /3 ranges from 2 to - 1.
122 3 TRANSFORMATION UNDER ROTATION

B. Consider once more the limiting case of diatomic photodissociation in the axial
recoil approximation. Show that P = 2 , that is, that I( 0, ,#,) is proportional
to cos2 B,, for a parallel-type transition ( p parallel to internuclear axis), while
P = -1 , that is, I( B,, 4,) is proportional to sin B,, for a perpendicular-type
transition ( p perpendicular to internuclear axis).
APPLICATION 7 INTRODUCTIONTO SYMMETRIC TOPS 123

APPLICATION 7
INTRODUCTION TO SYMMETRIC TOPS

Consider a symmetric top with total angular momentum J that makes a projection M
on the space-fixed Z axis and a projection K on the body-fixed z axis. Let $ j K M
denote its wave function. By symmetry, its permanent dipole moment (if it has one)
must lie along z. Then the probability of the dipole moment pointing between 8 and
8 + dB is given by

(1)
where 8 is the angle between the z and Z axes.

Show that

where P,(cos 8) is the nth-order Legendre polynomial. This result has been
obtained by S. E. Choi and R. B. Bemstein, J. Chem. Phys., 85, 150 (1986).

Find the orientation of this system by evaluating the expectation value of cos 8.
Show that
MK
(cos 8) =
J ( J + 1)

The effect of a static electric field on the energy level structure of a symmetric top
can be calculated by perturbation theory provided the external electric field is much
smaller than the internal electric field of the electron-nuclei interactions (as is usually
the case). Then the perturbation term is

where the electric field E has been chosen to lie along the Z axis in the space-fixed
frame. Thus, according to first-order perturbation theory,
 3 TRANSFORMATION UNDER ROTATION

Hence symmetric top states with K = 0 do not show a first-order Stark effect
(nor do linear molecules in their vibrational ground states with K = 0 or diatomic
molecules in 'Z+ or 'C - states). The factor (cos 0) = M K / [ J ( J + I)] controls the
rotational state selection of polar symmetric top molecules as achieved by the electric
hexapole focusing technique; see S. R. Gandhi, T. J. Curtiss, Q.-X. Xu, S. E. Choi,
and R. B. Bemstein, Chem. Phys. Lett., 132,6 (1986).

C. In second-order perturbation theory, the molecular wave function is changed by

the presence of the static electric field and the resulting energy change is given
by
= x
J' K' M'
( J K M I H' IJ'K'M') ( J 1 K ' M ' (H' I J K M )
EJKM - EJ'K~M'
(6)

where the summation is over all J'K'M' not equal to J K M . As will be shown
in Chapter 5,
E~~~=BJ(J+~)+(c-B)K~ (7)

where B and C are rotational constants. Show that the second-order energy is
affected only by the two neighboring states J' = J + 1 and J' = J - 1 with
K t = K and M' = M , and that the sum of their effects is

For K = 0 we obtain the second-order Stark shift for a linear molecule in the
ground vibrational level or for a 'C or 'C - diatomic molecule:
+

For the special case of the J = 0 level, only the J = 1 level can interact with it, and
we have
APPLICATION 8 RESONANCE FLUORESCENCE AND RAMAN SCATTERING 125
--

APPLICATION 8
POLARIZED RESONANCE FLUORESCENCE AND
POLARIZED RAMAN SCATTERING: CLASSICAL EXPRESSIONS

Because molecular systems are associated with states of large angular momentum J,
it is useful in many cdses to seek an understanding of their radiative properties in
the limit where J approaches infinity. Here we assume that we may describe the
absorption and emission process classically where we replace the electric dipole
transition moment p of the molecule by a Hertzian dipole oscillator vibrating at the
transition frequency and pointing in the same direction as p and attached rigidly
to the molecular framework. Let the incident light be linearly polarized with its
electric vector E pointing along F = X, Y, Z in the laboratory-fixed frame. Let
it excite an absorption oscillator p pointing along g referred to the molecule-fixed
axes. We choose g to be a unit vector along x, y, or z and denote the unit vectors
perpendicular to g by g' and g". The probability for excitation is then proportional
to IEF . pgI2,that is, to a2 where aFg is the direction cosine matrix element that
P
relates the two unit vector F and g having as their common origin the center of mass
of the molecule. Let the emission oscillator be along h where h does not necessarily
coincide with g. The probability for emission with light plane polarized along H is
proportional to IEH . phi2,that is, to @ i h .It is traditional to call Illthe intensity of
the light emitted plane polarized in the same direction as the electric vector of the
incident light and Il the intensity of the light emitted plane polarized perpendicular
to the direction of the electric vector of the incident light. Thus we have

and

I,. = A a;, a;,h

where @'is a unit vector perpendicular to fi, the bars represent an ensemble average
over all initial orientations of the molecules, and A is a proportionality constant.
It is traditional to define the degree of polarization P by

and the polarization anisotropy R by

126 3 TRANSFORMATION UNDER ROTATION

We wish to develop expressions for P and R in terms of cosZ7, the average cosine
squared of the angle 7 between the absorption and emission oscillators

We expand h in terms of the components of the g frame

Hence

and thus

Substitution of Eq. (8) into Eqs. (1) and (2) yields

and

We list below the only nonvanishing products of four direction cosine matrix
elements averaged over all molecular orientations:

A. Using Eq. (ll), show that

APPLICATION 8 RESONANCE FLUORESCENCE AND RAMAN SCAlTERlNG 127

from which it follows that

and

Hint : Make use of the identity

@,2,+ a;(,,= 1
B. Derive Eq. (11). Hint : It is readily verified that

where 0 is the angle between fi and g. Consider the two equations

Square each equation and carry out the average over all molecular orientations.
Complete this problem by considering the identity (CF@ig)( C F , @iJgl)= 1.

For a Q branch transition (A J = 0), group theory arguments show that the
transition moment lies along J while for P and R branch transitions (A J = -1
and +I, respectively), the transition moment lies in a plane perpendicular to J. Let f
denote the absorption process and 1 the emission process. Then for a ( Q f , QJ)
resonance fluorescence transition, the absorption and emission oscillators coincide so
t h a t 7 = O , c o s Z 7 = l , a n d P = i,whilefor(~f,~1),(Qf,~~),(Pf,Q~),and
(Rf ,Ql) resonance fluorescence transitions, the absorption and emission oscillators
are at right angles so that 7 = n / 2 , cos2 7 = 0, and P = -;. Finally for a
resonance fluorescence transition of the type ( P f ,P i ) , ( PT, Ri), ( Rf ,P J ) , or
(Rf , RJ) ,the absorption and emission oscillators are both in a plane perpendicular
to J. Because radiative lifetimes are typically thousands of rotational periods or
longer, the absorption and emission oscillators are distributed uniformly about a circle
(uncorrelated) and make an average (acute) angle 7 = n/4 so that cos2 7 = and i
P = +. Hence, although P can span the range -1 to 1, P can assume only three
values,i,4, i,
and - in the classical limit[l-51.

C. Often molecular resonance fluorescence is not resolved. Find an expression for

the apparent polarization if the resonance fluorescence process consists of two
types of emission, one of total intensity I1of polarization PI and the other of
total intensity I 2 of polarization P2.Hint: R is linear in cos2 7 while P is not.
128 3 TRANSFORMATION UNDER ROTATION

Off resonance we may observe what is called the Rarnan effect. Here the electric
field of the incident light E induces a dipole moment p according to the relation

where g is a second-rank Cartesian tensor called the molecular polarizability tensor.

The induced Hertzian dipole then radiates. Its frequency of oscillation may be just
that of the electric field impressed on the molecule (Rayleigh scattering), or it may
have other frequency components caused by other internal periodic motions of the
molecule that modulate the frequency of the Hertzian dipole oscillator (rotational
and vibrational Raman scattering).
When Eq. (15) is written out in full in the laboratory frame, it looks like

As in all tensors relating to physical quantities g is symmetric, that is, QAB = QgA.
Let us consider an excitation-detection geometry in which the molecular sample
is located at the origin, the incident light propagates along the X axis with its electric
vector pointing along the Z axis and the scattered light is detected along the Y axis
for polarization along Z (giving the signal Ill)or for polarization along X (giving the
signal IJ. In this arrangement

where the bars again denote an average over all molecular orientations.
The direction cosine matrix 9 expresses the transformation from the space-fixed
to the molecule-fixed frames. We have

Then the polarizability tensor in the molecule-fixed frame is given by

D. Use Eqs. (15H20) to show that

APPLICATION 8 RESONANCE FLUORESCENCE AND RAMAN SCATTERING 129

Among the possible molecule-fixed coordinate frames we choose x,y, z so that

#'" is diagonal

where x, y, and z are the principal axes of the polarizability tensor.

Let us introduce the quantities

and

where a is called the mean value or isotropic part of the polarizability and 72 is called
the anisotropic part of the polarizability.

E. Show that the depolarization ratio

is given by[6-81

Hint: Use Eqs. (21) and (22) to find expressions for QZZ and axz. Square
these expressions and use Eq. (11) to carry out the averages over all molecular
orientations.

The depolarization ratio p can range from 0 when r2vanishes to % when a

vanishes. The latter limit cannot be reached in Rayleigh scattering since this would
imply a,, = G,, = a,, = 0 and thus there would be no scattered light. However,
it can occur in Raman scattering for bands in which the symmetry of the normal
vibration differs from that of the ground state.

F. Calculate p for Rayleigh scattering from C&.

130 3 TRANSFORMATION UNDER ROTATION

Resonance fluorescenceand Raman scattering are only two of several phenomena

involving the interaction of two photons with a molecule[9, 101. Some other exam-
ples are two-photon absorption and two-photon emission. In all of these processes
(unlike one-photon spectroscopieswith randomly oriented molecules),the intensities
are dependent on the polarization of the photons involved. In fact, the form of the
polarization dependence is the same in all cases. If the molecules are isotropically
distributed in space and do not move appreciably during the two-photon process (as
would be the case in a viscous fluid or a glassy matrix, but not in a crystalline sample),
the dependence is given by

where I is the observed intensity, fiA,kB are complex unit vectors representing the
polarizations of the two photons, and TABis a second-rank Cartesian tensor describ-
ing the response of the molecule to the radiation fields. Here Kz is a proportionality
constant that in the dilute absorber approximation[ll] contains the number of ab-
sorbers as well as the intensities of the two applied fields. The subscripts A and B
are taken to run over the space-fixed coordinates X, Y, 2, and repeated indices im-
ply a summation over all Cartesian components. The brackets (. . .) indicate an av-
erage over all the possible orientations of a single molecule. The Cartesian tensor T
is most conveniently expressed (and evaluated) in terms of the molecular coordi-
nates x, y, z [12]. The coordinate transformation is

where ado
is the direction cosine between A and B and so on. Thus we want to
average

where we make use of the fact that the direction cosines are real quantities. The
expression for the intensity becomes

All polarization information appears in the first factor in square brackets, all molecu-
lar information in the last factor, and all orientation information in the middle factor.
Thus the general polarization dependence of a two-photon process is contained in
the rotationally averaged product of four direction cosines[l3]. There are formally
3 = 6561 such products. However, as Eq. (11) shows, only a small number of these
products (of five different types) are nonvanishing.
NOTESANDREFERENCES 131

Similar considerations apply to higher-order multiphoton processes[l4]. For

example, the general polarization dependence of a three-photon process is contained
in
(@A~@B~@C,@D~@E,@F~)

and of a four-photon process in

( @ A ~ @ B ~ @ c ~ @ D ~ @ Ef ~@@~Fg @ ~ h )

The explicit integrations are straightforward but so numerous (3 l6 different cases

for a four-photon process!) as to be overwhelmingly tedious[l5]. Fortunately, a
systematic method for evaluating direction cosine averages of this type has been
found by Andrews and Thirunamachandran[l6], and the results for three-photon and
four-photon processes are available in the literature[l6-201.
The polarization dependence is often used to determine the symmetry of the ex-
cited state, particularly for molecules in a rigid medium[21]. This serves to introduce
us to the topic of photoselection[22-241, which is a sequential process involving po-
larized absorption followed by a polarized probing step in a rigid isotropic medium.
In principle, either or both processes might be multiphoton in nature, although single-
photon absorption followed by a single-photon probe step is by far the most common
application. Note that the state that is ultimately probed need not be the same state
initially prepared since a rigid medium preserves the spatial anisotropy created in the
polarized absorption step.

NOTES AND REFERENCES

1. P. P. Feofilov, The Physical Basis of Polarized Emission, ConsultantsBureau Enterprises,

New York, 1961.
2. V. L. Levshin, Z. Phys., 32,307 (1925).
3. E Pemn, Ann. Phys. (Paris), 12, 169 (1929).
4. A. Jablonski, Acta Phys. Polonica, 4,371 (1935); 5,271 (1936); 10,33 (1950).
5. M. McClintock, W. Demtrader, and R. N. Zare, J. Chem. Phys., 51,5509 (1969).
6. A. Anderson, ed., The Raman Effect, Vols. I and II, Marcel Dekker, New York, 1967.
7. E. B. Wilson, Jr., J. C. Decius, and P. C. Cross, Molecular Vibrations,McGraw-Hill, New
York, 1955.
8. G. Placzek, in Handbuch der Radiologie, Vol. VI, E. Man, ed., Akademische Verlagsge-
sellschaft, Leipzig, 1934.
9. W. M. McCIain, Acc. Chem. Res., 7 , 129 (1974); W. M. McClain and R. A. Harris, in
Excited States, Vol. 3, E. C. Lim, ed., Academic Press, New York, 1977, pp. 1-56.
10. D. M. Friedrich and W. M. McClain, Ann. Rev. Phys. Chem., 31,559 (1980).
11. In this approximation each molecule contributes to the overall transition matrix element
a term with a different phase factor. Consequently, for a large number of molecules the
interferencebetween matrix elements of different molecules provides a vanishingly small
contribution to the magnitude squared of the overall transition matrix element. Thus
the N molecules contribute incoherently, and the contribution is equivalent to N times
the contribution of a single molecule.
132 3 TRANSFORMATION UNDER ROTATION

12. For example, in single-color two-photon (electric dipole) absorption, the second-rank
Cartesian tensor Tabhas the form

for a transition from the initial state i to the final state f via the intermediate states e with
lifetimes 1 /re. Here p is the electric dipole transition operator and Eeiis the energy
difference Ee - Ei.
13. P. R. Monson and W. M. McClain, J. Chem. Phys., 53,29 (1970).
14. D. L. Andrews and W. A. Ghoul, J. Chem. Phys., 75,530 (1981).
15. An appealing alternative is to reexpress the Cartesian tensors in terms of spherical
irreducible tensors (see Chapter 5), which transform under rotation as the Wigner rotation
matrices. Then the integration over all orientations is readily accomplished. See
K. S. Haber, Ph.D. thesis, Cornell University, Ithaca, New York, 1984. See also B. Dick,
Chem. Phys., 96, 199 (1985).
16. D. L. Andrews and T. Thirunamachandran, J. Chem. Phys., 67.5026 (1977).
17. S. J. Cyvin, J. E. Rauch, and J. C. Decius, J. Chem. Phys., 43,4083 (1965).
18. W. M. McClain, J. Chem. Phys., 57,2264 (1972).
19. D. L. Andrews and W. A. Ghoul, J. Phys. A., 14,1281 (1981).
20. J. R. Cable and A. C. Albrecht, J. Chem. Phys., 85,3145 (1986).
21. E DW, "Polarized Light in Spectroscopy and Photochemistry," in Creation and Detection
of the Excited State, Vol. I , A. A. Lamola, ed., Marcel Dekker, New York, 1971, pp. 53-
122.
22. A. C. Albrecht, J. Molec. Spectrosc., 6,84 (1961).
23. A. C. Albrecht, Progr. Reaction Kinetics, 5,301 (1970).
24. T. W. Scott and A. C. Albrecht, J. Chem. Phys., 78,150 (1983).
APPLICATION 9 MAGNETIC DEPOLARIZATIONOF RESONANCE FLUORESCENCE 133

APPLICATION 9
MAGNETIC DEPOLARIZATION OF RESONANCE FLUORESCENCE:
ZEEMAN QUANTUM BEATS AND THE HANLE EFFECT

Let us consider the resonance fluorescenceprocess when an external magnetic field H

is applied. This causes the excited state of the molecule to undergo Larmor precession
before radiating. During a time t the total angular momentum vector J precesses
about H by the angle W Lt , where the Larmor precession frequency is given by

Here po is the electronic Bohr magneton and g is the electronic Land6 factor
expressing the quotient of the magnitude of the average magnetic moment along J
divided by the magnitude of the total angular momentum of the system.
Let the direction of H define the Z axis in the space-fixed frame. Then from the
viewpoint of the molecule, the presence of the magnetic field causes the X and Y axes
to precess at the frequency WL:

We specialize our considerations to a mutually orthogonal geometry in which the

molecular sample is located at the origin, the magnetic field is directly along the
Z axis, the incident plane polarized light beam propagates along the X axis toward the
origin, with its electric vector pointing along the Y axis, and the resulting resonance
fluorescence is detected along the Y axis and analyzed for its polarization along the
X axis. Supposethat at t = to a light pulse of short duration is incident on the sample.
Then at time t the resonance fluorescence signal is given by

where A is a proportionality constant, O;, expresses the probability of exciting

the Hertzian dipole absorption oscillator along g, a;,( H, t - to) expresses the
probability that the emission oscillator along h emits light plane polarized along 2
at the field strength H at time t , the bar means that an average must be taken over
all initial molecular orientations, and the factor exp[ -( t - t o )/TI expresses the fact
that the excited-state population decays radiatively with a lifetime T.

A. Show that
134 3 TRANSFORMATION UNDER ROTATION

FIGURE 1 Time-resolved fluorescence decay following pulsed excitation of NO2 on the

F2909-100,10transition (16,840.90 cm-') of the 5933 A band in an applied magnetic field of
2.6 G. The calculated spectrum is displaced for clarity. The NO2 excited state level has three
hyperfine components (I = 1 for 14N),each with slightly different g values and hence slightly
different Larmor precession frequencies, w ~ This. causes the appearance of the super-beat
structure as the contribution from each excited state hyperfine level comes in and out of phase.
Reproduced with permission from P. J. Brucat and R. N. Zare, J . Chem. Phys., 78,100 (1983).

where P = ( Ill- I J /( Ill+ I L ) as defined previously. Hence the resonance flu-

orescence decays exponentially but is modulated at twice the Larmor precession
frequency. This modulation is called Zeeman quantum beats[l].The degree of
modulation equals the degree of fluorescence polarization P. An example of this
effect is shown in Figure 1.

B. Suppose that either the incident light is plane polarized along H (the 2 axis)
or the resonance fluorescence is analyzed for polarization along H (the Z axis).
Discuss how the resonance fluorescence signal depends on the magnetic field
strength for these excitation-detection geometries.
NOTES AND REFERENCES 135

C. Instead of a light pulse of short duration, consider a continuouslight beam. Show

that I(H ) has a Lorentzian shape as a function of magnetic field strength with
a half-width at half-maximum of

This type of magnetic depolarization of the resonance fluorescence is called the

Hanle effect [2,3]. Determination of the value of H i permits the measurement of
the product gr. Classically, at small field strengths the Larmor precession rate is
much slower than the rate of radiative decay so that the emission is polarized
with a value close to that for H = 0 . As the field strength increases, the
Larrnor precession frequency increases. For large field strengths, the molecule
precesses many times on the average before radiating and the fluorescence
becomes completely depolarized.

D. Suppose that the intensity of the incident light beam can be modulated at the
frequency w according to

I ( t o ) = lo($ + 4cos wto) (6)

Use Eq. (4), the response of the resonance fluorescence signal to a pulse of
light at time to, to determine the steady-state response of the system. Consider
what happens to the resonance fluorescence signal if the modulation frequency
is swept or if the magnetic field strength is swept[4]. Explain how to design an
experiment to determine g and r independently.

NOTES AND REFERENCES

1. For examples, see E. B. Alexandrov, Opt. Spectrosc., 17, 522 (1964); J. N. Dodd,
D. M. Warrington, and R. D. Kaul, Proc. Phys. Soc., (London), 84,176 (1964); P. J. Bmcat
and R. N. Zare, J. Chem. Phys., 78,100 (1983).
2. W. Hanle, 2.Phys., 30,93 (1924).
3. R. N. Zare, Acc. Chem. Res., 4,361 (1971).
4. J. N. Dodd and G. W. Series, "Time-Resolved Fluorescence Spectroscopy," in Progress
in Atomic Spectroscopy, W. Hanle and H. Kleinpoppen, eds., Plenum, New York, 1978,
Part A, Chapter 14, pp. 639-677.
136 3 TRANSFORMATION UNDER ROTATION

APPLICATION 10
CORRELATION FUNCTIONS IN MOLECULAR SPECTROSCOPY

Spectroscopy concerns the frequency response of a system subjected to radiation.

The most common picture of this process is a static one in which radiation with
frequency w is absorbed or emitted by the system provided the Bohr condition
w = w f i is satisfied where

This time-independent view stresses finding the possible energy levels of the system
and the selection rules that govern their radiative connections. In this picture it is
useful to define a lineshape function (for absorption or emission) by

Here the initial state li) has energy E,, the final state If) has energy E f , pi is the
probability of the system being in state li), and I (f 16 . p ti) I2 is the square of the
transition dipole matrix element connecting i to f , where C is a unit vector pointing
along the electric field of the light wave, and p is the transition dipole moment
vector of the molecule. By switching to a time-dependent picture, we can extract
the dynamics of the system undergoing the i to f transition[l, 21.
We introduce the Fourier expansion of the Dirac delta function

and recall that

where we take C and p as real functions.

A. Show that the lineshape factor may be written as

APPLICATION 10 CORRELATION FUNCTIONS IN MOLECULAR SPECTROSCOPY 137

The sum over initial states a weighted by pi is simply the ensemble average, which
we denote by (. . .):

For an isotropic system, the same result is obtained independent of the direction
of polarization C. Hence

We call

the dipole correlation function. Equation (8) expresses the general result that the
lineshape or frequency response of a system interacting with radiation is the Fourier
transform of its dipole correlation function.
The dipole correlation function may be obtained directly from experiment by
inverting the Fourier transform given in Eq. (8):

(W

where the integration is over the spectral band of interest. It is convenient to

normalize the correlation function to its initial value:

Thus we define a correlation function of a unit vector ii( t) along p ( t ) by

and a normalized intensity distribution by

Then

Some experimentally derived correlation functions (real part) obtained from the
shape of the near-infrared rotation-vibration band of CO in various environments are
 3 TRANSFORMATION UNDER ROTATION

0.2 0.4 0.6 0.8

Time (1 0-l2sec)
FIGURE 1 Carbon monoxide dipole correlation functions in various environments. Spectra
were obtained from the following references: CHC13 and wC7HI6. M. 0. Bulanin and
N. D. Orlova, Opt. i Spektroskopiya, 15, 112 (1963) [Opt. Spectr. USSR (Eng. transl.), 15,
208 (1963)l: CC4, J. Lascombe, P. V. Huong, and M. Josien, Bull. Soc. Chim. France, 1175
(1959); A. R. Coulon, L. Galalry, B. Oksengom, S. Robin, and B. Vodar, J. Phys. Radium, 15,
58,641 (1954). [x, CO in CHC13 (liquid); a,CO in CC14 (liquid); +, CO in wC7HI6(liquid);
A, CO in argon (gas, 510 amagat); EJ, CO in argon (gas, 270 amagat); -, CO in argon (gas,
66 amagat); - - -,CO free (calculated)]. Reprinted with permission from R. G. Gordon, Adv.
Magn. Res., 3, 1 (1968).

shown in Figure 1[1]. The negative correlations in this gas-phase system mean that
after a time t, it is more probable that the CO molecule has swung around to a point
more than 90" from its direction at t = 0. For CO in liquids we anticipate that the
dipole correlation function is always positive.
APPLICATION 10 CORRELATION FUNCTIONS IN MOLECULAR SPECTROSCOPY 139

B. Because I(w) is a real quantity, that is, I(w) = I*(w), show that

from which it follows that

and
Im G(t) = -1rn G(-t)

Thus the real part of the correlation function is an even function of the time variable t,
while the imaginary part of the correlation function is an odd function of t. This
permits us to calculate the correlation function for positive time only:

C. Suppose that the dipole correlation function varies sinusoidally in time with a
frequency wo, that is, G(t) = exp(iwot). Find I(w) and discuss its form.
D. Suppose that G(t) has a sinusoidal variation in time with frequency wo but also
decreases exponentially with a half-life T , that is,
G(t) = r - I exp(-1tll.r) exp(iwot)

Find I(w) and discuss its form.

E. As another example, consider the classical infrared spectrum[3,4] of an equilib-
rium distribution of linear molecules of moment of inertia I at a temperature T.
Here the transition dipole moment vector lies along the molecular axis, and 'if
the latter rotates at an angular frequency O , then

The rotational energy of a linear molecule is given to a first approximation by

140 3 TRANSFORMATION UNDER ROTATION

where J = 0 , 1 , 2 , and so on. The rotational partition function[5] is

where 2 J + 1 is the rotational degeneracy factor. For sufficiently large T

or large I , J may be regarded as a continuous variable, and this sum can be
approximated by the integral
8?r2I ~ T
Qmt=lm(2J+ [ [ - 8J n( J2 +
I k1T) h 2 ] d J = (22)

Then the fraction of molecules between J and J + d J is given by

f( J)dJ =
( 2 J + 1) exp [- J ( J + 1 )h 2 ] d J
&lot 8 n2I k T

- ( 2 J + l ) h 2 exp [ - J ( J + l ) h 2 ] d J
87r2 I k T 8?r21kT
To obtain the probability f ( Q ) dn of finding a linear molecule rotating with an
angular frequency magnitude between Q and Q + dn ,we identify J ( J + 1) x
i
h2/ 8 n2I with the classical rotational energy I Q 2 , from which it follows that
( J + i ) h equals approximately 2 n I l Q 1, where the absolute value sign reminds
us that classically the angular frequency can be positive or negative. Show then
that

for0 5 Q 5 00.
Thus the normalized dipole correlation function is found by taking the ensemble
average over Eq. (19):

Show that
NOTES AND REFERENCES 141

Hinr : Do not explicitly evaluate Eq. (25);instead, think Fourier transform pairs.

Equation (26) has a simple interpretation: the classical infrared spectrum of linear
molecules at a temperature T has the form of a broad symmetric doublet of P and R
branch envelopes, wider at the outer wings and narrower at the inner wings than
that of the corresponding Gaussian curve. This example illustrates how dynamical
information can be extracted from a spectrum without resolving individual lines!
The dipole correlation function can be used not only to determine the spectral
density I ( w ) but also the degree of polarization of scattered or emitted radiation.
For example, the polarization anisotropy of resonance fluorescence at time t may be
written as

[see Eq. (14) of Application 81, from which R may be determined by taking a
suitable average over the time history of the system[6-8]. Once it was understood
by spectroscopists that the spectral density I ( w ) and its depolarization features are
related to the dipole correlation function G(t) and its various moments, much work
has been done to explore these connections[Z, 9, 101 and even in some cases to
calculate infrared or Raman spectra from first principles[11, 121.

NOTES AND REFERENCES

1. R.G. Gordon, Adv. Magn. Res., 3,l(1968).

2. B. J. Berne and R. Pecora, Dynamic Ligh Scattering, Wiley, New York, 1976.
3. H. Shimizu, J.Chem. Phys., 43,2453(1965).
4. A. G. St. Pierre and W. A. Steele. Phys. Rev., 184. 172 (1969).
5. G.Herzberg, Spectra of Diatomic Molecules, Van Nostrand, New York, 1950.
6. R. G. Gordon, J. Chem. Phys., 45,1643 (1966).
7. J. Husain, J. R. Wiesenfeld, and R. N. Zare, J. Chem. Phys., 72,2479 (1980).
8. T.Nagata, T.Kondow, K. Kuchitsu, G. W. Loge, and R. N. Zare, Molec. Phys., 50.49
(1983).
9. R.E.D.McClung, Adv. Molec. Relaxation Interaction Processes, 10,83(1977).
10. J.-Cl. Leicknarn and Y. Guissani, J. Molec. Strut., 80,377(1982).
11. E.J. Heller,Acc. Chem. Res., 14,368(1981).
12. D.R.Fredkin, A. Komornicki, S. R. White, and K. R. Wilson, J. Chem. Phys., 78,7077
(1983).
 Chapter 4
COUPLING OF MORE THAN
TWO ANGULAR
MOMENTUM VECTORS

4.1 THE 6 - j AND 9 -jSYMBOLS

When we couple two angular momentum vectors j1and j2,the resultant angular
momentum states are completely characterized by (jm). However, if we couple
together three angular momentum vectors, jl, j2,and j3,the resulting state Ijm)
does not completely specify the system because there is more than one way in which
these vectors can be added to form the same resultant. Let us first couple j1and j2to
give j12,then add j12to j3to give the resultant j. The eigenfunctions are

where 11 m l ,j 2 m2 ,j 3 m3) z (jl m3) are eigenfunctions of the com-

ml) I j 2 m2) (j3
pletely uncoupled representations. Alternatively, we could add jl to the resultant j23
of having coupled j2with j3. The eigenfunctions are then
 4 COUPLING OF MORE THAN TWO ANGULAR MOMENTUM VECTORS

These two representations of the same set of states must be physically equivalent.
Hence they are connected by a unitary transformation ljust as the coupled and
uncoupled representations of the addition of two angular momenta are related; see
Eqs. (2.5)-(2.9)]. We may write

where each expansion coefficient (j12j3j ( j 1j23j') E (j12j3jmVl j23j'm') is the

scalar product between the eigenfunctions taken from the two coupling schemes.
Because of orthogonality, this scalar product vanishes unless j = j' and m = m'.
Moreover, its value is independent of m because a scalar product does not depend
on the orientation of the coordinate system. The expansion coefficients are also
called recoupling coeflcients because they describe the transformation between
two different coupling schemes, in this case the l j l j 2 3j m ) scheme in which the
operators j 2 , j,, j:, and ji3 are diagonal, and the lj12j3 j m ) scheme, in which the
operators j 2 , j,, jT2, and j: are diagonal.
In what follows we develop an explicit expression for the recoupling coefficients.
We begin by inserting Eqs. (4.1) and (4.2) into Eq. (4.3) and equating the coefficients
of ( j lml ,j2 mz ,j 3 m3) on both sides of this equation. We find

Multiply both sides of Eq. (4.4) by ( j , m, ,j2 m2I j i 2 m1 + m2) and sum over ml
with ml + m2 = ml2 held constant. By the orthonormality of the Clebsch-
Gordan coefficients, the sum over ml on the left-hand side gives 6j,2!;2,reducing
the sum over jlZ to one term with j12 replaced by j i 2 . We drop this prime and
make the remaining summation explicit, using the fact that ml2 and m3 are fixed
and ml + m2 + m3 = m. We obtain the result
4.1 THE 6-jAND 9 - j SYMBOLS 145

The same procedure may be used to "isolate" the recoupling coefficient. Multiply
both sides of Eq. (4.5) by ( j 1 2mi2,j3 m3Vm)and sum over m'12with m = mi2+ m3
held fixed. Dropping primes, we obtain

Here we have found an expression for the recoupling coefficients as the contraction
of the product of four Clebsch-Gordan coefficients. As the latter are known, we have
in principle obtained all the recoupling coefficients for the addition of three angular
momenta. In particular, Eq. (4.6) shows at once that the coefficients of this unitary
transformation are real.
The first to characterize these recoupling coefficients was Racah[l], who defined
his W coefficients (called today Racah coeficients) in terms of these recoupling
coefficients by

The Racah coefficients were introduced because of their symmetry properties com-
pared to the recoupling coefficients. However, Wigner[2] has defined a closely re-
lated quantity, the 6 -j symbol, which has even higher symmetry. It is related to the
Racah coefficient by a simple phase factor

In what follows we work exclusively with the 6 - j symbols, which appear to be

gaining favor over the Racah coefficients in literature usage. We list below the most
commonly used properties of the 6 - j symbols. All these relations can be verified by
expressing the 6 - j symbols in terms of 3 - j symbols (or Clebsch-Gordan coefficients)
and using their previously derived properties.
146 4 COUPLING OF MORE THAN TWO ANGULAR MOMENTUM VECTORS

(a) Symmetries: The 6 - j symbol is invariant to the interchange of any two

columns; the 6 -j symbol is also invariant to the interchange of the upper and
lower arguments in each of any two columns. For example,

and so on. For the 6 - j symbol to be nonzero, the following four triangle
conditions hold for the vector addition of its angular momentum arguments:
A ( j l j 2 j 3 ) ,A( j l j 5 j 6 ) , A ( j 4 j 2j6), andA(j4j5j3). These conditions can be
illustrated in the following way:

This has the consequence that, for example, ( - l)jl+j2+j3 = ( - 1) - j l - J ~ - j ~ and

that (-1)2(11+iz+i3) = 1.
(b) Orthonormality:

(c) Expressions relating 3 - j and 6 - j symbols: A 6 - j symbol may be written as

the contraction of four 3 -jsymbols:

We may use the symmetry properties of the 3 -j symbols along with the facts
that ml + m:! + m3 sums to zero and that jl + j2 + j3 sums to an integer to
rewrite Eq. (4.12) as
4.1 THE 6 - j AND 9 - j SYMBOLS

Actually, only two of the six summation indices in Eqs. (4.12) and (4.13) are
independent, as the magnetic quantum numbers in each of the four 3 - j symbols
must sum to zero. Hence if one index, call it 6,is omitted from the sum, the
right-hand side of Eq. (4.12) or (4.13) must be multiplied by ( 2 j, + 1).
Equations (4.12) and (4.13) may take on a number of guises, which at
first appear contradictory, if not confusing! For example, we can change m3
to -m3 and m5 to -m5 since we sum over all values of these magnetic
quantum numbers in any case. If we also drop the sum over m5, Eq. (4.13)
becomes

By using the orthogonality properties of the 3 - j symbols, we may develop a

number of other relations. For example, by multiplying both sides by the last
3 - j symbol on the right-hand side of Eq. (4.14) we have

where we have used Eq. (2.32). Next we make use of Eq. (2.33) to rearrange
Eq. (4.15) into the form
 4 COUPLING OF MORE THAN TWO ANGULAR MOMENTUM VECTORS

where the formal sum over m6 on the left-hand side and m3 on the right-hand
side of Eq. (4.16) is omitted since for fixed ml ,m;!, m4, and m5 ,these sums
are only single terms. Continuing in the same manner, we obtain

and

(d) Special case when one argument vanishes:

By using the symmetry properties of the 6 - j symbol, we can rearrange a 6 -j

symbol with a zero in another position for application of Eq. (4.19).

The most convenienttabulation of 6 - j symbols is that of Rotenberg et al.[3], which

gives numerical values of all symbols up to a maximum value of 8 for any argument.
Algebraic expressions for the 6 - j symbols are given by Edmonds[4] when one of the
arguments is $, 1, $ ,or 2. In Table 4.1 we reproduce for convenience the algebraic
4.1 THE 6 - j AND 9 - j SYMBOLS 149

expressions when one argument is less than or equal to 2. In the Appendix we present
a simple computer program for the numerical evaluation of 6 -jsymbols.
The recoupling of four angular momenta leads to the 9 -j symbol. Once again
there is more than one (understatement!) possible coupling scheme, and all these
coupling schemes are related by unitary transformations. For example, we might
couplejl,j2,j3,andj4togethersothatj14 = j l + j 4 , j u = j 2+ j 3 , a n d j =j14+j23.
The eigenfunctionsin this coupling scheme are denoted by ((jl j4) j l 4 (j2 j3) j23 jm).
Alternatively, we might couple j12 = jl + j2,j34 = j3 + j4, and j = jI2 + j34 with
eigenfunctions I( jlj 2 )j12 jm). In analogy to Eq. (4.3), we write
(j3j4) js4

where the 9 -jsymbol is defined by

In the same manner as before, the 9 -j symbol may be rewritten as the product of
6 -j symbols and ultimately as the product of 3 -jsymbols. We list below the most
commonly used properties:

1. Symmetries: The 9 -jsymbol is unchanged by an even permutation of rows or

columns but is multiplied by the phase factor
( - 1) x(all i) = ( - l ) i 1 + h + i l z + j , + 1 4 + i u c + j 1 4 + j a + j

by an odd permutation of rows or columns.

2. Orthonormality:
150 4 COUPLING OF MORE THAN TWO ANGULAR MOMENTUM VECTORS

-
3. Contraction of 3 -j symbols and 6 j symbols:

and

4. Special case when one argument vanishes:

By using the symmetry properties of the 9 -j symbol, a 9 -j symbol with a zero

in another position can be rearranged for application of Eq. (4.25). Numerical
evaluation of the 9 - j symbols can often be conveniently carried out using
Eq. (4.23) or (4.24) and the numerical values available for the 3 - j and 6 - j
symbols. The Appendix presents a simple computer program that carries out
this task.

Higher n-j symbols can be defined[5] by considering the recoupling of five or

more angular momenta, but we shall have no occasion to use them.
4.2 GRAPHICAL METHODS 151

4.2 GRAPHICAL METHODS

Many problems in angular momentum theory confront one with an angry sea awash
with monsters, Clebsch-Gordan coefficients (Wigner coefficients), 3 -j symbols, and
the like, containing horrid sums over a variety of magnetic quantum numbers-and
yet when these sums are evaluated, often the dependence on these magnetic quan-
tum numbers disappears. Although such calculations can always be carried out alge-
braically, there exist graphical procedures that simplify these tedious manipulations.
It is the purpose of this section to outline these diagrammatic methods in a manner
that allows their facile application to many of the problems commonly encountered
in angular momentum theory. We follow closely the graphical methods of Yutsis,
Levinson, and Vanagas[5] as revised and extended by others[&12].
A 3 -j symbol is represented by three lines meeting at a common vertex or node.
Each line is labeled by its value of j and its magnetic quantum number m. A plus
sign or a minus sign is placed on the node indicating the order in which the arguments
appear in the 3-j symbol, read left to right. A plus sign means that the arguments
in the 3 -j symbol appear in counterclockwise order about the node; a minus sign,
clockwise order. Hence

Note that the orientation of the graph in space does not matter, nor do the length
of the lines or the angles between the lines. From Eq, (2.30) we see that changing
the plus sign or minus sign on a node corresponds to multiplying by the phase
factor ( - l)j1+j~+j3.Thus the permutation symmetry of the 3 -j symbol has been
built into this graphical representation. In particular, a change in the order of the
lines about the node accompanied by a change in the sign of the node is no change at
all. An even permutation causes a simple rotation of the graph, and hence also results
in no change.
The present notation does not handle the Clebsch-Gordan coefficient except by
explicit use of Eqs. (2.27) and (2.28). The important difference is that in a 3 -jsymbol
all the arguments have the same standing while in a Clebsch-Gordan coefficient
one of the jm pairs is the resultant of coupling together the other two. To obtain
a graphical method that incorporates both, we follow Brink and Satchler[6] by
introducing an arrow notation whereby an undirected line means

jm ilm' ,6..
11' mm'

and a directed line means

im
- 'I
1m
I
= (-llj-6.. 6
11' m-m'
 4 COUPLING OF MORE THAN TWO ANGULAR MOMENTUM VECTORS

Equations (4.28) and (4.29) serve as definitions of ingoing and outgoing arrows from
a j m pair. The m value changes its sign when passing through an arrow and a phase
factor of ( - l)j-m is introduced if the arrow points toward the j m pair, ( -l)j+mif it
points away[l3]. It follows directly from (4.28) and (4.29) that reversing the direction
of an arrow introduces a phase factor of ( - 1) = ( - 1)2j:

For a line with two arrows, we find

It is often convenient to add or drop an arrow from a line in a diagram. From the
above we can state the following rules:

1. To add an arrow pointing toward or to drop an arrow pointing away from a particular
j m pair, multiply the diagram by ( - 1) and change the sign of m in the diagram,
2. To add an arrow pointing away or to drop an arrow pointing toward a particular
j m pair, multiply the diagram by ( - l)Jdmand change the sign of m in the diagram.

Additional properties of the arrow notation are easily developed. In particular, we

can add arrows on the lines coming from a vertex in a diagram provided they all point
the same way, either into or out of the vertex:
4.2 GRAPHICAL METHODS 153

The proof of this relies on Eq. (2.31) and the fact that ( - 1)m1+m2+m3 = 1 since
ml + m2 + m3 = 0 for any 3 - j symbol that does not vanish.
With these conventions, and the notation that [j ] = 2 j + 1 , we represent Clebsch-
Gordan coefficients as follows [see Eqs. (2.27) and (2.28)]:
 4 COUPLING OF MORE THAN TWO ANGULAR MOMENTUM VECTORS

In Eq. (4.33) we have made the operations explicit to illustrate graph manipulations.
The reader is urged to check the logic of each step before continuing with this section.
Of particular interest is the case in which one of the angular momenta of a Clebsch-
Gordan coefficient or a 3 - j symbol is equal to zero. The results are already known
(see the first entry in Tables 2.4 and 2.5). Hence in terms of diagrams we find that
( j l ml ,00Ij 3 m3) = 6j1j3 6, becomes
4.2 GRAPHICAL METHODS

Hence any node involving a zero line (a line with j = 0 ) may be removed by use of
Eqs. (4.34) or (4.35).
More complicated diagrams are constructed by joining together simpler diagrams.
Often we encounter calculations involving products of 3 - j symbols in which the same
angular momentum appears in two 3-j symbols with either the same or opposite
magnetic quantum numbers. We represent this by joining together the free lines
representing the same angular momenta. Thus, for example,
 4 COUPLING OF MORE THAN TWO ANGULAR MOMENTUM VECTORS

and similarly
4.2 GRAPHICAL METHODS 157

It is conventional to omit the m label on a line in representing a sum over magnetic

quantum numbers. Hence

and

It is useful to introduce some additional definitions. An internal line is one that

connnects two nodes. An external line has one end free and the other end connected
to a node. Closed diagrams have no external lines. We illustrate the meaning of the
above by constructing some simple diagrams.
The first orthogonality relation for the 3 - j symbols [see Eq. (2.32)] becomes
158 4 COUPLING OF MORE THAN TWO ANGULAR MOMENTUM VECTORS

while that for the second orthogonality relation [see Eq. (2.33)] becomes

Equation (4.40) shows how a loop inserted into a line can be removed by a diagram-
matic manipulation. Suppose we set j 3 = j i , m3 = m i , and join j 3 m3 to j; mi in
this figure and sum over m3. Then, of course, we get unity provided jl ,j2, and j3
form a triangle, denoted by A (j l j2 j 3 ) . This is representated graphically by

Equation (4.42)is our first encounter with a closed diagram. The symbol A ( j l j2 j3)
has the value 1 if j, ,j2, and j3 satisfy the triangle condition [see Eq. (2.1I ) ] and zero
otherwise.
As another example of the connected line summation convention we give a
graphical representation of the 6 -j symbol [see Eq. (4.12)],which is a closed diagram
with six angular momenta:

It is noted that all vertices have plus signs; the arrows go around the perimeter in
a clockwise manner; the 6 - j symbol is nonvanishing only if the angular momenta
meeting at each vertex satisfy the triangle condition [see Eq. (4.10)]; and the pairing
of the angular momenta is such that (j l ,j 4 ) , (j 2 , js ) , and ( j 3 ,j6 ) have no vertices
in common.
The symmetries of the 6 - j symbol are readily verified by reference to the graph
shown in (4.43) and with the help of the two identities: ( - 1)4~ = 1 for any integral
4.2 GRAPHICAL METHODS 159

or half-integral j ; and ( -1)2j*+2jb+2jc

= 1 when j,, jb, j, form a triangle (i.e., enter
or leave a node), since j, + jb + j, must then be an integer. We illustrate some of
these symmetries. For example, by rotating the diagram in Eq. (4.43) about an axis
perpendicular to the plane of the figure and passing through the central vertex, we
have

from which it is seen that the 6 - j symbol is invariant to cyclic interchange of its
columns. Consider next a rotation by 180" about one edge, for example, the j4 line.
We find that

where the phase change from reversing the directions of all the arrows cancels that
from reversing the signs on all nodes. Equations (4.44) and (4.45) establish that the
160 4 COUPLING OF MORE THAN TWO ANGULAR MOMENTUM VECTORS

interchange of any two columns of a 6 -j symbol leaves the symbol unaltered. Finally,
consider the operations of pulling the central vertex to the periphery:

from which we find that the interchange of the upper and lower arguments in each of
any two columns leaves the 6 - j symbol unaltered. Equations (4.44X4.46) are the
same as Eq. (4.9).
Often the 6 -j graph does not have the symmetric form illustrated in Eq. (4.43)
but is geometrically distorted with the wrong arrows and plus and minus signs. For
example, by pulling the central vertex in Eq. (4.43) to the left, we obtain the following
equivalent diagrammatic representations of the 6 -j symbol:

(4.47)
where we have added arrows to all the lines of the vertex on the left to obtain the
diagram on the extreme right. A useful way for finding the appropriate 6 - j symbol is
to note that the angular momenta at each node, with a plus sign, taken in a clockwise
manner, are in the same order as in the 6 -j symbol read left to right [see Eq. (4.43)].
4.2 GRAPHICAL METHODS 161

The next closed diagram in increasing order of complexity is that representing the
9 -jsymbol:

Please note that in Eq. (4.48) the lines do not form a node at the center. Equa-
tion (4.48) has the properties that (1) all the vertices have minus signs (or equiva-
lently all plus signs); (2) none of the lines have arrows, although the diagram can be
put into so-called normal form by adding arrows to every other vertex so that each
internal line has an arrow; and (3) the angular momentum arguments in each row of
the 9 -jsymbol and in each column of the 9 -j symbol meet at a vertex. Once more,
the symmetries of the 9 -j symbol are readily derived by graphical manipulations of
Eq. (4.48).
The power of the graphical method is very much enhanced by the following
techniques for the reduction of closed diagrams to products and sums of products
of simpler diagrams. Consider a diagram that can be decomposed into two parts,
called blocks, denoted by A and B, such that A has no external lines and is joined
to B by only a single angular momentum line j. We assume that A is in or can be
placed in so-called normal form with one and only one arrow on each line. In this
situation rotational invariance of the diagram causes j = 0. The reason for this is the
following argument. A closed block in normal form must represent a scalar because
of the summation over all magnetic quantum numbers associated with each line in
the closed block. For example, the 6 - j and 9 - j symbols are closed blocks. Hence
the closed block A is proportional to some number times the zero line j = 0 , m = 0.
If a single line j joins to the closed block A, then by Eq. (4.27) this line must also be
a zero line. We may apply Eq. (4.35) to erase the zero line and hence to factor the
graph, as shown:
162 4 COUPLING OF MORE THAN TWO ANGULAR MOMENTUM VECTORS

Here the dashed line on B indicates any number of external lines.

Equation (4.49) is of fundamental importance in reducing the complexity of
graphs. Therefore, we pause to present a very simple instance where j = 0. Consider
the expression

which has the value Em, -*

Sjo Smo ( 2 j' + 1 ) = ( 2 j' + 1) 4. This expression may
be represented graphically by a closed diagram with one external angular momentum
line:

0 - 1 - (4.50)

i'

Applying Eqs. (4.34) and (4.40) to Eq. (4.50), we find that

from which it is seen that the jm line is a zero line.

Next we examine the case where A and B are connected by two lines. We reduce
this diagram by application of Eq. (4.41) followed by Eq. (4.49):
4.2 GRAPHICAL METHODS 163

To ensure that the diagram to the left of the j = 0 line is normal, we incorporate any
mows on jl and j2 in the B block. If the B block is completely empty, Eq. (4.52)
becomes simply

This result shows that if a normal diagram has two free lines, these must represent
the same total angular momentum; moreover, the nonzero value of such a diagram is
independent of the magnetic quantum numbers on the free lines.
For the case where A and B are connected by three and only three lines, the
same procedure may be used to reduce the diagram into the product of two simpler
diagrams:
164 4 COUPLING OF MORE THAN TWO ANGULAR MOMENTUM VECTORS

Here we can drop the two oppositely directed arrows in going from the second to last
to the last line in Eq. (4.54) because the resulting product of phase factors cancels.
Again we assume that A is normal and that any arrows on jl ,j2, or j3 are incorporated
into B. Note that the plus and minus signs may be reversed with no change in the
overall value of Eq.(4.54).
As a corollary, if the B block is completely empty, Eq. (4.54) becomes simply

This result is essentially the Wigner-Eckart theorem, which is discussed in more

detail in the next chapter. It is seen that a normal diagram having three free lines may
be replaced by a product of two factors: one factor is a closed diagram that can be
put into normal form, and the other factor is a 3 - j symbol made up of the three free
lines. The first factor is independent of magnetic quantum number; the second factor
contains completely the magnetic quantum number dependence.
This graphical reduction process can be continued, but if A and B are connected
by four or more lines, the reduction must cause the appearance of sums of products
of simpler graphs since the maximum number of lines meeting at a node is three. We
illustrate this by reducing a diagram in which A and B are connected by four lines.
We assume once more that A is normal and that all arrows on jl ,j 2 , j 3 , and j4 are
incorporated into B:
4.2 GRAPHICAL METHODS

At this point the power of the graphical method may begin to become apparent. Of
course, to every graphical reduction or manipulation there exists a corresponding
algebraic operation, but the graphical method can have several advantages. First
the notation is more compact because redundant magnetic quantum numbers have
been suppressed. Second, reduction can be made by recognizing geometric patterns.
Finally, the graphical method can serve as a convenient roadmap of what should be
expected in carrying out some involved angular momentum recoupling operation,
and this roadmap is available even if no care is taken to preserve phase factors, that
is, even if no attention is given to the directions of lines containing arrows nor to the
signs of nodes.
166 4 COUPLING OF MORE THAN TWO ANGULAR MOMENTUM VECTORS

We conclude this section with some examples chosen to illustrate the use of the
graphical method. We begin by presenting a diagrammatic proof of Eq. (4.19), the
reduction of a 6 -jsymbol having one vanishing argument:

Here A (jl j4j6) reminds us that j, ,j, ,andj6 must form a triangle if this 6 -jsymbol
is not to vanish. The corresponding situation of a 9 - j symbol with one vanishing
argument [see Eq. (4.24)] is reduced diagrammatically as follows:
4.2 GRAPHICAL METHODS

As a final example we offer a diagrammatic proof of Eq. (4.1l), the orthonormality

property of the 6 - j symbols:
 4 COUPLING OF MORE THAN TWO ANGULAR MOMENTUM VECTORS

The first line of Eq. (4.59) uses (4.43), the second line uses Eqs. (4.44) and (4.45), the
third line uses Eq. (4.54), the fourth line uses Eq. (4.41), the fifth line uses Eqs. (4.31)
and (4.32), the sixth line uses Eq. (4.52), while the final line makes use of Eq. (4.42).
4.2 GRAPHICAL METHODS 169

TABLE 4.1 Algebraic Expressions for Some Commonly Occumng 6 - j Symbolso

170 4 COUPLING OF MORE THAN TWO ANGULAR MOMENTUM VECTORS

TABLE 4.1 (continued)

= (-1)'A;(2jz+ 2)AB(2j3+ 2)
x[2(s-2j2)(s-2j3) -(s+2)(s-2j1 - 1)]
4.2 GRAPHICAL METHODS 171

TABLE 4.1 (continued)

{:I iz
j3-2 j 2j3+ 1
= (-1)'2A2(2j2
1 + 4 ) A 2 ( 2 j 3+ 1)
x [ ( s + l ) ( s - 2 j l ) ( s - 2j2 - 2 ) ( s - 2j2 - l ) ( s - 2 j 2 )
x ( s - 2 j 3 + l ) ( s - - 2 j 3+ 2 ) ( s - 2 j 3 + 3 ) ] +

{:L j 3 - 2
i2
j 2j3+ 2 1
= (-1)'A2(2 j2 + 5 ) A 2 ( 2 j 3+ 1)
xi(s-2j2 - 3)(s-2j2 -2)(s-2j2 - l ) ( s - 2 j 2 )
x ( s - 2 j3 + l ) ( s - 2 j3 + 2 ) ( s - 2 j3 + 3 ) ( s - 2 j 3 + 4 ) ]1

{:I j2
j3-1 j 2j3- 1 1
= ( - 1 ) ' 4 A 2 ( 2 j z + 2 ) A 2 ( 2 j 3+ 2)
x [ ( j l + j 2 ) ( j 1- j 2 + 1 ) - ( j 3 - l ) ( j 3 - j2+ I ) ]
X [ S ( S + 1 ) ( s- 2 j l - l ) ( s- 2j l ) ]1

{$ j3h ;;}
= ( - l ) 5 2 A 2 ( 2 j 2 + 3 ) A 2 ( 2 j 3+ 2)
x [ ( j l + j 2 + l ) ( j l - j 2 ) - j:+ 11
x [ 6 ( s + l ) ( s- 2 j l ) ( s - 2 j 2 ) ( s - 2j3 + I ) ] )

{:I j 2j3+ 1
= (-1)'4A2(2j2
I + 4 ) A 2 ( 2j3 + 2)
x [ ( j l + j2 + 2 ) ( j l - j2 - 1) - ( j 3 - l ) ( j 2 + j3 + 2)l
xL(s-2j2 - 1)(s-2jz)(s--2j3 + 1)(s-2j3 + 2 ) ] )
{:I 2 ;}
= (-l)'2A2(2j2 + 3 ) A 2 ( 2 j 3 + 3 )
x [ 3 C ( C + 1) - 4 j z ( j ~+ l ) j 3 ( j 3 + I ) ]
where
s = jl + j2 + j3
A t ( X ) = [ X ( X - 1)(X -2)(X - 3)]-i
A 2 ( X ) = [ X ( X - 1)(X - 2 ) ( x - 3 ) ( x - 4)]-1
and
C = i l ( j l + 1) - j2(j2 + 1 ) - j3(j3+ 1 )

'Symmetry relations [see Eq. (4.9)] may be used in conjunction with this table to evaluate
all nonvanishing 6 - j symbols with one angularmomentum argument equal to 0, 1 , +,or 2 . i,
172 4 COUPLING OF MORE THAN TWO ANGULAR MOMENTUM VECTORS

NOTES AND REFERENCES

1. G. Racah, Phys. Rev., 62, 438 (1942); 63, 367 (1943). See also L. C. Biedenharn,
J. M. Blatt, and M. E. Rose, Rev. Mod. Phys., 24,249 (1952).
2. E. P. Wigner, "On the Matrices which Reduce the Kronecker Products of Representa-
tions of S. R. Groups," in Quantum Theory of Angular Momentum, L. C. Biedenharn
and H. Van Dam, eds. (Academic Press, New York, 1965),pp. 87-133.
3. M. Rotenberg, R. Bivins, N. Metropolis, and J. K. Wooten, Jr., the 3-3' and 6 -j
symbols (The Technology Press, MIT, Cambridge, MA, 1959). See also "Numerical
Tables for Angular Correlation Computations in a-,/3-, and 7-Spectroscopy: 3-j,
r
6 -j, 9 - j Symbols, F-and -Coefficients." in Landolt-Bornstein Numerical Data and
Functional Relationships in Science and Technology, K.-H.Hellwege, ed. (Springer-
Verlag, Berlin, 1968), Group I, Volume 3.
4. A. R. Edmonds, Angular Momentum in Quantum Mechanics (Princeton University
Press, Princeton, NJ, 1974).
5. See A. P. Yutsis, I. B. Levinson, and V. V. Vanagas, Mathematical Apparatus of the
Theory of Angular Momentum (Academy of Sciences of the Lithuanian SSR, 1960),
translated into English by the Israel Program for Scientific Translations, Jerusalem,
1962.
6. D. M. Brink and G. R. Satchler,Angular Momentum (Clarendon Press, Oxford, 1979),
Chapter VII.
7. J.-N. Massot, E. El-Baz, and J. LaFoucribre, Rev. Mod. Phys., 39,288 (1967); E. El-Baz
and B. Castel, Graphical Methods of Spin Algebras (Marcel Dekker, New York, 1972).
8. J. S. Briggs, Rev. Mod. Phys., 43, 189 (1971).
9. P. G. H. Sandars, "Graphical Methods in Angular Momentum Theory" in Atomic
Physics anddstrophysics, Vol. 1, M. Chitien and E. Lipworth, eds. (Gordon &Breach,
New York, 1971) pp. 171-216.
10. I. Lindgren and J. Momson, Atomic Many-Body Theory (Springer-Verlag,Berlin, 1982).
See especially Chapters 3 and 4.
11. R. L. Chien, J. Chem. Phys., 81,4023 (1984).
12. D. A. Varshalovich, A. N. Moskalev, and V. K. Khersonskii, Quantum Theory ofAngular
Momentum (Nauka, Leningrad, 1975), Chapters 11 and 12, pp. 349404 (in Russian).
13. This convention agrees with that of Lindgren and Momson[lO] but is opposite to that
of Brink and Satchler[6]and of Sandars[9].
PROBLEM SET 2 173

PROBLEM SET 2

1. The vector model allows once more a simple geometric interpretation of the
expression ( j l 2 j3jl j l jUj),, the square of the recoupling coefficient for the two
schemes j = j12 + j3 and j = j1 + j,. According to Eq. (4.3), the probability

represents the probability that a system prepared in a state of the coupling scheme
j = j12 + j3 (having fixed magnitudes of jl ,j, ,j3, j12, and j) will be found
to be in a state of the coupling scheme j = j1 + j, (having fixed magnitudes
of jl ,j2, j3, j23, and j) (see Figure 4.1). In Figure 4.1 it is seen at once that
(j12j3jl jlj23 j) is independent of the value of m (the direction of the z axis).
It is also seen that the six angular momenta jl ,j2, j3, j12,j23, and j may be
regarded as forming the sides of an irregular tetrahedron whose volume V is
given by one-third the area of a face times the slant height; that is,

If j, ,j, ,j3,j,, , and j are fixed, the magnitude of j23 is free to vary in such a
way that the locus of j23describes a circle in space centered about j12. Then the
dihedral angle 4 between the two planes containing (j,j12,j3) and (jl ,j2 ,j12)

FIGURE 4.1 Vector model for the coupling of three angular momenta.
174 4 COUPLING OF MORE THAN TWO ANGULAR MOMENTUM VECTORS

varies between 0 and 2 n at a uniform rate. The unit vectors normal to these two
planesare(j12 x j)/U12 x jl and(j12 x jl)/lj12 x jlI,respectively. Hence

COS 4 = (j12 x j) . (j12 x jd

lj12 x jllh x jll
and

A. Differentiate Eq. (4.62) with respect to time to show that

where d$/dt may be replaced by 2 n. Hint:Make use of the law of cosines

to make the identification

B. The probability density for a given value of lj23 I may then be found from

where the factor 2 arises from the fact that j23 takes on any of its values
twice as 4 goes from 0 to 2n. Hence show that in the classical limit for
large angular momenta the square of the 6 - j symbol has the limiting value

The behavior of the square of the 6 - j symbol for large values of the six argu-
ments is analogous to the semiclassical expression we found previously for
the square of the 3 - j symbol: in classically allowed regions where V > 0 ,
the 6 - j coefficient is rapidly oscillating with an envelope for the oscil-
lations being given by Eq. (4.69), while in classically forbidden regions
where V < 0 the square of the 6 -j symbol is exponentially decreasing. At
PROBLEM SET 2 175

the transition regions corresponding to V = 0,Eq. (4.69) is singular while

the quantum mechanical value for the square of the 6 -j symbol is a finite
fractional number.

2. Prove by diagrammatic means that

which is a variant of Eq. (4.15). Hint: Start by showing that the left-hand side
is given by

Next apply Eq. (4.55) to obtain the final result

+

This identity allows us to remove a mangle from a diagram.

176 4 COUPLING OF MORE THAN TWO ANGULAR MOMENTUM VECTORS

3. Use graphical methods to show that the sum

can be reexpressed as

Hint:Start by showing that the sum over the four 3 - j symbols has the graph
1

1,J' 1u
4. Use graphical methods to prove the Biedenharn-Elliott sum rule for 6 - j syrn-
bols:

This was derived independently by algebraic means by L. C. Biedenham,J.Math.

Phys., 31, 287 (1953) and J. P. Elliott, Proc. Roy. Soc. (London), A218, 345
(1953). This identity may be used to obtain recursion relations for the 6 - j sym-
bols. For example, setting one of the arguments to on the left-hand side of the
Biedenharn-Elliott sum rule causes the sum on the right-hand side
to reduce to two terms.
 5 Chapter

SPHERICAL TENSOR
OPERATORS

5.1 DEFINITION

So far we have concerned ourselves with the transformation properties of the angular
momentum wave functions IJ M) under rotation. We have found that when the
rotation R acts on a particular J JM) it is transformed into a linear combination
of a closed set of functions IJM') M' = -J , -J + 1,. .. , J. Thus the 2 J + 1
functions form a basis for an irreducible representation of the three-dimensional
rotation group of dimension 2 J + 1. In the language of group theory, the elements of
this irreducible representation are the Wigner rotation matrices ( I?). Viewed
in this group theoretic manner, for example, the Clebsch-Gordan series merely
states that the direct product D$~!D$ $, which spans a space ( 2 51 + 1) x
(2 J2 + 1), is a reducible representahon that may be decomposed into irreducible
representations ~ 2 , ~ where J3 runs h m J1 + J2 to I J1 - J 2 1, and moreover,
the coefficients of hireduction are just the Clebsch-Gordan coefficient products
(JIM;, J2MiIJ3Mi) (JiMi, J2M2IJ3M3).
The same idea can be applied to operators Q, which we recall [see Eq. (3.6)]
transform under a rotation of axes as RQR-'. We define a spherical irreducible
tensor operator[l] of rank k to be a set of 2 k + 1 functions T( k, q) with components
q = -k, -k + 1, . .. , k that transform under rotation of the coordinate frame as

that is, under the rotation R the operator T(k, q) is transformed into a closed set of
(3,k + 1) operators T( k , q'), and the coefficients of this expansion are the Wigner
rotation matrix elements D$,*(R). Comparison of Eq. (5.1) with Eq. (3.87) shows
that the tensor operators T( k , q) are proportional to the spherical harmonics Yk,.
178 5 SPHERICAL TENSOR OPERATORS

Let us discover more about the nature of these tensor operators by looking at some
special cases. Let us start with Ic = 0 , for which there is only one spherical irreducible
tensor operator T(0,O). Under rotation it behaves as

in other words, T(0,O) is unchanged by rotation. Hence T(0,O) behaves just as a

number and such a zero-rank tensor operator is called a scalar operator. It might at
first be thought that scalar operators are of trivial importance, but this impression is
not correct. Because the Hamiltonian of a physical system (in the absence of external
fields) is invariant to mtation, this means that all the terms in the Hamiltonian are
scalar operators, and our interest in scalar operators will be intense.
For k = 1 there are three operators T ( 1 , I), T ( 1,O),and T( 1 , - 1) that constitute
a first-rank tensor. Under rotation we have

We recognize that the T ( 1 , q) span the 2 k + 1 = 3 invariant subspace of the three-

dimensional rotation group and hence behave as components of a vector or dipole.
We call such a first-rank tensor a vector or dipole operator. A simple example is the
position operator r with Cartesian components r,, r,, r,. But recall that the Cartesian
components of a vector transform under rotation according to the direction cosine
matrix elements. To ensure (5.3), we need to rewrite these components in a spherical
basis set, that is, in analogy with Eqs. (3.101)-(3.103):

As reference to Eqs. (3.99) and (3.100) shows, the operators defined in Eq. (5.4) are
proportional to the spherical harmonics Y;,($, 4). Indeed, it is readily seen that the
spherical harmonics themselves transform like spherical irreducible tensor operators:

At first, it may seem troubling that the spherical harmonics Ye, can be used as
both functions and operators. However, this same duality applies to the func-
tions x, y, z , which are also used as position operators. As eigenfunctions of t2
and Q,, the Yern($,4) are clearly functions, while as part of the multipole expansion
of an electrostatic potential (see Eq.(1.4)of Application 4), they are certainly operators.
For k = 2 there are five operators, T ( 2 , 2 ) , T ( 2 , 1 ) , T ( 2 , 0 ) , T ( 2 , - I ) ,
T ( 2 , -2), which constitute a second-rank tensor, Again they transform under
5.1 DEFINITION 179

z(, 8,4) ,that is, like the five d orbitals or the quadrupole moments
rotation like the Y
of a charge distribution. Higher-rank tensors with k > 2 appear more seldom and
represent higher-order multipole moments. The definition of the spherical irreducible
tensor operators [Eq. (5. I)] does not exclude k being half-integral. However, if there
were tensor operators of half-integral rank, they could have matrix elements that
connect integral (boson) and half-integral (fermion) angular momentum states. As
long as bosons and fermions are not interconvertible, we need not bother ourselves
about the properties of such tensor operators. Hence we conclude that spherical
irreducible tensor operators are operators that behave like the spherical harmonics
under rotation.
Actually our definition is equivalent to defining the spherical tensor operators by
the commutation relations

and

This connection follows at once when it is realized that the T ( k, q) are proportional to
the Ykqand the proportionality constant is independent of rotation, that is, commutes
with the components of J.
We close this section by considering the always vexing problem of relating
spherical tensor components defined in the space-fixed and molecule-fixed frames.
Let us use the mnemonic that the component p is defined in the space-fixed frame
and the component q in the molecule-fixed frame. Then Eq. (5.1) becomes

which expresses how the components of the tensor T ( k , q) in the molecule-fixed

frame are related to those in the space-fixed frame T ( k, p) under the rotation R. We
derive the inverse of Eq. (5.8) by multiplying both sides by D:', ( R) ,summing over q,
and using the unitarity properties of the rotation matrices [Eqs. (3.77) and (3.78)]. By
dropping the prime on p we find that

Recall that the components of the angular momentum operator when referred to
molecule-fixed axes have anomalous commutation relations, while those referred to
the space-fixed frame have normal commutation relations (see Section 3.4). To apply
the spherical tensor methods to be developed in this chapter it is necessary that all
our tensor operators satisfy Eq. (5.6) in both frames. Some consistent procedure
is mandatory. We choose to adopt a simple solution first introduced by Curl and
Kinsey[2] and later strongly advocated by Brown and Howard[3], namely, to evaluate
180 5 SPHERICAL TENSOR OPERATORS

all spherical tensor matrix elements in the space-fixed frame, using Eq. (5.8) to refer
the molecule-fixed components to the space-fixed axis system. In this way explicit
consideration of the anomalous commutation relations is completely circumvented.

5.2 THE WIGNER-ECKART THEOREM

Probably you are asking at this point why bother to introduce spherical tensor
operators. The reason is that it vastly simplifies the computation of matrix elements
of these operators between states of sharp (definite) angular momentum, say, ( aJ M I
and lalJ'M'), for example. As we will show, such matrix elements factor into the
product of two terms, one of which expresses the geometry, symmetry, and selection
rules of the system, the other of which contains the dynamics. This disentanglement
of geometrical from dynamical behavior is the power and the glory of the celebrated
Wigner-Eckart theorem[4].
Consider the matrix element ( aJ M I T ( k, q) la1J'M'). The state vector
T( k, q) la1J'M')transforms under rotation according to the representation Dk@ID',
which spans a ( 2 k + 1)(2 J' + 1) space of the rotation group. We decompose
this into its irreducible representations. Specifically, under rotation the product
T( k, q) la' JIM') transforms as

Equation (5.10) shows that the ( 2 k + 1)(2 J' + 1) products T(k,q) la1J'M')
transform under rotation as the direct product representation Dk@ D ~ ' .
Recall the Clebsch-Gordan series in which the product of two rotation matrices is
a sum over some Clebsch-Gordan coefficients times a rotation matrix DK where K
ranges from k + J' to Ik - J'I. We are led to ask what linear combination of
the products T ( k , q) Ja'J'M') transforms as a particular state vector IPKQ) of
the 2 K + 1 functions forming a basis for the representation DK. The answer to this
question is that this problem is just what Clebsch-Gordan coefficients and angular
momentum coupling were designed to solve

IPKQ) = (kq', J'M1'IKQ)T ( k , q') I ~ J ' M " ) (5.11)

,&'
We form the matrix element with ( aJ M I :
5.2 THE WIGNER-ECKART THEOREM 181

By multiplying both sides by (kg, J'M'IKQ) and summing over K and Q, we find

But the scalar product (aJMIPKQ) vanishes unless J = K and M = Q. When

these conditions are satisfied, the value of the scalar product is independent of M,
that is, unaltered by rotation. We therefore can write Eq. (5.13) as

where K is replaced by J. Here (aJIITk IlalJ'), which equals (aJMIPJM), is

called the reduced matrix element of Tk. This result shows that the matrix elements
of T ( k, q) separate into the physical part of the problem embodied in the reduced
matrix element and the geometric part embodied in the Clebsch-Gordan coefficient.
If only angular momentum theory spoke to us with one tongue.. .but, alas, at
this crucial juncture there is another of those damnable convention problems. Some
authors, including Brink and Satchler[5], adopt the above, while others, such as,
Edrnonds[6], redefine the reduced matrix element to be a factor of ( 2 J + 1) larger!
We adopt the convention (Edmonds[6]) and state the Wigner-Eckart theorem as

where we remind ourselves of the(-1)"j-j'(2 j+l)112factor by writing (a1j'II T~1 Iaj)

in place of (a'jlll Tk 114). By knowing the value of the reduced matrix element
(crlj'II T~Ilaj), we can evaluate all the ( 2 j l + 1)( 2 k + 1)( 2 j + 1) different possible
matrix elements (a'j'm'l T ( k, q) Iajm) !
Some simple examples may serve as a guide to how to find reduced matrix
elements. A particularly easy matrix element is calculated both directly and by
the Wigner-Eckart theorem. Comparison of the two results yields the reduced
matrix element being sought. As an example consider the orbital and spin angular
momenta L and S . From their commutation rules it is readily verified that L and S
are first-rank spherical tensor operators, that is, satisfy Eq. (5.6) with k = 1. The
three components of L and S are
182 5 SPHERICAL TENSOR OPERATORS

and

The reduced matrix elements of L and S are easily found. For example,
(L'M'I L, ILM) is nonzero only for L' = L and M' = M , in which case

(LMI L, ILM) = M

By the Wigner-Eckart theorem

so that Eq. (5.18) becomes

Comparison of Eqs. (5.17) and (5.20) yields the result

( ~ ' 1 1 L'" llL) = [ ( 2L + 1)L ( L + l)]*s,,L

Similarly

( ~ ' 1 1 s") 11s)= [ ( 2 S + l ) S ( S + 1 ) 1 * 6 ~ , ~

which for the case S = i becomes

A trivial example of the use of these reduced matrix elements is the evaluation of

(331L+132)=(33)LZ+iL,132)=[3(3+1)-(2)(3)]*=6+ (5.24)
5.2 THE WIGNER-ECKART THEOREM

This may also be evaluated according to the Wigner-Eckart theorem by

Of course, no one would normally use the Wigner-Eckart theorem to evaluate the
matrix elements of L+, L-, or L,, but it certainly is reassuring to know that all the
pieces of this machinery do work!
Another example of practical importance is the reduced matrix element of a
spherical harmonic. We have previously seen that for the integral of the triple product
of spherical harmonics with ml = m2 = m3 = 0

Alternatively, application of the Wigner-Eckart theorem yields

Hence

(5.28)
In practice it is often more convenient to work with a modified spherical har-
monic C( k, q) ,defined by
184 5 SPHERICAL TENSOR OPERATORS

which differs from the spherical harmonic Ykq

by the use of a different normalization.
Then we have

We close this section by examining some consequences of the Wigner-Eckart

theorem applied to electric and magnetic moments. We begin by looking for what
conditions must be met for the system to possess a static moment. Let the system
be in a state of definite angular momentum so that its state vector is Iajm). We
next show that such a system cannot have a permanent dipole moment unless
j 2 or a permanent quadrupole moment unless j 2 1. Let p denote the
dipole moment operator and Q the quadrupole moment operator. For p and Q
to be "permanent" means that the expectation value of p and Q must be nonzero,
that is, ( a j m l p ( 1 , q ) lajm) # 0 and (cujml Q(2,q) Iajm) # 0 for at least one
component q. Recognizing that p and Q are spherical tensors of ranks k = 1 and
k = 2 , respectively, we apply the Wigner-Eckart theorem. For example,

which vanishes if j = 0 or if q # 0 . Hence only systems with j 2 have permanent

dipole moments, and for such systems the dipole moment points along the axis of
quantization. A similar argument pertains to the quadrupole moment operator, that is,
only states with j 2 1 can have a permanent quadrupole moment, and such systems
possess only one such moment, the expectation value of Q(2,O).
So far nothing has been said about the difference between electric and magnetic
moments. However, these operators behave differently under inversion of the spatial
coordinates. The electric dipole operator can be written as p(e)= xi eiri, where e i
is the charge of the ith particle with position vector ri. However, the magnetic
dipole operator looks like p(rn)= pogLL+ pogsS + pIgII + .. where po is the
a,

electronic Bohr magneton, ,UI is the nuclear Bohr magneton, and g ~,gs, , g~ are the
corresponding g factors for each type of angular momentum (current loop) possessed
by the system. Both p(=)and p(rn)are first-rank tensors (i.e., vectors), but p(rn)
is an axial vector (pseudovector) that does not change sign under inversion of all
spatial coordinates, while p(e)is a polar vector and does. We say that p(e)has
odd parity while p(rn)has even parity. For the matrix element of p(e)or p(m)to
be nonvanishing, its overall parity must be even. Therefore, in a state of definite
angular momentum and definite parity there can be no permanent electric dipole
moment, while there can be a permanent magnetic dipole moment provided j 2 i.
Hence nondegenerate states (excluding the ( 2 j + 1)-fold m degeneracy present in the
absence of external fields) cannot have permanent electric dipole moments and cannot
exhibit first-order Stark effects. In contrast, nondegenerate states generally have
permanent magnetic moments and show linear Zeeman effects, unless j = 0 . The
parity of the 2 multipole moments go as ( - 1) for electric moments and ( - 1)k-l
5.2 THE WIGNER-ECKART THEOREM 185

for magnetic moments so that there are in general only even electric multipole
moments and odd magnetic multipole moments. Thus we understand why nuclei
6
with I 2 have magnetic dipole moments while those with I 2 1 have electric
quadrupole moments. Moreover, ground-state nuclei do not have electric dipole
moments (neglecting weak forces, parity violation, and all that jazz).
So far we have focused our attention on the expectation values (diagonal matrix
elements) of the multipole moment operators. But the off-diagonal matrix elements
are also of interest, of course, particularly with regard to radiative transitions. Here
we see that electric dipole transitions can connect states of only opposite parity while
magnetic dipole and electric quadrupole transitions can connect states of only the
same parity. Moreover, the conditions for the nonvanishing of the 3 - j symbol

give us at once the familiar selection rules: namely, for electric dipole transitions
as well as magnetic dipole transitions A J = 0 , f1 and A M = 0 , f1, except that
J = 0 cannot be connected to J = 0 , and for electric quadrupole transitions (as
well as magnetic quadrupole transitions) A J = 0 , f1 , f2 and AM = 0 , f1 , f2 ,
except that transitions for which J + J' < 2 cannot occur. Whereas q must be zero
for static moments of systems in which J z is diagonal, in general three transition
dipole moment operators and five transition quadrupole moment operators must be
considered in working out possible radiative transitions. This example illustrates how
the Wigner-Eckart theorem allows at once a separation and identification of all the
geometric factors, in this case the common selection rules for radiative transitions.
In many problems we encounter the square modulus of matrix elements. This
motivates us to consider the Hermitian conjugate of the matrix elements of a spherical
tensor operator. We begin by taking the Hermitian conjugate of both sides of
Eq. (5.14):

Because we insist that T ( k, q) behave as a spherical harmonic of rank k and compo-

nent q, we define the adjoint of a spherical tensor operator T ( k, q) as

inanalogy w i t h ~ $ 0 , 4 ) = Y&(B,c#J)= (-1)-qYk-q(0,4). Then

 5 SPHERICAL TENSOR OPERATORS

Comparison of Eq. (5.32) with Eq. (5.34) yields the useful identity

5.3 SPHERICAL TENSOR PRODUCTS

As we have seen, the spherical tensors behave like spherical harmonics. Recalling the
transformation between coupled and uncoupled representations [Eq. (2.5)], it is not
surprising to find that the multiplication (also called the contraction) of two spherical
tensors T ( L l , q l ) and T(L2,9 2 ) of ranks Ll and LZ to form a spherical tensor of rank t
is given by

To establish that T ( l ,q) is a tensor of rank L we must show that the right-hand side
of Eq. (5.36) transforms under rotation according to

We apply the rotation R to both sides of (5.36):

5.3 SPHERICAL TENSOR PRODUCTS 187

where we have made use of the Clebsch-Gordan series [Eq. (3.105)] and the
orthonormality of the Clebsch-Gordan coefficients [Eq. (2.8)]. Comparison of
Eqs. (5.38) and (5.37) shows the validity of the identification made in Eq. (5.36).
The tensor product of T(El, ql) and T(E2, 92) may also be written as

This notation reminds us that the direct product of the sets of operators T(El, ql)
and T(E2,42) span the representation @' @ @, which can be decomposed into the
representation @1+e2 + D@ez-l + . . . + ~ l ~ ~ The - ~ ~particular
l . linear combination
of the products T ( & ,ql)T(Ez, 92) that transform like T(E, q) are given by the
Clebsch-Gordan coefficients (El ql ,E2 92 leg) according to Eq. (5.36). This result
shows that the coupling of spherical tensors is mathematically the same as coupling
angular momentum eigenvectors, for both use only the group theoretic properties of
the operators or states under rotation.
To illustrate Eq. (5.36) consider, for example, the compound irreducible tensor op-
erators formed by multiplying (contracting) two first-rank tensors (vectors) A and B,
which have the components of the form given in Eq. (5.4). By evaluating the expres-
sion

we find explicitly that for k = 0

for k = 1
188 5 SPHERICAL TENSOR OPERATORS

and for k = 2

[A( 8 B")] = T&[(A, B, + A, B,) I i( A, B, + A, B,)]

It is seen that the tensor product of rank k = 0 is related to the scalar product (dot
product) of A and B while the tensor product of rank k = 1 is related to the cross
product of A and B. Indeed, the latter is just ( A x B) .@,,where $1 = -2 - * ( a + i*),
6o = 2, and -*
= 2 ( 2 - i?) are unit vectors in the spherical basis that transform
as D' under rotation. Care must be taken here to use the definition for a scalar product
of vectors having complex components, that is,

from which it is seen that 6, . = 6,,1.

@,I

A particularly importantcase of (5.36) is that of compound tensor scalar operators:

[T(') 8 T(~')]: = (kq, k'q1100) T ( k, q)T( k', q')

424

Equation (5.46) shows that scalar operators can be built from tensor operators of
rank k by contracting two tensor operators of the same rank. Equation (5.41),
in which the scalar operator [A' 8 B']iO) was constructed from the two vector
operators A and B, is a special example of Eq. (5.46). It is traditional to define the
generalized dot product . B ( ~by)

Then . B( is also a scalar operator and is related to [A") 4 B(L)] by

~ ( ~ ) . ~ ( ~ ) = ( - 1 ) ~ ( 2 k + l ) (k) lo
* [(0)~ ( ~ ) @ B(5.48)

Both Ack) . B ( ~and

) 8 B(~)],$O) appear in the literature, so care is needed not
to confuse these two forms of the scalar contraction.
5.3 SPHERICAL TENSOR PRODUCTS 189

With Eqs. (5.40)-(5.44) it is possible to rewrite products of vector operators in an

alternative form by use of this angular momentum coupling algebra. As an example
(which will prove useful later in discussing electric dipole absorption and/or emission
of radiation) we consider the scalar operator ( a . b) ( c - d). According to Eq. (5.41),
this has the tensorial character

In many applications it is necessary to regroup the operators in such an expression so

that, for example, a(') and d l ) are coupled into one resultant and b(') and d(') into
another. This is accomplished by using the recoupling transformation of four angular
momenta [Eq. (4.20)], with jl = j2 = j3 = j4 = 1. Explicitly we have

Evaluation of the recoupling coefficient (( 11)k( 11)kO I( l1)0( 11)00) involves a

9 - j coefficient, three of whose arguments are zero. Application of Eq. (4.25)
followed by (4.19) yields the result that

where k is restricted to have the values of O,1, or 2. Hence

where Eq. (5.46) has been used to obtain the last line of the above. Equation (5.52)
has a very pleasing appearance in that it clearly displays the scalar contraction of the
spherical operators T(k, q) formed from the vector operators a, b, c, and d. This
vector identity may also be written as

This expression cannot be simplified further without writing out the actual vector
components.
190 5 SPHERICAL TENSOR OPERATORS

As remarked before, the Hamiltonian of the system must be invariant to coordinate

rotation (the energy of the system must be independent of the choice of coordinates)
and hence is a scalar operator. As such, it is expressible as a sum of scalars, each
of which is a contraction of spherical tensors of the same rank. It is no wonder,
then, that the interaction between a system of charged particles (atom, molecule)
and an external electric field (from a radiation source, from a neighboring charge
distribution, or from an applied external field) contains terns such as the dipole
moment of the system times the electric field, the quadrupole moment of the system
times the gradient of the electric field, the octupole moment of the charge distribution
times the gradient of the gradient of the electric field, and so on.
To conclude this section let us examine the scalar invariant that can be formed
from the tensor product Ckq( O,+) and Ckq( 9' , 4') of modified spherical harmonics.
Using Eq.(5.47), we write

Because ck. ckmust be unchanged by rotation, it follows that this scalar product
must be some function of the angle O included between the directions (O,C$) and
( O', #), as this angle is the only quantity that does not depend on the choice of axes.
In particular, let us select our axes so that (O', 4') becomes ( 0 , O), that is, so that
the ( 8', 4') direction coincides with the new Z axis. Then Ck-J01,4') becomes
Ck-q(O, 0) = SqO[see Eq. (1.55)] and the other factor in (5.54) Ckq(O,4) becomes
CkO(O,@)= Pk(~~sO),wherewerecognizethat(0,q5) isrotatedto(O,Q,) when
(O', 4') is rotated to (0,O). Hence

This result is the same as Eq. (3.85); that is, we have rediscovered the spherical
harmonic addition theorem, which we now recognize as simply the scalar contraction
of the spherical harmonics.
Spherical harmonics with different arguments may also be used to construct other
tensors in addition to the scalar tensor contraction. These tensors, called bipolar
harmonics, are defined by[3]
5.3 SPHERICAL TENSOR PRODUCTS

They satisfy the orthogonality property

and the closure condition

For K = Q = 0 , Equation (5.56) becomes simply another statement of the spherical

harmonic addition theorem.
The bipolar harmonics often appear in problems involving two directions (corre-
lation and coincidence measurements)[7-lo]. A simple example is the ejection of
two electrons from an atom or molecule, caused by some collision process or by
the absorption of radiation. The two-electron angular distribution may be expressed
ad111

where the angular dependence is contained solely in the bipolar harmonics and
the ALM(ll,12) express the dynamics of the electron photoejection process. In
Eq. (5.59) the sum is over the possible angular momenta el and f22 of electrons 1
and 2 , which form the resultant total angular momentum L with projection M on
the axis of quantization. The dynamical coefficientsALMmay depend on scalars
(energy or time of emission) of the two electrons considered separately but not on the
angle between their momenta. Other examples of two-vector correlations are readily
formulated in reference to photofmgment rotational and translational motions[l2] or
to the prototypical bimolecular reaction A + B C -+ AB + C. In the latter case one
may consider measuring the correlation between the initial relative velocity vector
of the reagents and the recoil direction of one of the reaction products AB or C
(the differential cross section), the correlation between the direction of the initial
relative velocity of the reagents and the direction of the B C angular momentum on
the reaction probability (the total cross-section anisotropy), the correlation between
the initial relative velocity vector and the angular momentum direction of the AB
product (the product polarization), or the correlation between the direction of the
angular momentum of the B C reagent and the direction of the angular momentum of
the AB product (the rotational tilt)[l3]. Correlations between three or more vectors
are also of interest and lead naturally to a generalization of Eq. (5.56) to n-polar
harmonics[7,8, 10, 131.
192 5 SPHERICAL TENSOR OPERATORS

5.4 TENSOR PRODUCT MATRIX ELEMENTS

In this section we take full advantage of the entire angular momentum coupling
machinery to derive general expressions for the matrix elements of spherical tensors
that are themselves products of spherical tensor operators. These expressions form
an important and powerful reference library in the application of angular momentum
coupling techniques to problems of interest, particularly those involving composite
systems. Such a composite system may consist of two or more separate systems or
separate spaces of the same system. We consider the most general case and then
specialize our results.
Let V m ) = ljl j2 j m ) denote a coupled state built from two component states
V l ml ) and 1 2 m2) in spaces 1 and 2 according to the tried-and-true recipe

or lovingly recast into a form with 3 -j symbols as

(5.61)
We also introduce spherical tensor operators T ( kl ,ql ) and U( k 2 , q z ) acting on
independent sets of variables in spaces 1 and 2 and hence commuting. We form the
composite spherical tensor X ( k,q ) built from T ( kl ,ql ) and U( k2 , q z ) according
to

which has as its 3 - j counterpart

(5.63)
We desire to evaluate the general matrix element

where we have applied the Wiper-Eckart theorem to obtain the right-hand side.
Here 7 and 7' represent all other quantum numbers of the states characterized by j
and j'.
5.4 TENSOR PRODUCT MATRIX ELEMENTS

We use Eqs. (5.61) and (5.63) to rewrite Eq. (5.64) as

Finally, we apply the Wigner-Eckart theorem once more and rewrite (5.65) as

We have two expressions [Eqs. (5.64) and (5.66)] for the general matrix elements
of X ( k, q ) . We equate them and multiply both sides by
( 1 ) (-mj q m'
194 5 SPHERICAL TENSOR OPERATORS

and sum over m, m', and q. With the help of Eq. (2.32) we obtain

We have a sum over six 3 - j symbols, which has the distinct smell of a 9 - j symbol.
Using either graphical methods[l5] or the symmetries of the 3 - j symbols to remove
the phase factor, we obtain the result

Everything now becomes a special case of Eq. (5.68)!

In particular, let us examine the matrix elements of the zero-rank compound tensor
X (0,O). Putting k = 0 into the 9 - j symbol of Eq. (5.68) causes it to vanish unless
j = j' and kl = k2. We conclude that a scalar operator (as one would find as terms in
the Harniltonian of a system) is diagonal in the total angular momentum. Moreover,
such scalar operators can be constructed only from the contraction of tensor operators
of the same rank. Similar conclusions have been reached previously. However, the
advantage of Eq. (5.68) is that when k = 0 , it reduces to a 6 -j symbol, giving us an
explicit expression for the reduced matrix element of a compound scalar operator:
5.4 TENSOR PRODUCT MATRIX ELEMENTS 195

Often, as we remarked before, T~ .ukis used in place of [ T k8 uk]AO)

as the scalar
operator. Using Eq. (5.48),we rewrite Eq. (5.69) as

Because of frequent occurrence of these scalar operators, we write out Eq. (5.64)in
full for this case:

We have placed a box around Eq. (5.7 1 ) because it is so useful.

Next, we examine the special case where X ( k, q) acts on only the variables in
space 1 or 2. For example, suppose that X ( k ,q) behaves as only T ( kl , q l ) . This
is the same as setting U(k 2 ,qz) to a constant times the identity operator, that is,
to U ( 0 , O ) .By putting kz = 0 in Eq. (5.68), we find that once again the 9 - j symbol
collapses to a 6 - j symbol and we must have j2 = ji and k = kl . We find

Similarly, if X ( k , q) acts only on space 2 , that is, if X ( k, q) = U ( k2,9 2 ) ,then

Equations (5.72) and (5.73)are also placed in boxes because of the heavy use they get
as they give us the reduced matrix elements of a single particle operator in a coupled
196 5 SPHERICAL TENSOR OPERATORS

representation. Note that the phase factors are different in these two equations
because of the order of coupling of j , and j2 to form j.
Finally, there is a more general case, which is, however, not encountered as
frequently as the preceding cases. This is the situation when the system is in
a sharp angular momentum state Iqjrn) that is not decomposable into subsystem
states and we desire to find the matrix elements of the compound tensor X ( k, q) =
[Tkl@ T k z ] i kwhere
), here X ( k , q) can act only on the set of variables of the system.
The same treatment as that used in deriving Eq. (5.68) yields the result

We conclude this section by working out some examples chosen to illustrate the
utility of this reference library [Eqs. (5.71)-(5.74)]that we have compiled. First, let
us consider the effect of spin-orbit interaction on the term energies of a two-electron
atom. Here electrons 1 and 2 have orbital angular momenta el and l2 and spin
angular momenta sl and s2. According to Russell-Saunders coupling, L = el + & ,
S = sl + s2, and J = L + S. The spin-orbit Hamiltonian is a sum of one-electron
operators of the form Hso = C:=,E(ri)ti . si, and for the particular problem under
consideration

The general matrix elements of Hso are given by applying Eq. (5.71):

where the radial integral over [ ( r ) has been abbreviated as

5.4 TENSOR PRODUCT MATRIX ELEMENTS 197

Next we apply Eqs. (5.72) and (5.73) to express the reduced matrix elements appear-
ing in Eq. (5.76) in simpler form:

The reduced matrix elements in Eq. (5.78) have been evaluated before [see Eqs. (5.22)
and (5.23)l. Making all these substitutions, we find
 5 SPHERICAL TENSOR OPERATORS

L
x{S'
S
L'
J
I){?'
L
e,
L
;){B f}
where we have used the fact that s, ,s,, s',,and $ all equal i.
Examination of Eq. (5.79) permits us to make a number of remarks regarding
the nature of spin-orbit interaction. We see that Hso can connect states of only the
same total angular momentum J and projection of the total angular momentum M j .
Moreover, Hso is independent of M j and cannot split the ( 2 J + 1)-fold spatial
degeneracy associated with each J level. The spin-orbit operator is a one-electron
operator and cannot connect the terms arising from configurations that differ by two
spin-orbitals. This has the simple consequence that for the Cnlel;4e, term all quantum
numbers associated with electron 2 must be identical while for the Cl44e2 term,
the same condition holds for electron 1. Finally, the leading 6 - j symbol of each
term vanishes unless the following two angular momentum addition equalities hold:
L + 1= L' and S + 1= S'; in other words, A ( L I L') and A ( S 18') must be satisfied.
This yields the selection rules for spin-orbit interaction:

These restrictions can be augmented further by considering parity. Since Hso is

invariant under spatial coordinate inversion, the spin-orbit operator can connect only
levels of the same parity. Let us go one step further. Consider the form of the Cn,e,;nie;
terms in Eq. (5.79) when .ti= L: = 0 , where i = 1 or 2. The L:l2 factor shows at
once that these terms vanish. Hence ns spin-orbitals cannot contribute to the spin-
orbit splitting; that is, there must be a nonzero orbital angular momentum (current
loop) for the electron to have a relativistic (magnetic) interaction with its own spin.
Viewed from the rest frame of the electron, the current loop results from the nucleus
orbiting the electron with an effective charge which depends on the distance between
the electron and the nucleus. Hence, the magnitude of the current loop increases
with Z for the same 4 value and is greater for L # 0 electrons which penetrate the
core.
In general one must diagonalize a secular matrix with the same J value to obtain
the different term energies E ( ~ ~ + j)' L. However, when the spin-orbit interaction
is small compared to the separation between different terms having the same J ,
we may include the effects of spin-orbit coupling by using first-order perturbation
5.4 TENSOR PRODUCT MATRIX ELEMENTS 199

theory. For the diagonal term (LSJ M j ( Hso (LSJMJ)

the leading 6 - j symbol of
each Sne = Sne;& term becomes

Furthermore, for a given 2 S + 1 ~ level the values of S and L are fixed as well as those
of sl , s2, el, and &. Hence the diagonal term depends only on J through the 6 - j
symbol given in Eq. (5.80). Thus we can write

where the spin-orbit constant A is the same for all J levels of a 2 S + 1 ~term. This
treatment yields the Land6 interval rule[l4]

that is, the separation is proportional to J.

It is worth remarking that when S = 0 (singlet term) or L = 0 (S term), there is
only one J level associated with the 2 S + 1 term.
~ Spin-orbit interaction cannot split
this level, but its position relative to the other levels can be shifted by the existence of
nonzero off-diagonal matrix elements of Hso connecting this level with other levels
of the same J .
Please also note that the replacement of Hso by AL - S is valid only for calculating
the diagonal matrix elements of Hso. For example, L . S cannot connect states
of different multiplicity; that is, singlet and triplet states are mixed by spin-orbit
interaction but not by AL - S.
For the special case of the ( d ) 2 configuration having two equivalent electrons,
the diagonal elements have the form

L S
x{s L
J
I}{: : ;}{! ;!}
We see that only one spin-orbit parameter id characterizes all the J-level splittings
of the levels arising within a 2 S + 1 ~term. This conclusion is also true for the
configuration ( d ) ' of p identical electrons.
200 5 SPHERICAL TENSOR OPERATORS

As a final example, we consider the following resonance fluorescence process:

Here radiation of polarization 6, is absorbed by the initial states {li)) of an atomic

or molecular system causing an electric dipole transition to the excited states {le)),
which then decay to the final states {If )) by an electric dipole transition in which
the reemitted radiation is detected with polarization e d . We wish to develop a
general expression for an arbitrary excitation-detection geometry and for arbitrary
polarizations $ and Sd. Then we will specialize this to the common situation where
the observation direction is at right angles to the excitation direction and the incident
and detected light are both linearly polarized.
Clearly, the resonance fluorescence process is an involved one and several simpli-
fying assumptions are required to make this problem tractable. Let all states be char-
acterized by sharp angular momentum quantum numbers. If we assume no hyperfine
structure, we may write li) = laiJiMi), le) = \aeJeMe),and If) = lafJfMf). We
imagine at time t = 0 a short pulse of light is incident on our sample which is at
sufficiently low pressure so that we can treat it as isolated, that is, that we can ignore
the effect of collisions. We can derive an expression for the intensity at subsequent
time t, denoted by I ( t ) , assuming that the following conditions are met:

1. The spectral width of the light pulse is much greater than the reciprocal of the
duration of the pulse. Then the resonance fluorescence process may be treated
as two independent steps, absorption followed by fluorescence.
2. The excitation process is sufficiently weak; thus the natural lifetime for the
excited system to decay is much less than the average time between two
successive photon absorptions by the system. This condition ensures that we
can neglect optical pumping effects.
3. The duration of the pulse is short compared to characteristic "precession"
times of the system, that is, compare to the reciprocal of Bohr frequencies
corresponding to excited-state energy differences.
4. The initial state is isotropic, that is, has not been prepared with a preferential
population in any of the M isublevels nor with special phase relations (coher-
ence) among these Zeeman components.
5. The radiative lifetimes of all excited-state sublevels are the same. In what
follows it will be denoted by re = T
;
'
.
5.4 TENSOR PRODUCT MATRIX ELEMENTS 201

Before the light pulse arrives, the wave function of the atom or molecule in some
particular initial state ai Ji Mi with energy E ( ai Ji Mi ) is

where wui = E ( a i Ji M i ) / A . At time t = 0 the light pulse with polarization C,

causes excitation of the system. The wave function of the excited state is proportional
to

where r is the electric dipole moment operator and C, is a unit vector having in general
complex components that specify the polarization of the electric field of the incident
radiation. The summation over Me takes into account that more than one excited-
state sublevel can be excited from a given ground state sublevel M i , depending on
the nature of the polarization of the incident light beam.
The excited state evolves in time according to

l $ ( a e J e ; t ) ) cx C ( a e J e . ~ e ~ & . r~l ~i i ~~ ai ~) J ~ ~ e[ -) i ew ~x , pt - r e t / 2 ]
Me

(5.86)
where W & = E( aeJe M e )/ti and spontaneous emission is taken into account by
inclusion of the damping factor exp[ - r e t / 2 ] , where re= 11% is the rate of natural
radiative decay. The factor of $ arises because the excited-state population, which is
proportional to the square of the excited-state wave function, decays at the rate re.
Suppose at time t that the excited state emits a photon of polarization Gd as it makes
a spontaneous transition to the final state af Jf M f . Then the instantaneous rate of
fluorescence with that polarization is proportional to

where the sum is over all possible final states to which the excited state might radiate.
In the foregoing we restricted ourselves to excitation of one particular sublevel of
the initial state. However, for an isotropic distribution of initial states we may include
the contribution of all of them by simply rewriting Eq. (5.87) as

Next we make use of the identity

 5 SPHERICAL TENSOR OPERATORS

where 0 is an operator and Ot is its adjoint. Substitution of Eq. (5.86) in (5.88) and
use of Eq. (5.89) yields the result

x exp [-i (wMe - wM;) t - Tet] (5.90)

This displays the expected (see Application 9) time behavior of an exponential decay
on which is superimposed quantum beats from the interference of the indistinguish-
able different paths through the sublevels of the excited state that connect the same
initial and final sublevels[l6]. The total rate is found by integrating Eq. (5.90) over
all time:

In terns of quantum electrodynamics[l7], C .r corresponds to the photon annihilation

operator and C* .r to the photon creation operator, that is, C* .ralways appears in front
of laeJeMe) or laeJ, ML), while Carappears in front of laiJiMi) or lafJfMf ). Note
that the denominator of Eq. (5.91) depends on the Zeeman sublevel energy splitting
of the excited state and the natural radiative lifetime. Equation (5.91) was derived by
Breit in 1933 and is often referred to as the Breit level-crossingformula[18].
Suppose that an external field (magnetic or electric) is applied along the axis of
quantization (the Z axis) of the system. Then the splittings between the magnetic
sublevels can be made to change. There are two limiting cases of much interest.
If the excited-state magnetic sublevels are split apart much larger than the natural
linewidth, that is, if (wMe- wM,) >> r e , all terms involving Me # Ml vanish and
the resonance fluorescence signil is proportional to the simple expression
5.4 TENSOR PRODUCT MATRIX ELEMENTS 203

where each level a, Je Me contributes independently (i.e., incoherently). The other

extreme is represented by ( w M e- wML).<< refor example, in zero external field.
Then the resonance fluorescence signal 1s proportional to

which has interference terms (cross terms) among the different Me, ML sublevels
provided that they are coherently excited in the resonance fluorescence process (zero-
field level crossing). Hence as we pass from low to high external field, this coherence
is destroyed and a study of the polarization intensity as a function of field strength
allows us to determine excited-state splittings with a resolution comparable to the
natural linewidth.
If desired, it is actually possible to achieve resolution within the natural width!
This is accomplished by biasing the detection of the fluorescence to favor those
excited species that have lived longer than the radiative lifetime, for example, by
having the detection system accept signal only after a delay which exceeds the
radiative lifetime[l9]. The narrowing of the "resonances" below the natural width
does not violate the uncertainty principle because the uncertainty principle only
expresses the Fourier transform pair between mean decay time and spectral linewidth.
However, the gain in resolution must in practice be balanced against the loss in signal-
to-noise ratio of the fluorescence, caused by the resulting decrease in the number of
fluorescence events detected at longer delays.
While there are numerous uses of this coherence sub-Doppler spectroscopy, we
turn our attention to the calculation of the degree of polarization in the absence of
external fields. Equation (5.93) is the starting point. It may be rewritten as

where the absorption and fluorescence matrix elements are

204 5 SPHERICAL TENSOR OPERATORS

and

respectively, and we have introduced the projection operators

and

The projection operators P are readily verified to be invariant under rotation, that
is, R-'PR = P, and we can ignore these scalars in our tensorial analysis. Hence we
are concerned with operators of the form (6 .r) (C* - r ) in Eqs. (5.95) and (5.96). This
has the tensor structure indicated in Eq. (5.52):

which conveniently separates into the scalar contraction of two tensors, the polariza-
tion tensor[20,21]

which contains only variables concerning the electric field direction and the tensor
ik',
[ d l ) x r(l)] which contains operators that act on the atomic or molecular system.
The nine components of the polarization tensor ~ , k ( @ 6*), for k = 0, 1, and 2 are
readily worked out using Eqs. (5.41)-(5.44).
Let us illustrate the evaluation of the absorption matrix element AMeMt in rather
complete detail. Using Eqs. (5.99) and (5.100) and applying the Wigner-Eckart
theorem, we have
5.4 TENSOR PRODUCT MATRIX ELEMENTS

as
Next we use our trusty library of results, namely, Eq. (5.74), to express AMeM;

where the summation over j" in Eq. (5.74) is restricted to jl1 = Ji because of the
presence of the projection operator P, , and where we have also used Eq. (5.35) to
express the result as involving the square of the reduced matrix element of the electric
dipole operator.
By similar means,

Then Eq. (5.94) contains

206 5 SPHERICAL TENSOR OPERATORS

where we have used Eq. (2.32) to simplify its form. Hence

where

The simple form of Eqs. (5.105) and (5.106) displayed above is a lovely example
of the power of the tensor method. It is a key result[l7,22, 231 in calculating the
resonance fluorescence intensity under general excitation-detection conditions.
Let us specialize to the case of linearly polarized excitation-detection, where 6,
and ed are unit vectors pointing along the polarization directions of the exciting
and detected beams, taken to be the quantization axes of the tensors E,k(ea,C;;l) and
E,k(C,, CT, ) , respectively. Then the only nonvanishing components of both tensors
are those for q = 0 . It follows from Eq. (5.100) that we can write

However, the quantization axes of the two polarization tensors differ in general
because 6, does not necessarily coincide with but instead makes the angle 8.
We refer the detection polarization tensor to the same axis as that of the absorption
polarization tensor by carrying out the rotation R = ( 0 , 8 ,0 ),that is,

Thus for a linearly polarized excitation-detection geometry

I l k
k
5.4 TENSOR PRODUCT MATRIX ELEMENTS

where the summation over k is restricted to three terms, k = 0, k = 1, and k = 2.

Equation (5.109) shows that the angular dependence of the resonance fluorescence
has a very simple form: the k = 0 term is isotropic, the k = 1 term vanishes because

is zero, and the k = 2 term varies as ( 3 cos2 8- 1)/2. Thus by choosing 8 so that the
k = 2 also vanishes, that is, by choosing 8 = arccos( 1/3) 'I2 54.7 ",the radiation
reaching the detector is independent of the linear polarization directions 8, and Cd.
This choice of 8 is called the "magic angle."
Now consider the traditional case of a right-angle excitation-detection geometry
in which the viewing direction is perpendicular to the direction of propagation of
the linearly polarized excitation beam. It is convenient to distinguish two cases.
In the first, C, is parallel to Cd, 8 = 0'. and the resonance fluorescence signal is
denoted by Ill; in the second, 6, is perpendicular to 6d, 0 = n/2, and the resonance
fluorescence signal is denoted by IL. Then one describes the polarization by the
degree of polarization

or by the polarization anisotropy

These two measures are related by

and

Using Eq. (5.109), we obtain for the orthogonal excitation-detection geometry ( 8 =

n/2) the particularly pleasing expression
 5 SPHERICAL TENSOR OPERATORS

TABLE 5.1 Resonance Fluorescence Process; Linear Polarization

Transition R( Ji P ( Ji)

This is readily evaluated for the nine possible branches PT, QT, or RT on excitation
and P1, Q l , or R1 on fluorescence, where A J = - 1,0 ,or + l is identified with P, Q,
or R, respectively. (Here A J = Je - Ji on excitation and A J = Je - Jf on
fluorescence). Algebraic expressions for the polarization are collected in Table 5.1
for the different branches by evaluating Eq. (5.114) for R and using Eq. (5.112) to
reexpress R in terms of P. These expressions have been found previously[24-261
but by much more cumbersome means. As Ji increases, the results in Table 5.1
approach the classical limiting forms found previously in Application 8, specifically:
R = 215 ( P = 112) f o r ( Q f , Q J ) ; R = 1/10 (P= 117) for(Pf,PL),(Pf,R.L),
( R f , P l ) , or (RT,Rl); and R = -115 ( P = -113) f o r ( Q T , P l ) , ( Q f , R 1 ) ,
( P f , QJ), or ( R f , QJ). The effects of hyperfine structure are easily included[l7,
22,231, easy that is, when the tensor methods of this section are used!
Another variant of the resonance fluorescence process occurs when the incident
radiation is circularly polarized, with either left- or right-hand helicity, and the
detected light is analyzed for circular polarization[24, 271. Here the propagation
direction of the incident and detected light beams may be chosen to define the Z axis.
This choice has the virtue that once again only one component of the polarization
5.4 TENSOR PRODUCT MATRIX ELEMENTS 209

vector expressed in the spherical basis set is nonvanishing, either e( 1 , l ) for left
circularly polarized light or e( 1 , - 1) for right circularly polarized light.
Recall that the electric field vector E of a monochromatic wave of frequency w
and wave vector k = 2 n/Xfi, where fi is a unit vector in the direction of motion, can
be written in the form

E(t) = A@exp[i(k.r- wt)] (5.115)

where A is the amplitude and & is a unit vector called the polarization vector. Because
of the transverse nature of electromagnetic waves, 6 is perpendicular to 8. Hence the
choice of ii to be along the Z axis confines & to lie in the XY plane. For light in
a pure state of arbitrary polarization pmpagating along the Z axis, its polarization
vector may be represented by

@(p,6) = cos p$ i ei6sin pP (5.116)

where p ranges from 0 to n and 6 is the phase necessary to describe elliptical

polarization. For example, for linear polarization S = 0 and /3 is the angle between
the polarization vector and the X axis. Circularly polarized light corresponds to equal
components of 6 along 2 and Y but vibrating out of phase by 90". We define left
circularly polarized light by P = n/4 and S = n/2

and right circularly polarized light by P = 3 n/4 and S = n/2

Thus the only nonvanishing spherical component for left circularly polarized light is
-e( 1 , l ) and for right circularly polarized light is -e( 1 , - 1).
Special care must be taken in interpreting the spherical components of @*. We
take e*( 1,p) to mean the ( 1,p) spherical component of the t*vector. Reference to
Eq. (5.116) for S = 7~/2shows that e( 1 , l ) = -(cos /3 i sin @ ) / a , e( 1, - 1) =
(cosP-sin/3)/fl,e*(l, 1) = -(cosp-sinp)/fi,ande*(l,-1) = (cospi
sin p) /fi. Hence

The polarization tensor ~ , k 2,( @*)can then be constructed according to Eq. (5.100)
with the help of Eq. (5.119). We find that for circularly polarized light the only
nonvanishing components of the polarization tensor are with q = 0 , specifically
210 5 SPHERICAL TENSOR OPERATORS

for left circularly polarized light and

for right circularly polarized light.

Let I(same) denote the resonance fluorescence signal having the same helicity as
the circular polarization of the incident beam and I ( opposite), the opposite sense of
circular polarization. It is useful to introduce the degree of circular polarization, C ,
which is defined in analogy to the degree of linear polarization, P, by
I(same) - I(opposite)
C=
I(sarne) + I ( opposite)
Then in the same manner as Eq. (5.109) was derived, it is found that

(5.123)
where 8 is the angle between the incident beam and the detection direction. Note
that C vanishes for a right-angle excitation-detection geometry (8 = n/2) but attains
a maximum value for a collinear geometry (8 = 0 or n). In particular, when the
excitation beam coincides in direction with that of the fluorescence beam detected
(8 = O), the degree of circular polarization is given by

Closed algebraic expressions and limiting forms for large Ji are listed in Table 5.2
for the nine possible resonance fluorescence branches.
For an isotropic sample, measurement of the degree of circular polarization in
molecular resonance fluorescence generally requires the resolution of individual ro-
tational branches in both the excitation and fluorescence steps. This is a consequence
of the fact that P and R branch transitions usually have the same strength from a given
excited level, and the sums C ( RT, RJ,) + C ( RT,P J ) and C( PT, RL) + C( P f ,PJ,)
are approximately zero for large Ji values. Moreover, resonance fluorescence
branches involving QT or QJ also make a negligible contribution to C in the high Ji
limit. For an oriented sample, nonzero values of C generally require only the res-
olution of individual rotational branches either in the excitation or the fluorescence
step.
5.4 TENSOR PRODUCT MATRIX ELEMENTS

TABLE 5.2 Resonance Fluorescence Process; Circular Polarization

Transition C( Ji ) C( Ji --+ CO)

We have assumed that the light beam is in a pure polarization state. This need
not be the case, and a beam of photons is said to be in a mixed state if it is not
possible to describe the beam in terms of a single state vector, as in Eq. (5.116). A
mixed state might result from two (or more) light sources emitting independently
with no definite phase relation between the sources. An example is a beam of totally
unpolarized light that may be regarded as equal incoherent combinations of two
beams of opposite polarization, that is, one beam of horizontal linear polarization
with another beam of vertical linear polarization having the same intensity or one
beam of left circular polarization with another beam of right circular polarization
having the same intensity. Mixed photon states are most conveniently treated by use
of density matrix methods[28].
In the foregoing we have also assumed that'the ground-state distribution of Ji
directions is random. This need not be the case, especially if the molecules are
formed in some directionalprocess such as in photodissociation or in molecular beam
scattering either with other molecules or with surfaces. The determination of the
angular momentum distribution by use of the resonance fluorescence process has
been considered elsewhere in some detail[29-321.
212 5 SPHERICAL TENSOR OPERATORS

NOTES AND REFERENCES

1. Wigner and Racah were the first to introduce irreducible tensor operators [see
E. P. Wigner, "On the Matrices which Reduce the Kronecker Products of Represen-
tations of S. R. Groups," in Quantum Theory of Angular Momentum, L. C. Biedenham
and H. van Dam, eds. (Academic Press, New York, 1965); E. P. Wigner, Am. J. Math.,
63,57 (1941); G. Racah, Phys. Rev., 61, 186 (1942); 62,438 (1942); 63,367 (1943)l.
The latter three papers by Racah are classics in the field of atomic spectroscopy because
they first illustrated the power of these tensor methods compared to previous treatments.
For an excellent introduction to this topic, see B. L. Silver,Irreducible Tensor Methods
(Academic Press, New York, 1976), which contains many worked-out examples,
particularly in regard to applications to atomic systems in spherical and point group
environments. Other extremely useful references are U. Fano and G. Racah, Irreducible
Tensorial Sets (Academic Press, New York, 1959); B. R. Judd, Operator Techniques in
Atomic Spectroscopy (McGraw-Hill, New York, 1963); B. R. Judd, Angular Momentum
Theoryfor Diatomic Molecules (Academic Press, New York, 1975); and I. I. Sobel'man,
An Introduction to the Theory of Atomic Spectra (Pergamon Press, New York, 1972).
Finally, for all you might want to know (and probably more!), there is the two-volume
set: L. C. Biedenharn and J. D. Louck, Angular Momentum in Quantum Physics
(Addison-Wesley, Reading, MA, 1981); The Racah-Wigner Algebra in Quantum
Theory (Addison-Wesley,Reading, MA, 1981).
2. R. F, Curl, Jr. and J. L. Kinsey, J. Chem. Phys., 35, 1758 (1961).
3. J. M. Brown and B. J. Howard, Molec. Phys., 31, 1517 (1976); 32,1197 (1976).
4. E. P. Wigner, Group Theory (Academic Press, New York, 1959); C. Eckart, Rev. Mod.
Phys., 2.305 (1930).
5. D. M. Brink and G. R. Satchler,Angular Momentum (Clarendon Press, Oxford, 1979).
6. A. R. Edmonds, Angular Momentum in Quantum Mechanics (Princeton University
Press, Princeton, NJ, 1974).
7. S. Devons and J. B. Goldfarb, "Angular Correlations," in Encyclopedia of Physics
Vol. XLII, (Springer-Verlag, Berlin, 1957), pp. 362-554.
8. L. C. Biedenham, in Nuclear Spectroscopy, E Ajzenberg-Selove, ed. (Academic Press,
New York, 1960), Part B, pp. 732-810.
9. H. Feshbach, in Polarization Phenomena in Nuclear Reactions, H. H. Barschall and
W. Haeberli, eds. (University of Wisconsin Press, Madison, WI,1971),pp. 29-38.
10. R. M. Steffen and K. Alder, in The Electromagnetic Interaction in Nuclear Spectroscopy
(North-Holland, Amsterdam, 1975),Chapters 12 and 13.
11. M. Peshkin, Adv. Chem. Phys., 18,1(1970).
12. R. N. Dixon, J. Chem. Phys., 85,1866 (1986).
13. D. A. Case and D. R. Herschbach, Molec. Phys., 30, 1537 (1975); D. A. Case
and D. R. Herschbach, J. Chem. Phys., 64, 4212 (1976); G. M. McClelland and
D. R. Herschbach, J. Phys. Chem., 83, 1445 (1979); J. D. Barnwell, J. G. Loeser, and
D. R. Herschbach, J. Phys. Chem., 87,2781 (1983).
14. E. U. Condon and G. H. Shortley, The Theory of Atomic Spectra (Cambridge University
Press, 1935); J. S. Griffith, The Theory of Transition-Metal Ions (Cambridge University
Press, 1964).
NOTES AND REFERENCES 213

15. Graphical methods are readily adapted to the evaluation of spherical tensor matrix
elements; see for example I. Lindgren and J. Morrison, Atomic Many-Body Theory
(Springer-Verlag, Berlin, 1982), Chapters 3 and 4. For example, the Wigner-Eckart
theorem becomes

16. For small molecules, like diatomics and triatornics, the excited state rotational level J,
is almost always unique since only one such rotational transition Ji + J, can occur for
a short pulse of light with a small frequency spread. For large molecular systems, it is
possible to prepare a coherent distribution of excited rotational levels with a short pulse
of light. This results in more than one indistinguishablepath from Ji to Jf and causes the
appearance of rotational quantum beats. See P. M. Felker, J. S. Baskin, and A. H. Zewail,
J. Phys. Chem., 90,724 (1986); 1. S. Baskin, P. M. Felker, and A. H. Zewail, J. Chem.
Phys., 84,4708 (1986); P. M. Felker and A. H. Zewail, J. Chem. Phys., 86,2460,2483
(1987);and N. F. Scherer, L. R. Khundkar, T. S. Rose, and A. H. Zewail, J. Phys. Chem.,
91,6478 (1987).
17. A. Omont, Progr. Qwntwn Electronics, 5,69 (1977).
18. G. Breit, Rev. Mod. Phys., 5,91 (1933). See alsoP. Franken, Phys. Rev., 121,508 (1961)
and M. E. Rose and L. Carovillano, Phys. Rev., 122, 1185 (1961).
19. P. Schenck, R. C. Hilbern, and H. Metcalf, Phys. Rev. Lett., 31, 189 (1973).
20. M. I. Dyakonov, Sov. Phys. J E W , 20,1484 (1965).
21. H. Kretzen and H. Walther, Phys. Lett., 27A, 718 (1968).
22. G. Gouedard and J. C. Lehrnann, J. Phys., 34, 693 (1973); M. Broyer, G. Gouedard,
J. C. Lehmann, and J. Vigu6, "Optical Pumping of Molecules," Adv. Atomic Molec.
Phys., 12,165 (1976).
23. R. Luypaert and J. Van Craen, J. Phys. B, 10,3627 (1977).
24. P. P. Feofilov, The Physical Basis of Polarized Emission (Consultants Bureau, New
York, 1961).
25. R. N. Zare, J. Chem. Phys., 45,4510 (1966).
26. G. W. Loge and C. S. Parmenter, J. Chem. Phys., 74,29 (1981).
27. H. Kata, S. R. Jeyes, A. J. McCaffery, and M. D. Rowe, Chem. Phys. Lett., 39, 573
(1976); S. R. Jeyes, A. J. McCaffery, M. D. Rowe, P. A. Madden, and H. Kat6, Chem.
Phys. Lett., 47,550 (1977); R. Clark and A. J. McCaffery, Molec. Phys., 35,609, 617
(1978); S. R. Jeyes, A. J. McCaffery, and M. D. Rowe, Molec. Phys., 36, 845, 1865
(1978); M. D. Rowe and A. J. McCaffery, Chem. Phys., 34, 81 (1978); ibid., 43, 35
(1979); J. McCormack, A. J. McCaffery, and M. D. Rowe, Chem. Phys., 48,121 (1980);
A. J. McCaffery, Specialist Periodical Reports, "Gas Kinetics and Energy Transfer," 4,
47 (1981); B. J. Whitaker and A. J. McCaffery, Chem. Phys. Lett., 86,185 (1982); ibid.,
J. Chem. Phys., 78,3256 (1983); A. J. Bain and A. J. McCaffery, J. Chem. Phys., 83,
2627 (1985).
214 5 SPHERICAL TENSOR OPERATORS

28. K. Blum, Density Matrix Theory and Applications (Plenum Press, New York, 1981).
29. D. A. Case, G. M. McClelland, and D. R. Herschbach, Molec. Phys., 35,541(1978).
30. C.H.Greene and R. N. Zare, J. Chem. Phys., 78,6741 (1983).
31. R.I. Altkom and R. N. Zare, Ann. Rev. Phys. Chem., 35,265(1984).
32. A. C.Kummel, G. 0.Sitz, and R. N. Zare, J. Chem. Phys., 85,6874 (1986);J . Chem.
Phys. 88,6707 (1988); J. Chem. Phys. 88,7357 (1988); J . R. Waldeck, A. C. Kummel,
G. 0. Sitz, and R. N. Zare,J. Chem. Phys. 90,4112 (1989).
PROBLEM SET 3 215

PROBLEM SET 3

1. Use angular momentum coupling algebra to prove the vector identity

2. Derive Eq. (5.123) from Eq. (5.122). Offer an interpretation why C(J i ) -+0 as
Ji -t CCJ for the resonance fluorescence process involving QT or QJ branches.
3. Explore an alternative derivation of Eq. (5.105), which expresses the resonance
fluorescence intensity in the absence of an external field. Start with Eq. (5.93).

x (- M ~ 1/J' Ml
Je)( MJe 1
Y Mf
Jt)( -Mf
Jf ' Me
Y1 Je)
216 5 SPHERICAL TENSOR OPERATORS

Use the results of Problem Set 2, part 3 to reexpress the summation S as

Show that on substitution this yields the expression for the resonance fluores-
cence intensity given by Eqs. (5.105) and (5.106).

4. A second-rank Cartesian tensor TAB(A = X , Y, Z and B = X, Y, 2 ) has

3 = 9 independent components. Decompose TABinto its irreducible spher-
ical tensor components. Show that TABis composed of a zero-rank irreducible
tensor, proportional to Txx + Tyy + TZZ = Tr[T] and having one indepen-
dent component, a first-rank irreducible tensor, proportional to the antisymmet-
ric tensor &(TAB- TBA) and having three independent components, and a
second-rank irreducible tensor, proportional to the symmetric traceless tensor
1
%( TAB+ TBA)- f s ~ [TI~andThaving ~ five independent components. Present
a general procedure for decomposing an arbitrary Nth-rank Cartesian tensor into
its spherical irreducible components. Hint:Any Cartesian tensor may be writ-
ten in a spherical basis using the correspondence between first rank Cartesian
tensors Tx ,Ty ,Tz and first rank spherical tensors T,('), T:'), T(:) :

with the inverse relation

PROBLEM SET 3 217

Thus the AB element of a second-rank Cartesian tensor TABis written in a

spherical basis as

The product Tj ')T:') is not irreducible but transforms under rotation according
to

and it follows that

Then solve for the coefficients C,, in the expansion. The generalization to
higher-order Cartesian tensors relies on repeated application of the procedure
shown above. The explicit decomposition of a third-rank Cartesian tensor into
spherical irreducible components is given in B. Dick, Chem. Phys., 96, 199
(1985).
218 5 SPHERICAL TENSOR OPERATORS

APPLICATION 11
THE ENERGY LEVELS OF A TWO-VALENCE-ELECTRONATOM
REVISITED

A. Work out the term energies for the p2 configuration, using spherical tensor
methods:

where we have made use of the expansion

Show in particular that

It is traditional to follow the practice of Condon and Shortley and to define

Fo = F0 and F2 = ( 1 /25) F2 so that
APPLICATION 11 ATOMIC ENERGY LEVELS REVISITED 219

B. By considering only the electrostatic interaction, the resulting energy expression

does not contain J, and hence the different J levels 2 S + ' ~ J of the term 2 S + 1 ~
have the same energy. However, the inclusion of the spin-orbit interaction term

in the Hamiltonian removes this degeneracy. Find the E ( ~ ~ +treating ~ L ~Hso

)
as a perturbation; that is, use only first-order perturbation theory.
Set up, but do not evaluate, the exact expressions for E ( ~ ~ + 'whenL ~ the
) spin-
orbit interaction is not small compared to the separation between term energies.
Express your energy level expressions in terms of the spin-orbit parameter

It may be shown that

As I'approaches 0, the energy levels follow LS coupling; if we keep only terms

first order in, ,C the energy level spacings follow the Land6 interval rule; and if
ST/ F2 tends to infinity, the energy levels follow jj coupling.

C. Fluorescent lighting utilizes the Hg 6s 6 p 3 ~ P-+ 6s2 'So W transition at

253.7 nm to cause a phosphor on the inside walls of a fluorescent lamp to glow
"white." The (strongly allowed) resonant transition of Hg corresponds to the
transition 6s 6p 'Pi' + 6s2 'SOat 184.9 nm.Present arguments to explain why
the Hg 253.7 nm line has significant intensity although singlet-triplet electric
dipole transitions are forbidden in the L S coupling limit.
220 5 SPHERICAL TENSOR OPERATORS

APPLICATION 12
DIRECTIONAL CORRELATIONS;
BREAKUP OF A LONG-LIVED COMPLEX

Consider a reactive scattering process

As a first approximation,we regard the reagents and products to be spinless particles

possessing no internal angular momenta. Let k and ii' specify the respective initial
and final relative velocity directions in the center-of-mass frame, and let t and t' be
the corresponding external orbital angular momenta whose directions are perpendic-
ular to k and k , respectively, and whose magnitudes are given by the product of the
reduced mass, the impact parameter, and the relative velocity magnitude of the colli-
sion partners. We denote the total angular momentum of the scattering system by j.
Conservation of angular momentum implies that

The direction correlation[l, 21 between k and kt, which is proportional to the

differential cross section, depends only on the included angle 8, where

that is, we may write the differential cross section as the function I(8). Next we
expand I ( 8) in a complete set of Legendre polynomials:

We may solve for the expansion coefficients A, by multiplying both sides of Eq. (3)
by P,~(cos8) and integrating over all solid angle elements cUZ = sin 8 d0 dd. With
the help of Eq. (1.59) we find that

Here the expansion coefficient A, is interpreted as the value of P,(k - k') averaged
over all possible configurations of the scattering system.
APPLICATION 12 DIRECTIONAL CORRELATIONS 221

We choose to define the z axis to be along f!. We denote the polar angles of k by
( 81, 41) and those of 1;' by (82, 42). Use of the spherical harmonic addition theorem
[Eq. (3.85)1, allows us to rewrite Eq. (4) as

So far these results are general. Now we introduce the condition that the reaction,
which is characterized by the angular momentum f!, is long-lived. If the time between
formation and breakup of the complex is long compared to the rotational period of the
complex, k will be uniformly distributed about f!' = f!. This has the consequencethat
the angle in the xy plane conjugate to l is uniformly distributed, that is, the collision
complex breaks up with a random value of 4, - 4 2 .

and hence that

Av = [pV(0)l2

For such a long-lived collision complex it follows that

Explain why this result implies that the product angular distributionhas forward-
backward symmetry in the center-of-mass frame, that is, is symmetric about
8 = 7r/2.

Show that I(8) is proportional to 1/ sin 8, that is, peaks at the poles and fans
out at the equator. Hint:Expand 1/ sin 8 in a Legendre series and compare to
Eq. (8).
We remove the restriction that the reactants and products are spinless particles.
The total angular momentum of the collision complex is then

where s = JA + JBC and s' = JAB + JC denote the entrance and exit channel spins
(angular momenta). In the classical limit each angular momentum value may be
222 5 SPHERICAL TENSOR OPERATORS

associated with a definite direction in space, specified by its corresponding unit

vector. Successive applications of the spherical harmonic addition theorem give

= ( ~ ( .kl)P,(&' . t ) P U ( l .fi put; . t ) )

av

where we assume that directional properties of the reagent and product trajec-
tories are uncorrelated except by the conservation of total angular momentum
(definition of a separable long-lived collision complex[2]). Here the first line of
Eq. (10) is a consequence of the random distribution of the dihedral angle f,bkjklr

and the second line is a consequence of the random distributionof f,bkej and +klelj,
which results because s is assumed randomly oriented with respect to k and st
is unobserved. Note that an unobserved direction is equivalent to a random one
since each leads to an unweighted average over the relevant dihedral angle. Show
once again that only even values of v contribute to 1(0), that is, that I(0) has
forward-backward scattering symmetry.
E. The preceding arguments are semiclassical in nature in that they ignore the dif-
ficulty of specifying the magnitude of e and its exact direction. A fully quantum
mechanical treatment shows that we must replace the Legendre function of the
angle between a quantized angular momentum l a n d an unquantized direction &
by an appropriate Clebsch-Gordan coefficient

for large &,and the Legendre function for the angle between the two quantized
angular momenta t? and 3 by an appropriate 6 -jsymbol

for large 4 and j. The derivations of Eqs. (11) and (12) may be found in
P. J. Brussaard and H. A. Tolhoek, Physica, 23, 955 (1957). The complete
expression for the differential cross section becomes
APPLICATION 12 DIRECTIONAL CORRELATIONS

An important consequence of the quantization is that the index u is restricted to

a finite number of even values. What conditions can be placed on the maximum
value of u?

These same considerations apply to the directional correlation of photons emitted

from an excited state, such as the angular correlation of successive gamma rays from
an excited atomic nucleus[3] or the angular correlation of two fluorescence photons
successively emitted from an electronically excited atomic or molecular state. We
consider the process j i + j followed by j -t jf , where ji, j , and jf are the
angular momentum quantum numbers of the initial, intermediate, and final states of
the system, respectively. We denote the two directions as ki and Gf and suppose that
in the first transition the emitted photon has the intrinsic angular momentum t i and in
the second transition tf. We also suppose that the intermediate state is not disturbed
(perturbed) between the time it is formed and the time it decays.
Any correlation in which only two directions of motion are measured (called
a direction-direction correlation) must be expressible as a Legendre series in the
angle Oif between the two directions ki and kf . In analogy to Eqs. (3) and (4) we
may write

Because the polarizations of the two photons are not measured, we may regard t i
and tf to be randomly and uniformly oriented. Repeated application of the spherical
harmonic addition theorem yields

The above gives the general form of the direction-direction angular distribution
as more rigorous arguments show[3]. Let us specialize this to the emission of two
electric dipole photons. In this case ti = tf = 1. Because the electric vector is
transverse to the propagation direction of the radiation, the projection o f t on k can
have only the values plus or minus one (net helicity of unity). We use semiclassical
224 5 SPHERICAL TENSOR OPERATORS

arguments to deduce the quantum transcription of Eq. (16). In particular [in analogy
to Eq. (11)l

and [in analogy to Eq. (12)]

with correspondingexpressions for P,(& &) and P,(& .i). Thus for the successive
emission of two electric dipole photons the direction-direction angular correlation
has the form

where a2 depends on ji,j,jf as

up to a normalization constant. Hence a measurement of a2 can be used to deduce

the value of j in the unobserved intermediate state!
A simple example[4] serves to illustrate Eq. (19). We consider the radiative
cascade process 'S -t 'P -t 'S in an atom corresponding to ji = 0 , j = 1, and
jf = 0. In the first transition oi --+ j) the three possibilities A m = 0 , Am = +1,
and Am = - 1 are equally probable. However, a Am = 0 transition in ji -t j must
be followed by a Am = 0 transition in j -t jf. Similarly, a Am = f1 transition in
ji -+ j must be followed by a A m = r~ 1 transition in j -t jf.
The directional photon distribution in each of these transitions has a dipole antenna
pattern. Specifically, for a Am = 0 transition the photon angular distribution
is sin2 8 = 1 - cos2 8, whereas for a Am = & l transition the photon angular
distribution is &(1+ cos2 8) ,where 8 is measured from the quantization axis. Hence
the radiation in the ji = 0 -t j = 1 transition is isotropic when summed over all
possible A m components weighted by their transition probabilities (which should be
no surprise since the state ji = 0 , mi = 0 is also isotropic).
The direction of the quantization axis may be arbitrarily chosen without any loss
of generality. In particular, we choose the z axis to lie along the direction of the first
photon; that is, z is parallel to ki . For this choice no A m = 0 transition can occur
since the photon angular distribution associated with the Am = 0 transition has a
node along the z axis; that is, for Am = 0 transitions no radiation is emitted in the
z direction. Hence the first transition is of the Am = f 1 type, and this must be
followed by a second transition of the Am = ~1 type. Therefore, it follows that
NOTES AND REFERENCES 225

where the normalization constant is chosen so that the total emission probability of
the second photon is unity. According to Eq. (21), the direction4irection correlation
for the emission of two successive electric dipole photons is twice as large when the
two photons are detected in the same or opposite directions (Oif = 0 or T ) than when
detected at right angles (eif = 7~12).

E. Yang[S] has used group theoretic arguments to prove a general theorem about
angular correlations between successive photons; namely, W (&f ) is an even
power of cos eif, and the highest power in this polynomial is smaller than or
equal to 2 4 , 2ef, or 2 j, whichever is the smallest. Show that the semiclassical
arguments presented in the preceding paragraphs lead to the same conclusion as
Yang's theorem.

NOTES AND REFERENCES

1. L. C. Biedenharn, in Nuclear Spectroscopy, F. Azjenberg-Selove, ed., Academic Press.

New York, 1960, Part B, p. 732.
2. G. M. McClelland and D. R. Herschbach, J. Phys. Chem., 83, 1445 (1979). These
authors show that even if a complex survives for many rotational periods, the product
angular distribution need not display forward-backward symmetry unless the complex
is separable. This replaces the "beguiling but misbegotten conclusion" that all long-
lived collision complexes yield forward-backward scattering symmetry. Unfortunately.
examples of nonseparable long-lived complexes appear not to have been observed.
3. There is an extensive literature on correlations in nuclear cascade processes, starting
with the pioneering work of D. R. Hamilton, Phys. Rev., 58, 122 (1940), who treated
7-ray correlations using time-dependent perturbation theory. See also G. Goertzel,
Phys. Rev., 70, 897 (1946), who extended Hamilton's treatment and D. L Falltoff and
G. E. Uhlenbeck, Phys. Rev., 79,323 (1950), who generalized this treatment to successive
radiations of arbitrary multipole moments. The first use of tensor operators is to be
found in G. Racah, Phys. Rev., 84, 910 (1951); for a review, see L. C. Biedenharn
and M. E. Rose, Rev. Mod. Phys., 25, 729 (1953). However, angular wrrelations are
most generally treated by using the powerful density matrix formalism; see S. Devon
and L. J. B. Goldfarb, "Angular Correlations." in Encyclopedia of Physics, Vol. XLII,
Springer-Verlag, Berlin, 1957, pp. 362-554, as well as R. M. Steffen and K. Alder, in The
Electromagnetic Interaction in Nuclear Spectroscopy, North-Holland, Amsterdam, 1975,
pp. 505-643.
4. This example is taken from J. M. Blatt and V. F. Weisskopf, Theoretical Nuclear Physics,
Wiley, New York, 1952, pp. 635-639.
5. C. N. Yang. Phys. Rev., 74.764 (1948).
226 5 SPHERICAL TENSOR OPERATORS

APPLICATION 13
ORIENTATION AND ALIGNMENT

Inelastic and reactive scattering, as well as photodissociation and photoionization

processes, typically produce anisotropic distributions of the angular momentum
of the collision products and photofragrnents. We consider how to describe the
angular momentum polarization. One natural choice is to make an expansion of the
angular momentum distribution into multipole moments. Here the monopole term is
proportional to the population. All odd multipoles (dipole, octupole, etc.) describe
the orientation of the angular momentum J, that is, which way J preferentially points
regarded as a collection of single-headed arrows, and all even multipoles (quadruple,
hexadecapole, etc.) describe the alignment of J, that is, the spatial distribution
of J regarded as a collection of double-headed arrows. If all multipoles (except the
monopole) vanish, the spatial distribution of J is said to be isotropic; that is, J is
uniformly distributed in space. In the IJ M ) representation, orientation corresponds
to a preferential population of some IJ M) versus some IJ -M ) level causing
the system to have a net helicity or spin. In the same representation, alignment
corresponds to a preferential population in the combined IJ M ) and IJ -M ) levels
compared to the combined IJ MI) and IJ -MI) levels. Note that a system can exhibit
both orientation and alignment.
To carry out the multipole moment expansion of J, we introduce the spherical
tensor operators J,( k ) . In accord with Eq. (5.15)

It follows from Eqs. (5.6) and (5.7) that

Prove that
k
= [JZ] (4)

Hint: Contract J,$) and J:; to form the spherical tensor JZ)and make an
inductive argument.
APPLICATION 13 ORIENTATION AND ALIGNMENT 227

Application of Eq. (3) to Eq. (4) can be used to generate the J,(k, for all q. We list
for reference purposes the Jlk)for k = 0 through 4 in terms of J*, J,, and J2 :

B. Use Eqs. (2) and (3) to derive JZ)and Ji2)from ~2).

228 5 SPHERICAL TENSOR OPERATORS

C. The angular momentum tensors satisfy the relation

By explicit reference to Eqs. (8), (9). and (lo), show that ( J $ ) ) ~ = J:?,
(J!?)~ = -J:), and (.Ti2))+ = ,Ti2).Hint: Recall that ( A B ) ~= B ~ A ~ ,
where A and B are two operators.

We are interested in calculating the expectation values of the operators J,(k). This
motivates a more general consideration of what is the minimum set of information
required to calculate the mean value of any operator Q for our system "prepared" by
collision with another system or a photon. This question is most readily answered by
using density matrix methods[l-31.
We suppose that our system is in a definite angular momentum state. If it is in a
pure state, then its state vector I$) can be expanded in terms, of the complete set of
eigenvectors IJ M )

where

The mean value of the operator Q is given by

In the general case our system is in a mixed state, which is represented by a

density operator p that is an incoherent superposition of a number of pure
states I$(')with statistical weights w(')

where

In other words, in this ensemble w(') of the states are in I$(')). Then
APPLICATION 13 ORIENTATION AND ALIGNMENT 229

The density matrix is defined as

so that the expectation value of Q becomes

Here Tr[A] indicates the trace of the matrix A, that is, the sum of the diagonal
elements of A.

D. Prove that

E. Prove that p is Hermitian, that is,

Clearly, a complete knowledge of p suffices for calculation of the expectation

value of any arbitrary operator. We now consider the number of independent
parameters needed to specify a given density matrix. Of course, this depends on
the number of orthogonal states IJ M ) over which the sum in Eq. (21) ranges, but for
angular momentum J this sum is restricted to 2 J + 1. For this case p is a ( 2 J + 1)
by ( 2 J + 1) square matrix with (2 J + 1)2 complex elements corresponding to
2(2 J + 1)2 real parameters. The hermiticity condition given in Eq. (30) shows that
the number of real parameters is ( 2 J + 1) 2 , and the normalization condition given
in&. (29)reducesthisnumberto(2J + 1)2 - 1 = 4 J ( J + 1).

F. Show that for a system in a pure state of definite angular momentum J the number
of real parameters necessary to specify the density matrix is 4 J .

The diagonal and off-diagonal elements of the density matrix have special signif-
icance. The diagonal element (JMI p 1 J M ) = xi w(')1a$)12 is the probability of
finding the system in the state I J M ) ; that is, the diagonal elements of p represent
the populations in the set of orthogonal eigenstates. However, the offdiagonal ele-
ments of p vanish unless some two eigenstates I J M ) and IJ MI) are in phase so that
their ensemble average is nonzero; that is, the off-diagonal elements of p represent
coherent preparation of the system when expressed in the I J M ) representation.
230 5 SPHERICAL TENSOR OPERATORS

Note that if a pure state in one representation is transformed to another representa-

tion, then in general the density matrix has both diagonal and off-diagonal elements.
For a mixed state, the density matrix also has both diagonal and off-diagonal elements
in general, but there is a difference in that the mixed state has lost phase information
through the combination. Only for pure states is ~ r [ p =~ Tr[p] ] = 1; for mixed
states Tr [p2] is less than 1 and in the chaotic limit becomes 1/ ( 2 J + I), where the
dimensionality is 2 J + 1. To prove this, recall that the trace of a matrix equals the
sum of its eigenvalues Xi and that the eigenvalues of the density matrix correspond to
populations and hence are positive definite. Then we may write Tr[p] = Xi Xi = 1
a n d ~ r [ =~ CiX;.
~] <
~ e c a u s e C ~ X5; (CiXi12, w e c o n c l u d e t h a t ~ r [ ~ ~1.
]
The minimum value for Tr [p2] occurs in the uniform mixed state for which each
eigenvalue Xi equals 1/(2 J + 1). The pure states and the uniform mixed state of
an ensemble are at opposite extremes with respect to the extent of order. A pure
state represents a completely ordered ensemble, whereas the mixed state that is
uniform is in a state of maximum disorder. A more fa-
miliar measure of the lack of information or disorder is the entropy S, defined as
S = -Tr [ p In p] . Then for a pure state S = 0,whereas for a state of maximum
disorder S = l n ( 2 J + 1 ) .
We return to the problem of describing the orientation and alignment of a system
in a definite state of angular momentum. This is accomplished by calculating the
mean value of the angular momentum tensor components J,(k ) . Hence the multipole
moments of the system are proportional to

where the brackets (. -) in the last line of Eq. (31) remind us that this is an ensemble
average. The multipole moments (J: k ) ) contain all the dynamical information that
can be learned about the system as to how the angular momentum vector points in
space. As Eq. (3 1) shows, its full calculation requires a complete knowledge of the
system's density matrix, a daunting task in general!
In many collision experiments we have axial symmetry, and when this is the case
calculation of the ( J , ( ~ )is) simplified. For example, there is cylindrical symmetry
along the beam axis of beam-gas scattering experiments (involving isotropic projec-
tiles incident on isotropic targets), along the relative velocity vector of two crossed
(isotropic) beams, along the propagation direction of a beam of unpolarized light
incident on an isotropic gas sample, or along the electric vector of a beam of lin-
early polarized light incident on an isotropic gas sample. In such cases it is natural
to choose the Z axis to lie along the axis of cylindrical symmetry. It then follows
that the physical properties of the ensemble are independent of the choice of the X
APPLICATION 13 ORIENTATION AND ALIGNMENT 23 1

and Y axes. Thus the density matrix is diagonal in M in this coordinate frame, and
Eq. ( 31) takes on the particularly simple form

where the second line of Eq. (32) was obtained from the first by use of the Wigner-
Eckart theorem [Eq. (5.14)]. Reference to the 3 - j symbol in Eq. (32) shows that it
vanishes for all nonzero values of q. Hence we conclude that in a system having
axial symmetry the only nonvanishing multipole moments are (JAk))(provided the
quantization axis coincides with the symmetry axis). Conversely, the existence of a
q f 0 multipole moment implies that the system was prepared through a collision
process not having axial symmetry.
For reference purposes we list the reduced matrix elements ( J I I J ( ~II)J ) for k = 0
through k = 4 , on the basis of the convention of Eq. (5.14):

( J I IJ ( ~I )I J )= [ ( J - l)J(J+ l)(J+2)(2J-1)(2J+ 1)(2J+3)

10

( J - l)J(J+l)(J+2)(2J-3)(2J- 1)(2J+ 1)(2J+3)(2J+5)

70 1'
G. Verify Eq. (35).
H. Show that an axially symmetricsystem that also possesses the additional symme-
try of a plane of reflection perpendicular to the axial symmetry axis has only non-
vanishing multipole moments with even rank k, that is, only shows alignment. It
follows that photodissociation or photoionization of an isotropic gas sample by
a beam of plane polarized or unpolarized light can only produce photofragments
232 5 SPHERICAL TENSOR OPERATORS

that are aligned but not oriented. Similarly, beam-gas or beam-bearn scattering
in which a beam of isotropic particles impinges on an isotropic sample or beam
of atoms or molecules can only produce products that are aligned (in the coordi-
nate frame in which the quantization axis coincides with the axis of cylindrical
symmetry).
I. To make the foregoing less abstract, consider a J = 2 system having axial
symmetry. Let the magnetic sublevel populations N ( J, M) be
N(2,2) = $
N(2,l) = &
N(2,O) =O
N(2,-1) =&
N ( 2 , -2) =

Find all the nonvanishing multipole moments (Jjk))of this system.

For many problems it is convenient to introducethe dipolar orientation vector 0;')

and the quadrupolar alignment tensor AY). Of course, 0;') is proportional to (J,( '))
and At2) to (,Ti2)). Specifically, they are defined[4] by

and
A:')( J) = (6);~J( J + I)]-'Re (( J IJi2)I J ) ) (39)

In the high J limit, where cos 9 = M I [ J ( J + I)] ' I 2 , we find that 0;') equals
the ensemble average of Pl (cos 9) and its value ranges from 1 to - 1 and that Ah2)
becomes the ensemble average of 2 P2 (cos 9) and its value ranges from 2 to - 1.

J. Use Eq. (39) to find the value of the alignment tensor component AF)
for the
system described in part I. What is the average [root-mean-square (RMS)] value
of the cosine of the angle that J makes with the quantization axis?

As a further illustration of the power of these methods, we consider at some length

the orientation and alignment produced by photoionization[5]. First we introduce the
angular momentum persona:

jo-angular momentum of target state before photoelectron ejection

APPLICATION 13 ORIENTATION AND ALIGNMENT 233

1 - angular momentum of photon causing electricdipole-allowed transition

s, -spin angular momentum of photoelectron (s, = i)
L, -orbital angular momentum of photoelectron (& = 0 , 1 , 2 ,...)
j, -total angular momentum of photoelectron u, = L, + s,)
j, -angular momentum of target state after photoelectron ejection

We wish to determine the spatial distribution of jl .

The total angular momentum of the system j is given by jo + 1 prior to photoion-
ization and by j, + j, after photoionization. Conservation of angular momentum
implies
j=jo+l=je+jl (40)

However, it is much more convenient to characterize the photoionization process

not by the total angular momentum but by j,, the unobserved angular momentum
transferred from the target to the photofragments:

When so defined, j: commutes with the angular momenta associated with the ob-
sewed anisotropy, and hence the contributions to the anisotropy from each j, add
incoherently[6].
The photoionization cross section a ( j l m l ) for leaving the target in the state
I jl ml ) is proportional to the square of the dipole moment matrix element, averaged
over all initial orientations of the target, and summed over all possible (undetected)
photoelectron final states:

In Eq. (42) KOis a proportionality constant and the index q describes the polarization
state of the dipole photon (q = 0 corresponds to linearly polarized light, and q = f1
corresponds to left- and right-handed circularly polarized light, respectively).
234 5 SPHERICAL TENSOR OPERATORS

K. Show that

from which it follows that

that is, the cross section a( jl ml ) has the form of an incoherent summation over
the angular momentum transfer quantum number. Hint:Use the Wigner-Eckart
theorem and the identity (4.16).

L. We wish to calculate the orientation and alignment

and

and find the contribution for each angular momentum transfer continuum chan-
nel. With the help of Eq. (#), we may rewrite these expressions as

and
APPLICATION 13 ORIENTATION AND ALIGNMENT

Find expressions for OA1)(jl ,jt ;q ) and AF)(jl ,jt ;q) ,and show that

and that

Hint:Use the results of Eqs. (1.97H1.99) of Problem Set 1.

A number of conclusions follow from Eqs. (49) and (50). For example,
oh1)(j l ,jt ;q ) vanishes for q = 0, which implies that interaction of the (isotropic)
target atom or molecule with linearly polarized light cannot produce orientation of
the target following photoionization. Moreover, oh1)(jl ,jt ;q) changes sign if the
sign of the incident photon circular polarization is changed. Unpolarized light may
be regarded as an equal weighting of q = +1 and q = - 1. Therefore, interaction
of the (isotropic) target with unpolarized light also cannot produce orientation in the
target.
The orientation and alignment of the target state following photoionization can
be expressed in terms of the contribution from three angular momentum transfer
continuum channels jt = j l + 1, jt = jl , and jt = jl - 1. Let the fractional
contribution of each channel be denoted by f (jt) ,where

Then the orientation and alignment of the final state of the target may be written as
236 5 SPHERICAL TENSOR OPERATORS

Hence a measurement of 0:')(jl ) and AF)(jl ) provides sufficient information to

extract all the continuum channel contributions f (j t ) . Equations (51)-(53) com-
pletely disentangle the dynamicsof the photoionization process from geometricterms
involving the angular momentum of the resulting target state and the polarization state
of the absorbed photon.
We return once more to the description of orientation and alignment in a state of
definite angular momentum in order to develop a different point of view. Following
Fano[2], we introduce the operators

which have the matrix elements

The nonvanishing of the Clebsch-Gordan coefficient ( J MI, J -MI KQ) in the

definition of T ( J ) K Q shows that K is restricted to integral values extending from 0
to 2 J and that Q ranges from -K to K in integral steps.

M. Consider a rotation R acting on T( J ) K ~Show

. that

from which it follows that T ( J)KQ is a spherical tensor operator of rank K and
component Q. The T(J)xQ are called state multipoles, a designation whose
appropriateness will become evident as we proceed.

Clearly, the T( J)KQ span the ( 2 J + 1) by ( 2 J + 1) space of the density matrix:

Equation (54)is readily inverted by multiplyingboth sides by ( J N t ,J -N IKQ)and

summing over all values of K and Q:
APPLICATION 13 ORIENTATION AND ALIGNMENT

Specifically we may decompose p as

N. Invert Eq. (59) to show that

Hint: Use the orthonormality properties of the Clebsch4ordan coefficients to

prove that

where ( J M I T( J ) k Q IJM') = ( J M I I T ( J ) K QIJM)*.

The state multipoles TKQtransform as the spherical harmonics YKQand the expan-
sion coefficients ~ K =
Q ( T k Q )must be interpreted as the multiple moments of the
state (up to some undetermined constant that depends on the normalization chosen).
Substitution of Eq. (58) in Eq. (57) and comparison with Eq. (59) shows that

( T i Q )= C ( - I ) ~ - (~J M ' , J - M I K Q ) (JM'l p l J M )

M1.M
(62)

Alternatively, Eq. (62) can be inverted to yield the expression

Hence the two descriptions of a system in terms of density matrix elements and its
multiple moments are equivalent. To determine one is to determine the other.
However, we have previously shown that the ( J j k ) )also describe the multipole
moments of the system. Clearly, the ( J j k ) )and the (T(J ) i Q ) must be related.
Indeed, this must be the case because both transform as YKQ.Hence to determine
( J j k ) )is equivalent to determining the state multiples of a system.
We establish the proportionality constant. Application of the Wiper-Eckart
theoremtoEq.(55)showsthat(~II~~ll~) = ( 2 K + 1)'12. ~herefore
238 5 SPHERICAL TENSOR OPERATORS

and it follows that

that is,

and so on.

0.Consider an electronically excited atom or molecule with initial angular momen-

tum quantum number Ji that makes a spontaneous electric dipole transition to
a final state characterized by the angular momentum quantum number Jf. Let
the Mi population of the initial state be distributed according to the probability
law

that is, the initial state has cylindrical symmetry with an orientationcharacterized
by the value or)( Ji) and an alignment characterized by the value ~r)(
Present arguments to indicate that the probability P( Jf, Mf) of populating the
Ji).

final state Jf , Mfby the radiative transition is

Explicit evaluation of Eq. (68) shows that[7]

APPLICATION 13 ORIENTATION AND ALIGNMENT

The values of the proportionality constants are for Jf = Ji - 1

for Jf = Ji

and for Jf = Ji + 1

Hence the orientation and alignment of the final state is proportional to the
orientation and alignment of the initial state following a spontaneous electric
dipole transition from Ji to Jf;moreover, for J values large compared to the
angular momentum carried off by the photon, C(') ( Ji ,Jf) and Cf2)( Ji, Jf)
both approach unity; that is, the degree of orientation and alignment is unchanged
by the spontaneous electric dipole transition.

As a concluding example, consider the time evolution of a composite system

having angular momenta J and I. Here we have in mind a molecule with rotational
angular momentum J and nuclear spin angular momentum I. We suppose that our
system initially has no preference for how J or I points in space. Suppose at time
t = 0 that a photon or particle collision causes J to become oriented or aligned but
does not act on I. Consequently, the system acquires through the collision process
a multiple moment of its J distribution, (J;k, (t = 0)) = (J,(')) but an isotropic
distribution of its I distribution ( I ~ O ) ) . After the collision, J and I couple to form
240 5 SPHERICAL TENSOR OPERATORS

the resultant F, which commutes with the system Harniltonian H. According to the
vector model, the coupling of J and I causes J to precess about F. Hence we expect
that the presence of hyperline structure, even if unresolved, causes ( J j k )( t )) to be
less than ( J i k ) ( 0 ) ) that
, is, depolarizes the multipole moments of J. We wish to
determine the time dependence of ( ~ , ( ~ ) (and t ) )make these comments quantitative.
Using the Wigner-Eckart theorem we replace the matrix elements of one spherical
tensor for those of another with the same rank and component:

Thus the time dependence of (( JI J j k ) ( t ) I J ) ) is contained entirely in the time

dependence of the reduced matrix element (JI I ~ ( ~ )11 J() t. )We use Eq. (5.68) to
evaluate this. Let

X i k ) ( t ) = [ f k ) ( t )€9 (79)
4

be a spherical tensor of rank k and component q constructed from the angular mo-
mentum tensor operators in J space and I space. Then in the coupled representation

We solve Eq. (80) for ( J 1 1 ~ ( ~ ) 11(J t) )by using the orthononnality properties of the
9 - j symbols:

But x ( k )(t ) is related to x ( k ) ( t = 0 ) by a unitary transformation involving the time

evolution operator U ( t ) [see Eq. (3.16)]
NOTES AND REFERENCES 241

where H is the system Hamiltonian, which is diagonal in the coupled representation

with eigenvalues EF.Thus it is readily shown that

where

Under interchange of F' and F, the imaginary terms are odd and thus cancel in the
sum. Hence

Substitution of Eq. (83) into Eq. (78) shows that

We see that the interaction between J and I does not mix multipole moments of J
but causes their value in general to be reduced. If the hyperfine splitting is small, the
F' # F terms oscillate rapidly about zero and the time-averaged value of G ( ~ )t)( is
given by the F' = F terms in Eq.(85), namely,

NOTES AND REFERENCES

1. K. Blum, Density Matrix Theory and Applications, Plenum Press, New York, 1981.
2. U. Fano, Rev. Mod.Phys., 29,74 (1957).
3. D. ter Haar, Rep. Progr. Phys., 24,304 (1961).
4. The definition of Af)( J) is from U.Fano and J. H. Macek, Rev. Mod. Phys., 45,553
(1973). The definition of Oil)(J) is [ J( J + I)] ' I 2 times larger than that of Fano and
Macek.
242 5 SPHERICAL TENSOR OPERATORS

5. C. H. Greene and R. N. Zare, Ann. Rev. Phys. Chem., 33, 119 (1982).
6. The theory of angular momentum transfer is developed in U. Fano and D. Dill, Phys.
Rev., A6, 185 (1972);D. Dill and U. Fano, Phys. Rev. Lett., 29, 1203 (1972);D. Dill,
Phys. Rev., A7, 1976 (1973);D. Dill and J. L. Dehmer, J . Chem. Phys., 61,692(1974);
J . L.Dehmer and D. Dill, J . Chem. Phys., 65,5327(1976);D. Dill, S. Wallace, J. Siegel,
and J. L. Dehmer, Phys. Rev. Lett., 41,1230(1978);ibid., 42,411 (1979).See also U.Fano
and A. R. P. Rau, Atomic Collisions and Spectra, Academic Press, Orlando, IT,1986.
7. R. Bersohn and S. H. Lin, Adv. Chem. Phys., 16,67(1969).
APPLICATION 14 NUCLEAR QUADRUPOLE INTERACTIONS 243

APPLICATION 14
NUCLEAR QUADRUPOLE INTERACTIONS

Armed with a reference library for evaluating spherical tensor matrix elements, let us
put these to use in one important application, the interactionof the electric quadrupole
moment of the nucleus with the electric field gradient at the nucleus caused by all
electronic (and other nuclear) charges in the atom (or molecule)[l-6]. Specifically,
we will examine how nuclear quadruple interactions affect the appearance of pure
rotational spectra of molecules and how such spectra can provide information about
the electronic charge distribution in the molecule. The rotational spectrum of the
linear triatomic molecule, cyanogen chloride (ClCN), is chosen as an example.
The rotational energy levels of a rigid rotor have the form

where B is the rotational constant and J the rotational quantum number. For electric-
dipole-allowed rotational transitions, A J = f1 and the so-called pure rotation
spectrum consists of a series of lines located at

These considerations predict a simple spectrum of equally spaced lines having a

frequency separation equal to twice the rotational constant.
Indeed this is what is observed for 35C1 14N and 37C112c1 4 under ~ low
resolution, each molecule having a slightly different rotational constant because of
the differing masses of the " ~ land 3 7 ~isotopes.
1 However, under higher resolution
the pure rotation spectrum is actually far from simple! Figure 1 shows the 35C1CN
microwave spectrum (the "cICN microwave spectrum looks qualitatively very
similar) taken at two different pressures, 50 and 1 mTorr, using Stark modulated
phase sensitive detection to yield first-derivativelineshapes. These have been kindly
recorded for us by Professor John S. Muenter, Department of Chemistry, University
of Rochester, who suggested this example. The static electric field is parallel to the
electric field of the microwave radiation, and in the absence of nuclear quadrupole
interaction, Figure 1 corresponds to a J = 0,M = 0 to J = 1, M = 0 transition.
Clearly, there is not one line, as predicted by Eq.(2), but several closely spaced lines.
Why? Has the rigid rotor model failed?
The answer is that our treatment is incomplete; it does not consider the interaction
of the rotational angular momentum of the molecule with the spins of the nuclei. In
ClCN, C1 has I = (which holds for 3 5 ~ 1and 3 7 ~ 1N) has I = 1 (which holds
, C has I = 0 (which holds for 12c).Figure 1 shows that the J = 0 to
for 1 4 ~ )and
J = 1 transition of ClCN has t h e lines for just the 35 c~CNspecies spread out over
35 MHz.Under higher resolution two of these lines split up into triplet. and one into
a doublet with these new splittings ranging from 1.2 MHz to 50 kHz. So there are
eight lines instead of just one.
244 5 SPHERICAL TENSOR OPERATORS

The J = 0 level does not split into different components because, as will be shown
below, a rotational level with zero rotational angular momentum cannot couple with
nuclear spins. For the J = 1 level, the major coupling is with the C1 nucleus and the
more minor coupling is with the N nucleus. The vector sum of J and I is traditionally
labeled F. In the case of a single nuclear spin, F is the total angular momentum. If
there are two nuclear spins, as in ClCN, we need to introduce a coupling scheme.
We label the two nuclear spins I1 and 12. Then J + I1 = F1 and F1 + I2 = F.
We call F1 the intermediate angular momentum and F still the total. We have two
possible coupling schemes, depending on whether we associate I1 with the nuclear
spin of C1 or N, For ClCN, the interaction between J and the spin on the C1 nucleus
5
is dominant and we take I1 = and I2 = 1. Thus in the J = 1 level, Fl can
be i, ;
$, or corresponding to the three lines seen in Figure 1 at lower resolution.
4 i
Coupling Fl = with 12 = 1 gives F = or $, coupling Fl = $ with I2 = 1 gives
F= a, t ,or ,; and coupling Fl = $ with 1 2 = 1 gives F = $, $ or $.Hence the
J = 1 level is split into three levels by the C1 nuclear spin angular momentum (upper
trace of Figure 1)which is what is observed at normal sample pressure since pressure
broadening is sufficiently large to obscure the smaller splittings from N. However,
at reduced pressure (lower trace of Figure 1)the N nuclear spin angular momentum
causes each of the three levels to split further, one into a doublet (unresolved) and
the other two into triplets. The J = 0 to J = 1 transition of ClCN has been studied
by Lafferty, Lide, and Toth[7] and we list below the line frequencies associated with
( Fl ,F) for the J = 1 level of 35 ClCN:

The last two transitions are calculated to be only 50 lcHz apart and were not resolved.

A. Even these elementary considerations suffice to show that the coupling of J

with I(C1) is dominant in ClCN. Discuss how the appearance of the J = 0
to J = 1 transition of ClCN would differ from that in Figure 1 if the coupling
of J with the nitrogen nuclear spin dominated. Would the total number of lines
change? Of course, both coupling schemes provide a complete set of basis
functions for describing the nuclear quadrupole couplings and would give the
same answer when all interactions are fully considered. However, as is often
the case, one coupling scheme gives a better zero-order approximation to the
observed splittings. We turn next to the form of the coupling between J and I
that causes the hyperfine structure seen in Figure 1.
APPLICATION 14 NUCLEAR QUADRUPOLE INTERACTIONS

FIGURE 1 Microwave spectrum of the 35C1C~ J = 0,M = 0 + J = 1, M = 0 rotational

transition, taken by J. S. Muenter at two different pressures.

The classical expression for the electrostatic interaction of a nuclear charge density
pn(rn) with position vector r, and an electron charge density p,(r,) with position
vector re is

where the center of the nucleus is taken as the origin of the coordinate system. For
simplicity, we ignore other nuclei in the molecule (or surroundings) and assume
that the charge distribution of the electrons is entirely outside the nuclear charge
distribution, that is, re > r n . Making use of the expansion (see Application 11)
246 5 SPHERICAL TENSOR OPERATORS

we have

where it should be recalled that Chk)(O,4) = [4 n/(2 k + I)] 'I2 Ykq(O,4) and (e)
and ( n ) stand for the electronic and nuclear coordinates, respectively.
Here the k = 0 term represents the Coulomb term of attraction between two point
charges of separation re, the k = 1 term vanishes because nuclei in their ground states
do not possess permanent electric dipole moments, and the k = 2 term represents the
interaction of the electric quadrupole moment of the nucleus with the electric field
gradient at the nucleus caused by the electron charge distribution. The k = 3 term
vanishes for the same reason as the k = 1 term (recall that nondegenerate nuclei have
only even electric moments), and the k = 4 term is much smaller than the k = 2
term. Hence we need only consider the k = 2 term in what follows. We write the
Hamiltonian for the nuclear quadrupole interaction as the contraction of two second
rank spherical tensors

where the components of the electric field gradient tensor are

and those of the nuclear quadrupole tensor are

To make Eqs. (7) and (8) into valid quantum mechanical expressions, we need to
replace the charge densities by expectation values of the appropriate operators. We
begin with the electric field gradient at the nucleus. For a nucleus at the center of
the coordinate system and an electron located on the 2 axis a distance re away, the
electric field gradient can be written in terms of the potential V = e/re as

a2v 2e
Qzz = = -
r,3

If the electron is located off the Z axis by an angle Oe, then

APPLICATION 14 NUCLEAR QUADRUPOLE INTERACTIONS 247

and the generalization of Eq. (10) to an electron charge density distribution pe (re) is

The quantum mechanical transcription of Eq. (11) is readily obtained. We define the
electric field gradient coupling constant as

where I a J J ) denotes the rigid rotor wave function with J making a maximum
projection on the space-fixed Z axis of MJ = J.

B. Show that q,, vanishes for J = 0 or J = and show that for J 2 1

In a similar manner it is traditional to define the electric quadrupole moment of

the nucleus as

and just as in Eq. (12) to write

where I 11) is the nuclear spin function with I making a maximum projection on the
space-fixed Z axis of MI = I. In exact analogy with Eq. (13) it follows that the
i,
nuclear electric quadrupole moment vanishes for I = 0 or I = and for I 2 1

Reference to Eq. (14) shows the nuclear quadruple moment takes on a positive value
if the nuclear charge distribution is prolate (elongated),takes on anegativevalue if the
nuclear charge distribution is oblate (flattened), and vanishes for a spherical nuclear
charge distribution as in the states I = 0 and I = i.
248 5 SPHERICAL TENSOR OPERATORS

C. Consider the coupled states IJIFMF). Find an expression for the nuclear
quadrupole interaction energy using first-order perturbation theory. Hint : Show
that

where

The factor eqzzQ is called the quadrupole coupling constant. If qZz can be
calculated from the electronic structure, then a measurement of the nuclear
quadrupole coupling constant allows a determination of the nuclear constant Q.

Equation (17) is a celebrated result that can be applied to both atoms and mol-
ecules. It assumes that the quadrupole splitting is much smaller than the spacing
between different J levels. When this is not the case, we must solve a secular deter-
minant involving interacting levels with the same F but different J . This introduces
matrix elements of the form (J1IFMFIv(') - Q(')I J I F M F ) and can cause levels
with A J = 0,f1, and f2 to interact. Fortunately for ClCN, the J = 0 to J = 1
rotational spacing (N 12,000 MHz)is so much larger than the hyperfine splittings
of the J = 1 level (N 25 MHZ) that we need not go beyond first-order perturbation
theory.
However, when applying Eq. (17) to molecules we encounter the awkward prob-
lem that the electric field gradient coupling constant qZz is referenced to the space-
fixed Z axis, causing the value of qzz to depend not only on the electron charge
distribution in the molecule but also on an average of the orientation of the molecule
with respect to J. Consequently, for molecular applications it is more convenient
to express the electric field gradient coupling constant in terms of the principal in-
ertial axis system fixed in the molecule since the field gradients with respect to the
molecule-fixed system are constant independent of the rotational state, For a linear
molecule, the principal-axis system has its z axis along the internuclear axis and we
can relate q,, E q to qzz by
APPLICATION 14 NUCLEAR QUADRUPOLE INTERACTIONS 249

iq
In the high J limit, qZZ = - since classically the internuclearaxis is perpendicular
i.
to J and the expectation value of ( 3 cos 0. - 1)/2 comes - Hence the first-
order nuclear quadruple energy for a linear molecule is given by

D. Estimate the value of eqQ for 35 C1 in 35 ClCN from the &ta provided. Show
that this is approximately -83 MHz. Hint:Use Eq. (20) to evaluate the matrix
(4 I I
elements (FMFI HQ I FMF), and show that MF HQ $MF) = -eqQ/4,
I I I
(* MF HQ MF) = eqQI5, and ($MF HQ $MF) = -eqQ/20. I

This value of eqQ(35C1) = -83 MHz in ClCN immediately allows us to draw

some conclusions concerning the nature of the C-Cl bond in the molecule. Contribu-
tions to q in a molecule may arise from valence electrons of the nucleus under study,
distortion of the closed shells of electrons around this nucleus, and charge distribu-
tions associated with nearby atoms. Because q refers to the electric field gradient
at the nucleus, vanishes for spherically symmetric charge distributions, and depends
inversely on the third power of the charge separation, the overwhelmingly dominant
contribution to q is from the nature of the valence electrons. For pure s valence elec-
trons q = 0 (again because the charge distribution has spherical symmetry), while
for pure p valence electrons the value of q is large because the electron density at the
nucleus is rapidly changing. Therefore, eqQ provides an excellent measure of the
p character associated with the bonding involving the nucleus under study. Note that
for an isolated free C1 atom (pure p valence electrons) eqQ(35C1) = -110 MHz,
while for a C1- anion (closed valence shell) e q ~ ( ~ * c=l )0. Thus, for example, in
KCl, eqQ(35~1)is less than 0.04 MHz while in CH3C1, FC1, and ICl, eqQ(35C1)
is -75 MHz, -146 MHz, and -83 MHz, respectively. Hence we conclude that
the C-Cl bonding in ClCN is covalent rather than ionic and that CN behaves as a
pseudo-halogen. This is one example of the rich chemical information available from
high-resolution studies of molecular hyperfine structure.
As Figure 1shows, the pure rotation spectrum of ClCN contains more information,
namely, the "super hyperhe splittings" caused by 14N with I = 1. We treat this
problem by working in the I( J I l ) F l 12FMF) coupling scheme where J + I1 = F1
and F1 + 12 = F and set up the secular determinant involving all levels with the
same F but different Fl . We now need to evaluate matrix elements of the form
250 5 SPHERICAL TENSOR OPERATORS

For ClCN J = 1 there are eight hyperbe levels and the secular determinant
is 8 x 8. However, by taking advantage of the fact that F is a good quantum
number, we can factor this problem into a 2 x 2 block for F = q,
a 3 x 3 for
F = %,another2 x 2 f o r F = $ , a n d a l x l f o r F = $. Eachofthesecanbe
diagonalized individually since there are no off-diagonal elements connecting these
subblocks. Thus by assigning the spectrum and measuring at least some of the small
splittings, it is possible to determine the J = 0 to J = 1 rotational spacing (2 B)
and the two nuclear quadruple coupling constants eqQ(35Cl)and eqQ(14~).In a
serious spectroscopic study one would make a least-squares fit to the line positions
in which B, eqQ(35C1),and eqQ(14N) were adjustable parameters. In this manner
Lafferty, Lide, and Toth found for 3 5 C 1 1 2 ~ 1eqQ(35~1)
4~ = -83.39 f 0.20 MHz
andeqQ(l4PJ) = -3.37 f0.26 MHz.

E. Set up, but do not solve, the 2 x 2 secular determinant for the F = block.
Hint: Evaluate Eq. (21) to show that the nonvanishing matrix elements in this
block are

NOTES AND REFERENCES

1. C. H. Townes and A. L. Schawlow. Microwave Spectroscopy, McGraw-Hill, New York,

1955, Chapter 9.
2. J. E. Wollrab, Rotational Spectra and Molecular Structure, Academic Press, New York,
1967, Chapter 5.
3. E. A. C. Lucken, Nuclear Quadrupole Coupling Constants, Academic Press, New York,
1969.
NOTES AND REFERENCES 251

4. W. Gordy and R. L. Cook, Microwave Molecular Spectra, Wiley-Interscience, New

York, 1970.
5. R. L. Cwk and E C. De Lucia, Am. J . Phys., 39,1433 (1971).
6. H.W. Kroto, Molecular Rotation Spectra, Wiley, New York, 1975.
7. W. J. Lafferty, D. R. Lide, and R. A. Toth, J. Chem. Phys., 43, 2063 (1965). There
appears to be a typographical m r for the assignment of the 11924.89 MHz line for
which F is listed as rather than $.
 Chapter 6
ENERGY-LEVEL STRUCTURE
AND WAVE FUNCTIONS
OF A RIGID ROTOR

Although no real molecule is a rigid rotor, the deviations from rigid rotor behavior are
often so small that nonrigid effects (vibration-rotation interaction) can be described
quite accurately by perturbation theory. Consequently, solutions to the rigid rotor
form the basis for more sophisticated treatments, and it is incumbent upon us to
examine the quantum mechanical description of a rotating body all of whose subpart.
are at fixed angles and distances from one another. This serves as the starting point
in discussing the rotational energy levels of polyatomic molecules and the radiative
transitions they undergo. The rigid rotor problem was first solved in the late 1920s
and early 1930s by several investigators[l-lo], but the original literature is difficult
to read because these authors could not avail themselves of what has now become
standard angular momentum machinery. After a brief review of the classical motion
of rigid bodies, we follow closely the expositions of King, Hainer, and Cross[ll-141
and of Van Winter[lS].

6.1 MOMENTS OF INERTIA: CLASSIFICATION OF RIGID ROTORS BY

TOP TYPES

Figure 6.1 shows two reference frames for describing the n particles of a rigid body,
a coordinate frame located in the laboratory (space-fixed frame), and a coordinate
frame located at the center of mass of the rigid body. The position vectors ri and Ti to
the ith particle (having mass mi) originate from the center of mass and the arbitrary
lab-frame origin, respectively. They are related by
254 6 ENERGY-LEVEL STRUCTURE AND WAVE FUNCTIONS OF A RIGID ROTOR

---* Y

-Y

X
FIGURE 6.1 Laboratory and center-of-mass coordinate frames for a rigid body consisting of
.
i = 1 , 2 , . . ,nparticles of mass mi having the position vector ri in the center-of-mass frame
and Fi in the lab frame. The vector R is the position vector to the center of mass in the lab
frame. The XYZ and xyz axes are chosen to be parallel to each other.

where R is the position vector from the arbitrary origin to the center of mass. The
total mass M is obtained by summing over all n particles

By definition, the center-of-mass position vectors satisfy the condition

and it follows that

6.1 MOMENTS OF INERTIA: CLASSIFICATION OF RIGID ROTORS BY TOP TYPES 255

that is,

The angular momentum of the ith particle about the origin of the lab frame is
given by Fi x pi, where pi = m J , = d/dt(miFi)is the linear momentum measured
in the same frame and iii = V + vi is the velocity. Hence the angular momentum with
respect to the lab frame of the rigid body is

Thus the total angular momentum in the lab frame J is the sum of the total angular
momentum in the center-of-mass frame J and the center-of-mass angular momentum
contribution R x P as if all the mass were concentrated at the center of mass. In
particular, let us choose to use a coordinate frame in which the center of mass is at
rest. Then

For a rigid rotor, all particles have the same angular velocity w about an instan-
taneous axis of rotation passing through the center of mass. The velocity of the ith
particle is related to w by

Substitution of Eq. (6.8) into (6.7) and use of the vector identity A x ( B x C ) =
B( A . C ) - C( A . B) gives

= 2
m i [ wri - ri(ri . w ) ]
i

whose Cartesian components are

256 6 ENERGY-LEVEL STRUCTURE AND WAVE FUNCTIONS OF A RIGID ROTOR

In matrix notation, Eq. (6.10) may be rewritten as

where 1 is called the inertia tensor (a second-rank Cartesian tensor) whose elements
are

where j, k represent x, y, z . Note that if the moment of inertia tensor is computed

with respect to the lab frame, we replace ri by 7, - R in Eq. (6.13) and obtain the
expression

In Eqs. (6.13) and (6.14) the diagonal elements I j j are known as moments of inertia
and the off-diagonal elements, as products of inertia. The moment of inertia tensor 1
in Eq. (6.11) must be regarded as acting on the vector w and not on the coordinate
system. The vectors J and w are two physically different vectors having different
dimensions. Unlike the rotation operator R, 1 has dimensions (g cm2) and does not
satisfy the conditions of a unitary transformation.
From Eq. (6.13) or (6.14) we see that the components of the inertia tensor 1 are
real and symmetric so that the tensor is self-adjoint or Hermitian. Consequently, there
exists a particular orientation of the rigid body (transformation of the coordinates) for
which the inertia tensor is a 3 x 3 diagonal matrix. The diagonal elements are called
the principal moments of inertia, and the axis system is termed the principal axis
6.1 MOMENTS OF INERTIA: CLASSIFICATIONOF RIGID ROTORS BY TOP TYPES 257

system. The three principal moments of inertia are designated I,,, lab,
Icc,which are
the roots of the determinantal equation

By convention, I,, is the smallest moment while Iccis the largest:

With respect to the principal axes, the components of J involve only the correspond-
ing components of w :

Molecules as rigid rotors are classified in terms of their three principal moments
according to the following types:
I,, = 0 , I b b = I, linear
L a < balb = I C c prolate symmetric top
L a = I b b < Ice oblate symmetric top
I,, = I b b = Ice spherical top
Z o < l a b < Icc asymmetric top

For example, C02 is a linear top, CH3C1 is a prolate symmetric top (football-like,
American style), CHC13 is an oblate symmetric top (frisbee-like), CC4 is a spherical
top, and CH2 Clz is an asymmetric top (which happens to be a "near" prolate top, i.e.,
I b b M I,).
In the center-of-mass frame, the rotational kinetic energy is

where use is made of the vector identity A - ( B x C) = B . ( C x A). In the principal-

axis system, Eq. (6.18) may be written with the help of Eq. (6.17)
258 6 ENERGY-LEVEL STRUCTURE AND WAVE FUNCTIONS OF A RIGID ROTOR

where A, B, and C are rotational constants, which in units of Hertz (Hz) are

Note that the inequality I,, 2 Ibb 5 ICcimplies that

Thus there are three mutually perpendicular directions for which the moment of
inertia is a maximum or a minimum. If 1~7''~
is plotted against the respective
principal axes, an ellipsoid is obtained, the so-called ellipsoid of inertia. If the
molecule has symmetry, the direction of one or more of the principal axes can easily
be found since axes of symmetry are always principal axes and a plane of symmetry
is always perpendicular to a principal axis. Molecules having a threefold or higher
axis of rotational symmetry are symmetric tops, while the presence of more than one
such axis causes the molecule to be a spherical top. Molecules without a threefold
or higher axis of rotation cannot be symmetric tops, although many molecules with
less symmetry are near-symmetric tops. It is interesting to note that prolate or near-
prolate tops are more common than the oblate or near-oblate variety because in most
molecules the light hydrogen atoms are located farthest from the center of mass.

6.2 CLASSICAL MOTION OF A FREE TOP

The Euler angles 0,$, x describe the position of the triad of orthogonal unit vectors
%,$,8 fixed in a rigid body and turning about a common origin 0 relative to the
triad of orthogonal unit vectors %,P, 2 fixed in space. The motion of the body is
determined when 0 , 4 , x are known as functions of the time t . This motion can also
be described by the angular velocity w ( t ),where, referred to the body-fixed frame,

We seek expressions for w,, w,, w, in terms of 0,4, x and their rates of change[l6].
In what follows we make extensive use of Eq. (3.43). The angular velocity 8 is
along the line of nodes; its components are

The angular velocity 8, is along the Z axis; its components are

4, = -4 sin 0 cos x
6.2 CLASSICAL MOTION OF A FREE TOP 259

$, = 4sin 8 sin x

Finally, the angular velocity is along the z axis; its components are

x, = x
Hence, collecting the components along each axis, we find

Let us choose the xyz axes to be the principal axes of the rigid body. Then the
rotational kinetic energy is given by

Let us further specialize this to a symmetric top for which I, = I, 5, I,. Then
substitution of Eq. (6.26) into Eq. (6.27) yields

Let us choose the Z axis of the space-fixed frame to be along the direction of the total
angular momentum L of the top. Then the components of L in the body-fixed frame
are

L, = I,w, = I, (-6 sin 9 cos x + 9 sin X)

However, we also have

L, = QZzL = -L sin 9 cos x

260 6 ENERGY-LEVELSTRUCTURE AND WAVE FUNCTIONS OF A RIGID ROTOR

Hence

I, (-6sin 8 cos x + 8 sin X) = - L sin 8 cos x (6.31)

I, (4sin 8 sin x + 6' cos X) = L sin 8 sin x (6.32)

Because the line of nodes is perpendicular to the Z axis, the component of L along
the line of nodes vanishes:
LN = 0 (6.34)

This component can be resolved into components along x and y as

= L, sin x + L, cos x

where we have replaced L, and L , by the expressions given in Eq. (6.29). Combining
Eqs. (6.34) and (6.35), we have the result

provided I, is not equal to zero.

On substitution of Eq. (6.36) into Eq. (6.31) or Eq. (6.32), we find the condition
for the rate of change of 4 with time

On substitution of Eq. (6.37) into Eq. (6.33),the condition for the rate of change of x
with time becomes

Equations (6,36)-(6.38) admit of a very simple interpretation. Equation (6.36)

shows that the angle between the figure akis of the top and the direction of L is
constant; Eq. (6.37) shows that the figure axis precesses about L at a constant angular
velocity L l I , , while Eq. (6.38) shows that at the same time the top rotates about its
figure axis at a constant angular velocity, ( L I I , - LII,) cos 8.
The preceding analysis leads to a picture of two cones rolling without slipping
about one another. At any instant L, w , and Z are coplanar. The angle between L
6.2 CLASSICAL MOTION OF A FREE TOP

space
~ody 1 -
cone
FIGURE 6.2 The classical motion of an
oblate symmetric top (I, > I, = I,,)is rep-
resented by a body cone in contact with a
fixed space cone with the space cone lying
inside the body cone. The body cone rolls
at a uniform rate without slipping about the
space cone. The line of contact is the in-
stantaneous axis of rotation, and the com-
mon vertex of the two cones is located at
the center of mass of the top.

(the Z axis) and the figure axis of the top (the z axis) is 8. Let a be the angle that w
makes with respect to 2 so that

Similarly,we have

Thus

We distinguish two cases: if I, > I,, then a > 8. The axis of rotation makes a
constant angle ( a - 8) with respect to L so that the axis of rotation is a generator of a
right circular cone, fixed in space, of angle ( a - 8). But the axis of rotation w makes
a constant angle a with the figure axis and traces out in the body a right circular cone
of angle a. Thus the motion is given by the rolling of a cone of angle a, which is
fixed in the body, on a cone of angle ( a - B), which is fixed in space. The smaller
cone touches the larger cone internally (see Figure 6.2). If I, < I,, it follows that
a < 8. A cone of angle a, fixed in the body, rolls on a cone of angle ( 8 - a ) fixed
in space. Then the smaller cone touches the larger cone externally (see Figure 6.3).
In each case the figure axis describes a cone of angle 8 about L. It is seen from
both figures that viewed from the molecular frame, L and w continuously change
position.
262 6 ENERGY-LEVEL STRUCTURE AND WAVE FUNCTIONS OF A RIGID ROTOR

w FIGURE 6.3 The classical motion of a pro-

late symmetric top (Iz < I, = I,) is repre-
sented by a body cone in contact with a fixed
space cone with the space cone lying outside
the body cone. The body cone rolls at a uni-
e form rate without slipping about the space
cone. The line of contact is the instantaneous
axis of rotation, and the common vertex of
H+H the two cones is located at the center of mass
H of the top.

Let us return to the general case of an asymmetric top with I,, 5 Ibb 5 I,,. From
the conservation of energy and angular momentum, we have

and

where E is the rotational kinetic energy and L~ is the square of the magnitude of
the angular momentum of the top. In what follows we outline a construction due to
Poinsot for describing the top motion.
Let the vector w be represented by a line segment OQ, drawn from the center of
mass 0 to the terminus Q. Let O P be a line drawn from 0 along the direction of L.
Draw the perpendicular QN of OQ on OP (see Figure 6.4). Then ON = w j? =
w .LIL = 2 EIL. Thus line segment ON is a constant (invariable line) and point N
is fixed during the motion of the top. Therefore, the plane through N perpendicular
to L is a fixed plane, called the invariable plane. The terminus of w moves on the
invariable plane.
To an observer riding in the top frame, the unit vectors %, y,%, taken along the
principal axes a , b, c, are fixed, but both w and L are changing direction in time. Let

FIGURE 6.4 The invariable line and the

invariable plane.
6.2 CLASSICAL MOTION OF A FREE TOP

FIGURE 6.5 Motion of the moment of inertia ellipsoid relative to the invariable plane.
Reprinted with permission from H. Goldstein, Classical Mechanics, 2nd ed., Addison-Wesley,
Reading, MA, 1953, p. 206, Fig. 5-4.

the terminus Q have the body-fixed coordinate ( x,y,2 ) . Then Eqs. (6.42) and (6.43)
become

and

These two equations are ellipsoids; the first of these two ellipsoids, called the Poinsot
ellipsoid, has the same axes as the moment of inertia ellipsoid but is scaled by the
factor (2 E)-'. To an observer moving with the top, the terminus of the angular
velocity w describes a curve that is the intersection of these two ellipsoids, fixed
in the body. As we have seen, the invariable plane touches the Poinsot ellipsoid
at the terminus of the angular velocity vector. Hence we can picture the motion of
the asymmetric top as being such that the moment of inertia ellipsoid rolls without
slipping on the invariable plane, with the center of the ellipsoid maintained at a
constant height above the plane (see Figure 6.5). The curve traced out by the point
of contact on the moment of inertia ellipse is called thepolhode; the curve traced out
on the invariable plane is called the herpolhode. In the special case of a symmetric
top, the moment of inertia ellipsoid is an ellipsoid of revolution so that the polhodes
degenerate to circles about the symmetry axis; that is, w moves on the surface of a
cone.
Let us return to Eqs. (6.42) and (6.43) but rewrite them in terms ofthe components
of L:
G +L; +L' -A
1 -
2 EI,, 2 EIbb 2 EI,
264 6 ENERGY-LEVELSTRUCTURE AND WAVE FUNCTIONS OF A RIGID ROTOR

FIGURE 6.6 Some possible paths of L on the kinetic energy ellipsoid fixed in the body axes.
Adapted from L. D. Landau and E. M. Lifschitz, Mechanics,2nd ed., Pergamon Press, New
York, 1969, p. 117, Fig. 51.

Equation (6.46) is the equation of an ellipsoidwith semiaxes ( 2 EI,,) 'I2, ( 2 EIbb) 'I2,
and ( 2 EI,) 'I2 ;Eq. (6.47) is the equation of a sphere of radius L. When L moves
relative to the principal axes of inertia, its terminus traces out the intersection of these
two surfaces. These paths change as the magnitude of L varies from ( 2 EI,,) 'I2 to
(2 EI,,) ' I 2for a given value of E. When L~ is only slightly greater than 2 EI,,, the
sphere intersects the ellipsoid in two small closed curves around the a axis. When L2
increases, these curves become larger, and for L2 = 2 EIbbthey become two plane
curves (ellipses) that intersect at the poles of the ellipsoid on the b axis. As L2 in-
creases further, two separate closed paths appear again, but now around the poles on
the c axis.
Because these paths are closed, the motion of L relative to the top axes is periodic.
However, there is a sharp difference in the nature of the paths near the various
poles. Around a and c the paths lie close together, but around b the paths make
great excursions before returning. This difference corresponds to a difference in
the stability of top rotations about its three principal axes. Rotation about c and a,
corresponding to the largest and the smallest moments of inertia, are stable in that
slight deviations produce a motion close to the original one. However, rotation
about b, corresponding to the intermediate moment of inertia, is unstable[l7]. This
6.2 CLASSICAL MOTION OF A FREE TOP

FIGURE 6.7 Rotational energy surfaces for two types of molecules: (a) rigid prolate sym-
metric top RE surface; (b) rigid asymmetric top RE surface. Contour lines represent angular
momentum trajectories; dotted lines, separatrix curves that separate different types of trajec-
tories; dashed lines, meridians of spherical energy surfaces that intersect the highest-energy
trajectories. The yz meridian is a tunneling path. Adapted from W. G. Harter and C. W. Pat-
terson, J. Chern. Phys., 80,4241 (1984), Fig. 1.

can be demonstrated very easily by spinning a book (taped shut) about its three axes.
A molecular analog may be the behavior of some polyatomic molecules in which,
with increasing internal excitation, vibration-rotation interaction alters sufficiently
the moments of inertia to cause large changes in the molecule's rotational motion[l8].
The time dependence of 8,4, x can be solved analytically for an asymmetric top, but
the solutions are quite complex (see Figure 6.6).
Another useful representation of top motion is to make a radial plot of the energy
as a function of the direction of L as viewed in the xyz body-fixed frame subject to
the constraint that L2 = L: + L; + L; is a constant. This is called a rotational energy
(RE) surface[l9]. Rotational energy surfaces for a rigid prolate symmetric top and
a rigid most asymmetric top (where the intermediate moment of inertia is halfway
between the other two) are shown in Figure 6.7. In Figure 6.7 the contour lines
correspond to angular momentum trajectories where each point on the RE surface
represents a direction L in the body-fixed frame. Recall that the direction of L is
fixed in space, but as the top rotates in the space-fmed frame, L moves from point
to point on the RE surface. At each point the radial height of the RE surface is
defined to be proportional to the rotational energy. Note that for L located on the
266 6 ENERGY-LEVEL STRUCTURE AND WAVE FUNCTIONS OF A RIGID ROTOR

equator of a symmetrictop RE surface the contours collapse to points, denoted by x's

and the precessional frequency of L vanishes. An equatorial line of fixed points is
the only part of the RE surface that remains for a linear rotor (diatomic molecule
or linear polyatomic molecule) in a 2 state. Thus the RE surface is most useful
in describing internal angular momentum trajectories and dynamics of nonlinear
polyatomic molecules.
In particular, the RE surface provides a good means of visualizing why the high-J
fine structure of nonrigid tops, particularly those of spherical tops with tetrahedral
or octahedral symmetry, show a clustering of the rovibrational sublevels into nearly
degenerate multiplets[l9-231. The total degeneracy of each cluster corresponds to
the number of distinct classical trajectories with the same energy on equivalent parts
of the RE surface. The classical frequency of L precession is roughly proportional
to the fine structure spacing between neighboring clusters; the superfine structure
splitting within a cluster is related to the rate at which L delocalizes or tunnels to
other equivalent trajectories[l9].

6.3 SYMMETRIC AND ASYMMETRIC TOP ENERGY LEVELS

The energy levels of a rigid rotor are found by solving the Schrodinger equation

where, according to Eq. (6.19), the Hamiltonian for the rotational motion of a free
rigid rotor is

In the oblate top limit I,, = Ibb< Icc,

where we make the identification of axes a ++ x, b ++ y, and c tt z (the figure axis).

The eigenfunctions I J K M) are given in Eq. (3.125), and the eigenvalues are

For each J level there are 2 J + 1 different K levels, but those with K # 0 are
doubly degenerate. For a prolate top (Iaa < Ibb = Ice), we redefine the body-fixed
coordinates a ++ z (figure axis), b ++ x, and c tt y, and the rotational Hamiltonian
takes the form
6.3 SYMMETRIC AND ASYMMETRIC TOP ENERGY LEVELS

FIGURE 6.8 Energy-level structure for (a) prolate top and (b) oblate top. All levels with
K # 0 are doubly degenerate.

The eigenvalues then become

which again show what is called K-type doubling for K # 0. Note that the
coefficient in front of K' changes signs in going from an oblate to a prolate top,
as it is negative in the former and positive in the latter. Figure 6.8 illustrates the
energy-level structure for these two types of symmetric tops.
In the general case of an asymmetric top (I,, # Iab# I,) the eigenfunctions are
linear combinations of symmetric top wave functions [see Eq. (3.124)l and we write
268 6 ENERGY-LEVEL STRUCTURE AND WAVE FUNCTIONS OF A RIGID ROTOR

TABLE 6.1 Group Characters for the Dz(V) Group

Group Operations
Symmetry Designationsa E C2(c) C2(b) C2(a)

T h e first column is the traditional group theoretic designation; the second column places a
subscript on B to denote the axis about which the rotation is made. We prefer the latter because
it is so easy to remember what it means.

On substituting Eq. (6.54) into (6.48), we obtain the expression

In Eqs. (6.54) and (6.55) summation over J and M is unnecessary since they are
good quantum numbers for an asymmetric top. After multiplication by ( J KIM1 and
integration, we find that the energies are the roots of the 2 J + 1 by 2 J + 1 secular
determinant

where

A nifty simplification can be made in the evaluation of Eq. (6.56) by using the
symmetry inherent in the moment of inertia ellipsoid to factor the secular determi-
nant into four subblocks (the Wang transformation[6]). The rotor Harniltonian [see

TABLE 6.2 Possible Identifications of Principal Axes a , b, c with Body-Fixed Coordinates

x , y, z (Both Frames Are Righthanded Coordinate Systems).

Representation
Body-Fixed Coordinates Ia 1 1 ~ IIIC

'Most natural in the limit of a prolate top.

bMostnatural for a top intermediate between the two limiting symmetric top cases.
CMostnatural in the limit of an oblate top.
6.3 SYMMETRIC AND ASYMMETRIC TOP ENERGY LEVELS 269

TABLE 6.3 Cartesian Coordinate and Euler Angle Transformations under Operation of the
D2 (V)Group
Cartesian Coordinate Euler Angle
Group Operationa Transformation Transformation

OFor identification of z, y, z with a, b, c, see Table 6.2.

Eq. (6.49)]has the property that a rotation by 180" about a, b, c leaves the Hamilto-
nian unchanged. Thus any asymmetric top belongs to the point group D2,also known
as the Viergruppe (V). The group operations are twofold rotations about a,b, c, de-
noted by Cz(a), Cz ( b), Cz (c),and the identity E. The group characters are shown
in Table 6.1.
We wish to discover how the IJ K M) transform under the operations of the Dz
(V) group. However, the (JK M) are. defined with respect to the x y z frame fixed in
the body where z is chosen to be along the figure axis, while the group operations
refer to the principal axes a,b, c. As shown in Table 6.2,there are three possible ways
in which the a,b, c may be identified with the x,y,z (where both frames refer to right-
handed coordinate systems). The K value in ( JK M) depends on the representation
that is used. For a near-prolate top representation I is used while for a near-oblate top
representation 111is most natural. For an asymmetric top intermediate between the
two symmetric top limits, representation 11brings out the asymmetry most clearly.
Once the top representation is selected, the effect of the operations of the D2 (V)
group may be found by determining the change in the Euler angles. For example,
C2(z) causes the transformation x -+ -x, y --+ -y, z -+ z.Then by inspection or
by using the direction cosine matrix elements QF, (#,O,X) that relate the Cartesian
vector components 7-p and r, [see Eq. (3.37)], it is readily verified that the C2(z)
270 6 ENERGY-LEVEL STRUCTURE AND WAVE FUNCTIONS OF A RIGID ROTOR

operation corresponds to 4 -t 4 , 8 -+ 0, x -+ n + X. The results of this analysis are

collected in Table 6.3. Thus we find that

= ( - I ) ~I J K M )

= ( - I ) ~ - IJ~ - K M )

and

-
C ~ ( XI J) K M ) = [2:; '
11 e ~ ~ ' * "d ~ , ( - 0)em'-x)

where we use Eq. (3.67) and the identity ein = -1.

Equations (6.58H6.60) suggest that we use, instead of 1 J K M ) , a symmetry basis
set constructed from linear combinations of I J fK M ) that transform the same way
under the operations of the Dz(V) group. We define for K # 0

where we may take s = 0 or 1, and for K = 0

IJOMs) = (-1)'IJOM) = IJOM) (6.62)

where we may take only s = 0 . Then we have

E IJKMs) = IJKMs)

C2(z)IJKMs) = (-llK IJKMs)

6.3 SYMMETRIC AND ASYMMETRIC TOP ENERGY LEVELS

TABLE 6.4 Asymmetric Top Symmetry Species

- -

Secular Symmetry Designation0

Determinant K s Jeven Jodd

E+ Even 0 A BZ
E- Even 1 B, A
o+ Odd 0 B, BY
0- Odd 1 By Bz
T h e identification of x, y, z with a, b , c depends on the top representation chosen (see
Table 6.2).

Hence the ( 2 J + 1) x ( 2 J + 1) secular determinant factors are divided into four

blocks; the eigenvalues and eigenfunctions of each block belong to one of the
symmetry species of the Dz (V) group. We list in Table 6.4 the results for the various
values of J , K, and s.

Note that Eq. (6.58), (6.59), and (6.60) are valid for integer and half-
integer J and K, whereas Eq. (6.63) holds true only for integral J and K.

We now turn to the calculation of the matrix elements of 7 i R appearing in the

secular determinant subblocks. Recalling that the commutation rules for J referred
to the molecular frame are anomalous, we apply Eq. (3.49) and obtain

We then use matrix multiplication to evaluate the desired elements, for instance
272 6 ENERGY-LEVEL STRUCTURE AND WAVE FUNCTIONS OF A RIGID ROTOR

We find for the diagonal elements

( J K M I J: I J K M ) = (JKMI J; I J K M ) = [ J ( J + 1) - K 2 ]

( J K M I 5: ( J K M )= K 2 (6.66)

and for the off-diagonal elements

( J K M I J: IJK f 2 M )

= - ( J K M I 5; ( J K It 2 M )

To be specific, let us choose to work in the prolate symmetric top basis set (repre-
sentation I in Table 6.2) since most molecules are closer to the prolate top limit. We
have

and

Note that only even or odd K levels interact with one another.
We illustrate the foregoing by working out explicit forms for the energy levels
E( A, B, C) of an asymmetric rotor for some of the lowest J values. We designate
the levels by JK-,K , , where K-1 is the value of I K I the top would approach in the
limit of a prolate top (A > B = C) and K1 is the value of 1 K 1 in the limit of an
oblate top (A = B > C). Figure 6.8 shows that K-1 runs from J to 0 and K1
from 0 to J with decreasing energy. Hence the possible energy levels are JJo, J J 1 ,
JJ-1,1, J3-1,2, and so on. It follows that we can label the energy levels instead by
the index 7, where

Then 7 runs from J to -J in unit steps in order of decreasing energy. However, for
symmetry purposes it is more convenient to use the K-1 K1 label. Note that K-l
is associated with a rotation about the a axis and K 1 , the c axis. Then the D2 (V)
6.3 SYMMETRIC AND ASYMMETRIC TOP ENERGY LEVELS 273

group designations may be identified with the (even, odd) parity of the K-1 Kl pair
as follows (see Table 6.1):

The symmetry about the b axis may be obtained from the parity of the sum K-l + K1.
For J = 0 there is only one level, &, and its energy is seen at once to be 0.
For J = 1 there are three levels 110, 111, and lol . The hypothetical loolevel of
A symmetry comes from the E- secular determinant subblock, which can have no
entries for K = 0 and s = 1. The energy of the 1lo level of B, symnxuy is calculated
from the 1 x 1 O+ secular determinant

from which we find

Similarly, the energy of the 111 level (Bb symmetry) is calculated from the 1 x 1
0- secular determinant

which has the solution

274 6 ENERGY-LEVEL STRUCTURE AND WAVE FUNCTIONS OF A RIGID ROTOR

and the lol level (B, symmetry) from the 1 x 1 E+ secular determinant

whose solution is

For J = 2 there are five levels, denoted as 220, 221, 211, 212, and 202. The
levels 221, 211, and 212 each belong to a different symmetry designation of the
Dz (V) group, namely, B,, Bb, and B,, respectively. Hence the energy for each
of these levels involves only the evaluation of a 1 x 1 secular determinant, as in the
J = 1 case. It is easily shown that the energy is 4 A + B + C for the 221 level,
A + 4 B + C for the 211 level, and A + B + 4 C for the 212 level. However, the 220
and 202 levels both are of A symmetry, and their energies are the roots of the 2 x 2
E+ secular determinant:

(22 MO I XR 120 MO) (22 MO

which has the explicit form

T h e r o o t s a r e 2 ~ + 2 ~ + 2 C + 2 [ ( ~ -+(A-c)(A-
C)~ B)]'/' f 0 r t h e 2 ~level
~
and 2A + 2 B + 2 C - 2 [ ( B - c)' + ( A - C)(A - B ) ] ~ /for ~ the h2level.
The energy assignments are made by identifying 220 with the higher energy and 202
with the lower energy in the prolate top limit A > B = C (see Figure 6.8). Explicit
algebraic expressions up to J = 3 are given in Table 6.5, and we sketch in Figure 6.9
how the energies of an asymmetricrotor vary between its prolate and oblate top limits
for these levels. We note that no two levels cross, a result that may be proved to hold
for all J[5, 101.
As first pointed out by Ray[lO], it is advantageous to introduce an asymmetry
parameter ti, defined by
2B-A-C
ti, =
A-C
in calculating the energy level structure of an asymmetric top. Note that K ranges
between -1 for a prolate top (B = C) and +1 for an oblate top (A = B). When
K = 0 , B = ;(A + C ) , and as Figure 6.9 shows, the energy levels are symmetrically
distributed about their mean value (middle level).
The motivation behind this choice for K is that the energy levels of any asymmetric
top can be parameterized in terms of this single variable. This may be shown as
6.3 SYMMETRIC AND ASYMMETRIC TOP ENERGY LEVELS 275

TABLE 6.5 Asymmetric Top Energy Levels for J 3

Symmetry

a J K (prolate) K( oblate)

follows. Let a and p be scalar quantities. With a simple change of variables the
Hamiltonian becomes

and the corresponding eigenenergies are given by

E ( a A + P , aB + p, aC + p) = a E ( A ,B , C ) + p J ( J + 1) (6.82)

In particular, let us choose

-2
A-C
276 6 ENERGY-LEVEL STRUCTURE AND WAVE FUNCTIONS OF A RIGID ROTOR

FIGURE 6.9 Energy levels of an asymmetric top as a function of the asymmetry parameter n,
calculated for A = 5 and C = 3 .

and
p= -(A + C ) / ( A - C)
so that
6.4 NONRIGID BEHAVIOR: THE VAN VLECK TRANSFORMATION 277

Substitution of Eqs. (6.83X6.85) into Eq. (6.82) yields the result

2
E ( ~ , I E , - 1 )= EJ(K) = - E(A, B, C) - -
A J( J + 1)
+ (6.86)
A-C A-C
which may be rearranged to read

Hence the energy levels of an asymmetric top, labeled by the index r [see Eq.(6.70)],
may be determined from knowledge of the universal energy function EJ( n) of a rigid
rotor with angular momentum J having the three rotational constants A = 1, B = rc,
C=-1.
Values of EJT(lc) for all levels with J 5 12 and for values of rc from 0 to 1 in
steps of 0 .O1 have been computed by Turner, Hicks, and Reitwiesner[24] and are
listed in Appendix IV of the monograph Microwave Spectroscopy by Townes and
Schawlow[25]. Energy levels tables with J 2 40 and for values of rc from -1
to 0 in steps of 0.1 have been compiled by Erlandsson[26] and are reproduced in
Appendix IV of the monograph Molecular Vib-Rotors by Allen and Cross[27]. Only
positive or negative values of rc need be tabulated since it may be shown[lO] that (see
Figure 6.9)

The energy levels of a rigid asymmetric rotor have also been calculated by semiclassi-
cal methods with the use of a uniform semiclassical quantum condition[l9,28]. The
results give the correct number of rotational levels for a given J and are remarkably
accurate.

6.4 NONRIGID BEHAVIOR: THE VAN VLECK TRANSFORMATION

Although the rotational motion of a molecule closely resembles that of a rigid body,
real molecules vibrate (many times) as they rotate, causing what is called vibration-
rotation interaction. This interaction has the effect of causing the intramolecularbond
angles and bond distances to change as the molecule rotates, thus coupling nearby
vibrational levels within the same electronic state. Also, as the molecule vibrates
and rotates there is the possibility of interaction between different electronic states
as the electronic and nuclear motions are not exactly separable (failure of the Born-
Oppenheimer approximation). All these effects cause the rigid rotor model to break
down. We seek in this section a way of accounting for nonrigid behavior.
As is well known, a polyatomic molecule with N nuclei has 3 N - 6 vibrational
degrees of freedom (3 N -5 if all the nuclei lie on a straight line). For small vibrations
about the equilibrium configuration, it is always possible to choose an orthogonal set
of internal coordinates {Qk) = Ql , Q2,. .. , Q3N-6rSO that the system undergoes
simple harmonic motion at some vibrational frequency vk along the vibrational
278 6 ENERGY-LEVELSTRUCTURE AND WAVE FUNCTIONS OF A RIGID ROTOR

coordinate Q k ,which is called a normal mode[29]. Under the conditions where we

ignore higher-order anharmonic effects, the vibrational motion of the system along
each of its normal modes is independent. Then the vibrational energy is a sum of the
i)
energies hvk(vt + of each normal mode with vibrational quantum number vk, and
the vibrational wave function is a product of the individual wave functions $, ( Q k )
of each normal mode (which are harmonic oscillator wave functions), that is,

and

Note that we let the index V denote the collection of vibrational quantum num-
bers { v k ) .
We suppose as a good startingpoint that the Hamiltonian for our nonrigid molecule
can be separated into two parts:

where H v describes a vibrating oscillator and 3CR a rigid rotor. Then we introduce
the wave functions

as a convenient basis set where

Here R denotes the collection of rotational quantum numbers T, J , K, M nec-

essary to specify the rotational wave function with rotational energy ER. We could
also denote the different electronic states, but for simplicity we will consider only
the interactions between the levels of one electronic state and, therefore, drop this
designation in what follows.
Let us concentrate our attention on the rotational level structure within one
specific nondegenerate[30] vibrational state of a nondegenerate[31] electronic state
by considering only those levels with the same total angular momentum J. There
are in general 2 J + 1 such levels with the same space-fixed projection of J, and we
refer to these levels as the IV;J K M ) block. Once again it is convenient to use the
symmetric top wave functions 1 J K M)as a basis set because they span the ( 2 J + 1)
b y ( 2 J + 1) spaceofR.
Nonrigid behavior makes H R vary with nuclear motion, causing the IV;J K M )
block to be perturbed. However, a perturbation, like beauty, is in the eye of the
beholder! Occasionally, the perturbation can be ascribed to a small number of
6.4 NONRIGID BEHAVIOR: THE VAN VLECK TRANSFORMATION 279

interacting levels belonging to different IV; J K M )blocks. Then the effect of the
perturbation can be treated by diagonalizing the matrix of interacting levels (all
having the same total angular momentum). This procedure, which amounts to all
orders of perturbation theory, requires that the location and symmetry designations
of the perturbing levels be known. The result is that the energy-level scheme deviates
from the regular pattern of the other unperturbed levels. Some choose to regard this
phenomenon as a nuisance; others regard this as an opportunity to have a "window"
on often otherwise hidden levels.
More generally, the perturbations are small in magnitude but large in number.
Moreover, they involve many interacting levels from distant blocks whose energy
positions are often poorly known (if at all). In this latter case, an exact treatment
(diagonalization of a semi-infinite number of interacting levels) is impractical, if not
impossible, and resort must be made to some approximate procedure.
The Van Vleck transformation, also called a contact transformation, is such an
approximation that treats the interaction between the ( 2 J + 1) by (2 J + 1) blocks
by incorporating their effects on the IV; J K M) block of interest[32-351. When this
procedure is applicable, there is no longer the need to diagonalize the supermatrix of
interacting blocks. In what follows, we outline the Van Vleck transformation, which
bears a strong resemblance to second-order perturbation theory, and illustrate its use
by consideringthe effects of centrifugal distortion on the energy levels of a symmetric
top.
Let the molecular Hamiltonian be written as

where X denotes the order of the perturbation. The parameter X is a bookkeeping

device; when we are finished, we will set X = 1. We assume that the matrix elements
of 7fo lie entirely within the diagonal IV; J K M )blocks while the matrix elements
of Xlmay lie inside the outside diagonal blocks. The idea behind the Van Vleck
transformation is to apply a unitary transformation such that it removes to first order
in X the matrix elements of 7 f 1 that lie outside the diagonal blocks while at the same
time preserves to first order in X the matrix elements of X1that lie inside the diagonal
blocks. Then the remaining matrix elements of X1that lie outside the diagonal
blocks are second order in X and contribute to the energy only to fourth order in A.
Thus through third order in A, the transformed matrix of X,which we call G,consists
of diagonal blocks of the same dimensions as found in Xo.If desired, G may be
subject to a second Van Vleck transformation, and so on.
We turn to the steps necessary to cany out this procedure. Any unitary transfor-
mation T can be expressed as

where S is a Hermitian matrix. Moreover, we can expand T in powers of X and obtain

the series
280 6 ENERGY-LEVEL STRUCTURE AND WAVE FUNCTIONS OF A RIGID ROTOR

where I is the identity matrix. Hence the transformed matrix B can be written as
P = T-'XT

= P o + x g 1 + x 2 9 2 + x3g3+ ...

Equating like powers of A, we find

Po = 310

Let the levels of the IV; J K M ) block of interest be indexed by j, k, C, and

so on, and the levels of other blocks by a, /3, 7, and so forth. The Van Vleck
transformation must meet two requirements. The first requirement is that (j18 1 k)
equals (j)'Hl )k).Hence we demand that

= i ( E i - E k ) (j[S[k) (6.99)

which is satisfied in general by setting

(A s lk) = 0 (6.100)

The second requirement is that (jI 8 1 la) vanishes. We have

0 = (jl Xl la) + i (jlXoS - SXo la)

= (jlXI 1. ) s
+ i ( E j - E d (3'1 (a) (6.101)

which is satisfied in general by setting

6.4 NONRIGID BEHAVIOR: THE VAN VLECK TRANSFORMATION 281

Equations (6.100) and (6.102) define the nature of S that generates the unitary
(contact) transformation T. Applying these relations to Eq. (6.98), we obtain the
matrix elements of the transformed Hamiltonian within the same ( 2 J+ 1) by (2 J + 1)
block I V; J K M ) :

(il Xl la) ( 4 1P) (PI X I lk)

( Ej - Ed ( E k - Ep)

-I ce
(A XI I4 ( 4 xi le) (el
(Ee - Ea)(Ej - E,)
jk)
I (6.103)

In most applicationsit suffices to retain only the second-ordercorrection. A further

simplification results if the energy separation between interacting blocks is much
larger than the energy separation between levels in the same block so that the levels
within the same block are quasi-degenerate compared to level separations between
blocks. Then to good approximation the correction matrix X:.:) = (j1 Gz lk) may be
written as

where EvJ - EvtJ is the energy difference between the centers of the unperturbed
blocks I V; J K M ) and I V'; J K M ) . Equation (6.104) closely resembles the form
of second-order perturbation theory. We can factor 31;:) as fjk( J , K)Ajk, where
fjk( J, K) contains the dependence of 311:) on the quantum numbers J and K
and Ajk is a term independent of these quantum numbers. As a result, the Van Vleck
transformation introduces new terms into the effective molecular Hamiltonian to ac-
count for the perturbation Xl on the levels of the IV; J K M) block. The "con-
stants" Ajk are then found empirically by comparison with spectroscopic data.
An example will help to make the above less abstract. Consider how centrifugal
distortion affects the energy levels of a prolate symmetric top. The rotational
Hamiltonian is, according to Eq. (6.52), given by XR = C J ~+ (A - C) J:,
where for a rigid body A and C are rotational constants inversely proportional to
282 6 ENERGY-LEVEL STRUCTURE AND WAVE FUNCTIONS OF A RIGID ROTOR

twice the moment of inertia about the a and c axes, respectively, where a is the
figure axis (the z axis in the molecule-fixed frame). When the molecule vibrates,
3-IR can be written in the same form; however, A and C are no longer constants but
depend on the positions and conjugate momenta of the nuclei[36]. Hence we must
regard A and C as operators in the space of molecular vibrations. We then identify
the expectation value (V I A I V) with Av and (V I C I V) with Cv. Application of the
Van Vleck transformation via Eq. (6.104) is particularly simple in this case because
for the prolate symmetric top rotational Hamiltonian there are no off-diagonal matrix
elements in K. We find for the second-order centrifugal distortion correction
(V; JKM13tl IV'; J'K'M') (V'; J'K1M'I3tl IV; J K M )
742)
VJKiVJK = C EVJ- Ev' J

x C ( J K M IJ~ IJ'K'M') (J'K'M'I J~ ~ K M )

J'K'M'

(Vl C lV') (V'l A - CIV) + (VIA- CIV') (V'I C J V )

+C V' EVJ - EV'J

x C
J'K'M'
( J K M IJ: IJ'K'M') (J'K'M'I J: IJKM) (6.105)

This is essentially equivalent to introducing a centrifugal distortion Hamiltonian for

the level V of the form

where we identify[37]
6.5 LINE STRENGTH FACTORS 283

D ~ = -( VC
I A - CIV') (V'IA- CIV)
Ev - Ev)
V'

as three phenomenological centrifugal distortion constants.

The rotational energy level structure of a vibrating prolate symmetric top then
becomes

which is a result first obtained by Slawsky and Dennison[38]. The correction terms
for centrifugal distortion depend on only even powers of the angular momentum.
This is so because the distortion effects do not depend on the direction of rotation
about any axis. The term D J KJ ( J + 1) K 2 has the effect that the K stack levels
(levels of different J but the same K) will no longer coincide exactly when shifted
by an appropriate amount.
Centrifugal distortion corrections for asymmetric rotors have been worked out in
a similar manner, only here the centrifugal distortion Hamiltonian is of the form[39]

where a, p, 7 , S may each represent either x, y, or z. The full vibration-rotation

problem is very complex and treated in detail elsewhere[36,40-421. By use of the
Van Vleck transformation it is possible to describe quite complicated molecular
interactions and reduce them to a small number of phenomenological molecular
constants.

6.5 LlNE STRENGTH FACTORS

The line strength S ( J ' ; J " ) of the electric-dipole-allowed molecular transition con-
necting the upper level characterized by the total angular momentum J' to the lower
level J" is defined in the same manner as in atomic spectroscopy[43], namely,
284 6 ENERGY-LEVEL STRUCTURE AND WAVE FUNCTIONS OF A RIGID ROTOR

where the summation is over all magnetic sublevels of both levels. It may be
shown[43] that the line strength factor may be related to the Einstein coefficient for
spontaneous emission by the relation

where v is the frequency of the J' + J" transition, h is Planck's constant, and c is the
speed of light. Thus the central issue in comparing molecular line intensities is the
calculation of the line strength factor S ( J ' ; J " ) ,which involves the evaluation of the
matrix elements of the electric dipole operator p between the wave functions + j t M t
and + p M u .
The operator p consists of a sum over all charges, nuclear and electronic,weighted
by their position vectors measured from a common origin. For an electric dipole
transition the fundamental interaction is p E, where E is the electric vector of
the radiation field. The line strength refers to what is called "natural excitation"
whereby the excitation is isotropic as opposed to directional[43]. Then we may
rewrite Eq. (6.111) as

because each space-fixed component of p contributes equally to the line strength

factor. It is more convenient to refer p in the space-fixed frame ( F = X, Y, 2 ) to the
molecule-fixed frame ( g = x,y , z ) . According to Eq. (5.9)

and for pz

Hence the line strength factor is given by

We begin by specializing our consideration to a symmetric top. For the transition

J'K' ++ J1'K" there is only one component of p in the molecule-fixed frame
which can contribute, namely, the component q for which K' - K" = q. We set the
magnitude of the square of the dipole moment to unity so that our factors refer simply
6.5 LINE STRENGTH FACTORS 285

to relative rotational line strengths independent of the nature of the molecule. We are
then able to carry out the evaluation of Eq. (6.1 16) in a straightforward manner:

S (J'K'; J I I K I I )

=3 c
M1,M"
[ ( 2J' + 1 ) (2 J" + 1 ) ]
8 n2 D$,,
JN
c i ~
I*

Using Eq.(2.33),we obtain at once the sum rules:

s (J'K';J'K') = 2 J" +1 (6.118 )

J1,K'

and

The S (J'K';J" K") are called Honl-London factors, named after the two work-
ers who first derived their value but by quite different means involving "old" quantum
theory[44]. We introduce the standard spectroscopic notation:
A J = jl- j"=-l P branch
A J = J1- J"=O Q branch
A J = J I - jl'=+l R branch
286 6 ENERGY-LEVEL STRUCTURE AND WAVE FUNCTIONS OF A RIGID ROTOR

TABLE 6.6 The Honl-London Factors"

Transition s (J'K'; J"K")

"Note that the P ( 0 ) and Q ( 0 ) transitions do not exist.

Table 6.6 lists the Honl-London factors for the nine possible types of symmetric top
transitions.
Often we are interested in the form of the rotational line strengths when J" >>
K". Table 6.6 shows that for a A K = 0 transition (with the transition dipole parallel
to the figure axis of the top) P (J") 21 J", Q (J") N 0, and R (J") N (J" + I ) ,
that is, the P and R branch lines are of comparable strength while the lines of the
Q branch have appreciable strength only for the lowest J1'values. For a A K = f1
transition (with the transition dipole perpendicular to the figure axis of the top)
P(J") 21 (J"- 1 ) / 4 , ~ ( ~ N " )(2JI1+ 1 ) / 4 , a n d ~ ( J " )21 (J1I+2)/4,that
is, the lines of the P and R branches for a given J" value have comparable strength
but the lines of the Q branch for the same J" value are about twice as strong.
6.5 LINE STRENGTH FACTORS 287

For an asymmetric top we may write the rotational wave function belonging to the
energy level E J , ( K )as a linear combination of symmetric top wave functions:

and

where the coefficients a,K are real and satisfy the condition

line strength of an asymmetric top transition T'J' -

On subsititution of Eqs. (6.120)and (6.121)into (6.116). we obtain for the rotational
T" 5''

Once again the S (dJ'; T" J") factors may be shown to satisfy the sum rules:

and

Equations (6.124)and (6.125)are a consequenceof Eqs. (6.118). (6.119),and (6.122).

Thus the calculationof rotational line strengthsfor asymmetrictop transitions reduces
to the problem of finding the values of the coefficients a + t p and a+Kt. As shown in
Section 6.3, this is equivalent to finding the elements of the columns (vectors) of the
unitary matrix that diagonalizes the asymmetric top Hamiltonian.
To the P, Q, R branch notation used to designate A J = - 1 , O ,+1 it is convenient
to add as subscripts on the right the values of A K-, and A K+l to distinguish
288 6 ENERGY-LEVEL STRUCTURE AND WAVE FUNCTIONS OF A RIGID ROTOR

different subbranches[45]. For example, P3,4indicates A J = - 1, A K-l = 3 ,

and A K+1 = -2. In pure rotational or vibrational-rotational spectra where the top
geometry almost does not change between the initial and final states, the principal
subbranches are those for which A K-l and A K+1 change by 0 or f 1 and are
thus allowed in both the prolate and oblate top limits. The next most important
subbranches are those for which one of the indices changes by 0 or f1 but the change
in the other index is greater than one. Such transitions are allowed in one of the two
top limits. The least important are those transitions forbidden in both top limits.
For an asymmetric top there are in general three nonvanishing components of the
transition dipole, pa, pb, and p,, which transform as B,, Bb, and B,, respectively,
under the symmetry operations of the Dz (V) group. For the matrix element
(T'J'I p Id1Jl1)not to vanish it is necessary that the product of the three factors be
symmetric, that is, belong to A, under all the group operations. Making use of this

pa #0 A - -
fact, we obtain the following selection rules for asymmetric top transitions:
B, Bb B, A K , = O , f 2 ,...

p,#O A-B, B,- Bb A K a = f l , f 3 ,...

e e w o e eo-oo A K c = 0 , f 2 ,...

We have completed the calculation of the rotational line strength factors for a rigid
body undergoing an electric dipole transition. We take up the more general problem
of finding the line strengths of molecular transitions. A full explication of this topic
is beyond our scope but the following overview is intended as an outline of the
most important features. Once again we begin by invoking the Bom-Oppenheimer
approximation[46] and writing the total wave function $ J J M as a product of an
electronic, vibrational, and rotational part:

Here E, V, and R stand for the complete set of electronic, vibrational, and rotational
quantum numbers, respectively. Each $ J E ,$Jv,$JRform a complete orthonormal basis
set spanning their respective spaces. Hence when more accurate calculations are
required, we can expand J M in terms of these functions
6.5 LINE STRENGTH FACTORS 289

For atomic transitions one chooses the origin to be the atomic nucleus; then p
depends almost always on the electronic coordinates alone (disregarding 7-ray tran-
sitions). For molecular transitions, p is always a sum of an electronic part pe and a
nuclear part pn:

~1 I+JNMII)
(+JIM~I R II+EN$V~RM)
= ( + E ~ + v ~pe +(+E~Jv~
pnR~I I+E~Jv~R~~
(6.128)
Because pn is independent of the electronic coordinates, the second integral in
Eq. (6.128) vanishes for E' # E" since ($,yl$,yt) = bEIEn. Hence for electronic
transitions in molecules, only the electronic part of the electric dipole operator needs
to be considered, while for transitions within the same electronic state, the full dipole
moment operator contributes. Indeed, in the latter case (E' = E"), the operator is
the expectation value of the dipole moment in the electronic state, (p),.
Equation (6.128) readily factors into the product of two terms so that

SF
is the strength of the E'V' -
where J,is the rotational line strength factor we discussed previously, and S:y,/
E1'V" vibronic band.
First let us discuss the form of s::, for E' # El'. Here we have

where
DElE" = (Ell peIE")

is the transition dipole moment connecting the two electronic states and is a weak
function of the nuclear coordinates in that the electronic wave functions and + E ~ ~
depend only parametrically on them. Consequently, DErE~,may be treated as a
constant (Condon appmximation[47]) and Eq. (6.130) then becomes [see Eq. (6.90)]:

where the square of the vibrational overlap integral

is called the Franck-Condon factor for the normal mode k. Because the vibrational
wave functions are oscillatory functions of nuclear displacement,the Franck-Condon
factors vary markedly with vibrational quantum number depending on whether the
overlap of the vibrational wave function is in or out of phase. This quantum
mechanical effect has been termed "internal diffraction" by Condon, who remarked:
290 6 ENERGY-LEVEL STRUCTURE AND WAVE FUNCTIONS OF A RIGID ROTOR

"internal diffraction is just as real and forceful a proof for the wave nature of
nuclear motion as any of the basic external diffraction experiments ... which also
are determined by a phase relation between initial and final wave functions[47]." For
more accurate work the variation of DEIEu with nuclear displacement must also be
taken into account[48,49].
Consider next the electric dipole transitions connecting V'R' and V"R1' within
the same electronic state E. The electronic dipole moment operator ( p j E depends
both on electronic and nuclear coordinates suggesting that we consider an expansion
of dipole moment relative to one of the molecule-fixed axes g = x, y , z about the
equilibrium configuration

Then the V' - V1' band strength factor becomes

Note that terms involving (VIIV1')vanish unless V' = V", whereas terms involving
(V'I Qk IVI1) vanish unless V' # V" since (VI Qk IV) is an odd function in the
I
normal coordinate Q k . Then to lowest order, the term (v' I ((P,) E ) , IV") givesl2
rise to a pure rotation spectrum (V' = V") that is seen to require a nonvanishing
permanent dipole moment along at least one of the principal axes g = a, b, c. And to
lowest order, the term Ixk ( a ( l r g ) E / a ~ k )(V'l l2
o Qk lV1') gives rise to a vibration-
rotation spectrum (V' # V) that is seen to require a changing dipole moment along
some normal coordinate Qk. If the vibrational wave functions are represented as
products of harmonic oscillator wave functions, as in Eq. (6.90), then only A vk = f1
transitions are allowed in which all the vibrational quantum numbers but one stay
fixed. This result is a consequence of the fact that

has nonzero values for only v; - vg = f1. Such transitions are calledfundamental
transitions because they are the most intense features of the vibration-rotation
spectrum. To obtain overtone transitions in which A vk > 1 or combination bands
in which the quantum numbers associated with more than one normal mode change,
it is necessary to consider both higher order terms in the normal mode expansion
of ( , u ~ (electrical
)~ anharmonicity) and coupling between normal modes caused by
mechanical anharmonicity.
NOTESANDREFERENCES 291

NOTES AND REFERENCES

1. F. Reiche and H. Rademacher, Z. Phys., 39,444 (1926); H. Rademacher and F. Reiche,

Z. Phys., 41,453 (1927).
2. D. M. Dennison, Phys. Rev., 28,318 (1926); Rev. Mod. Phys., 3,280 (1931).
3. R. L. Kronig and I. I. Rabi, Phys. Rev., 29,262 (1927).
4. E. E. Witmer, Proc. Natl. Acad. Sci. (USA), 13,60 (1927); Phys. Rev., 74, 1247, 1250
(1948).
5. H. A. Kramers and G. P. Ittman, 2.Phys., 53, 553 (1929); 58, 217 (1929); 60, 663
(1930).
6. S. C. Wang, Phys. Rev., 34,243 (1929).
7. 0. Klein, Z. Phys., 58,730 (1929).
8. H. B. G. Casimir, Rotation of a Rigid Body in Quantum Mechanics (J. B. Wolter's, The
Hague, 1931).
9. H. H. Nielsen, Phys. Rev., 38, 1432 (1931).
lo. B. S. Ray, Z. Phys., 78,74 (1932).
11. G. W. King, R. M. Hainer, and P. C. Cross, J. Chem. Phys., 11,27 (1943).
12. P. C. Cross, R. M. Hainer, and G. W. King, J. Chem. Phys., 12,210 (1944).
13. G. W. King, J. Chem. Phys., 15,820 (1947).
14. R. M. Hainer, P. C. Cross, and G. W. King, J. Chem. Phys., 17,826 (1949).
15. C. Van Winter, Physica, 20,274 (1954).
16. My favorite reference is J. L. Synge and B. A. Griffith, Principles of Mechanics
(McGraw-Hill, New York, 1959), pp. 374-383.
17. An asymmetric top set in motion with angular momentum pointing near the b axis will
spin for awhile and then precipitously flop almost completely over spinning about the
opposite axis for awhile before repeating this cycle. See W. G. Harter and C. C. Kim.
Am. J. Phys., 44, 1080 (1976), who describe an apparatus for demonstrating this
behavior as well as show how to calculate the time between reversals (top flops).
If the top system has dissipative energy losses (as it would if filled with a fluid),
then the only stable rotation is about the axis associated with the largest moment of
inertia. On a molecular scale, such losses may be caused by extensive vibration-rotation
coupling in a nonrigid molecule for which the time scale for recurrence exceeds the
relaxation time (caused by collision or radiative emission, for example) so that the top
flop is irreversible. For more information on the treatment of radiationless transitions
in isolated molecules, see M. Bixon and J. Jortner, J. Chem. Phys., 48, 715 (1968);
D. P. Chock, J. Jortner, and S. A. Rice, J. Chem. Phys., 49, 610 (1968); P. Avouris,
W. M. Gelbart, and M. A. El-Sayed, Chem. Rev., 77,793 (1977); and K. F. Freed, Acc.
Chem. Res., 11,74 (1978).
18. G. M. Nathanson and G. M. McClelland, J. Chem. Phys., 81,629 (1984); Chem. Phys.
Lett., ll4,441 (1985)
19. W. G. Harter and C. W. Patterson, J. Chem. Phys., 80,4241 (1984).
20. A. J. Domey and J. K. G. Watson, J. Molec. Spectrosc., 42, 135 (1972).
292 6 ENERGY-LEVEL STRUCTURE AND WAVE FUNCTIONS OF A RIGID ROTOR

21. K. Fox, H. W. Galbraith, B. J. Krohn, and J. D. Louck, Phys. Rev. A, 15,1363 (1977).
22. W. G. Harter and C. W. Patterson, J. Chem. Phys., 66,4872,4886 (1977).
23. W. G. Harter, C. W. Patterson, and F. J. da Paixao, Rev. Mod. Phys., 50,37 (1978).
24. T. E. Turner, B. L. Hicks, and G. Reitwiesner, Ballistics Research Laboratories Reprt
No. 878, Ballistics Research Laboratories, Aberdeen, Maryland (1953).
25. C. H. Townes and A. L. Schawlow,Microwave Spectroscopy (McGraw-Hill, New York,
1955).
26. G. Erlandsson,Arkiv Fysik, 10,65 (1956).
27. H. C. Allen, Jr. and P. C. Cross, Molecular Vib-Rotors (Wiley, New York, 1963).
28. S. M. Colwell, N. C. Handy, and W. H. Miller, J. Chem. Phys., 68,745 (1978).
29. G. Herzberg, Infrared and Raman Spectra of Polyatomic Molecules (Van Nostrand,
Princeton, NJ, 1945).
30. Polyatomic molecules may have vibrations whose motions behave as internal angular
momenta. An example is the doubly degenerate bending mode of a linear triatomic
molecule. Here two mutually perpendicular bends with a phase difference of fn/2
are equivalent to a pure rotation of the bent molecule about the longitudinal axis. It
may be shown that vibrational angular momenta occur only for degenerate vibrational
modes. Furthermore, the degenerate vibrational modes are easily identified from the
point group symmetry of the molecule (see reference 1291). In particular, if a molecule
has a threefold or higher axis of rotation as a symmetry element, there are degenerate
vibrational modes present.
3 1. Degenerate electronic states in polyatomic molecules introduce a number of interesting
complications. See G. Herzberg, Electronic Spectra and Electronic Structure of
Polyatomic Molecules (Van Nostrand, Princeton, NJ, 1966) for more information. In
brief, for a degenerate electronic state of a linear polyatomic molecule, the interaction
of the electronic and vibrational angular momenta (associated with degenerate bending
motion) causes the potential energy function to split into two when the molecule is bent.
This is called the Renner-Teller effect. See Ch. Jungen and A. J. Merer, Molec. Phys.,
40, 1 (1980). For a degenerate electronic state of a nonlinear polyatomic molecule,
there are nontotally symmetric displacements of the nuclei that cause a splitting of the
potential energy function such that the potential minima are no longer located in the
symmetrical position. This vibrational-electronic interaction is called the Jahn-Teller
effect. See T. A. Miller and V. E. Bondybey, in Molecular Ions: Spectroscopy, Structure
and Chemistry, T. A. Miller and V. E. Bondybey, eds. (North-Holland, Amsterdam,
1983), pp. 201-229.
32. J. H. Van Vleck, Phys. Rev., 33,467 (1929).
33. 0. M. Jordahl, Phys. Rev., 45,87 (1934).
34. See E. C. Kemble, Fundamental Principles of QuantumMechanics (McGraw-Hill, New
York, 1937), pp. 394-396.
35. D. R. Herschbach, unpublished notes; see J. E. Wollrab, Rotational Spectra and
Molecular Structure (Academic Press, New York, 1967), Appendix 7. In the treatment
we present, it is made explicit that the expression for Hij includes more than one
interacting block, but we have omitted a tern in (il Sj Ij) that is ( Ei - Ej) /( E, - Eg)
times smaller than the other terns, where a and p belong to different blocks.
NOTES AND REFERENCES 293

36. The vibration-rotation Harniltonian is derived in E. B. Wilson, Jr., J. C. Decius, and

P. C. Cross, Molecular Vibrations(McGraw-Hill, New York, 1955) and in D. Papougek
and M. R. Aliev, Molecular Vibrational-Rotational Spectra (Usevier, Amsterdam,
1982).
37. We drop the J dependence on the energy difference (which would show up as a higher
order correction).
38. Z. I. Slawsky and D. M. Dennison, J. Chem. Phys., 7,509 (1939).
39. J. B. Howard and E. B. Wilson, Jr., J. Chem. Phys., 4 , 260 (1936); E. B. Wilson, Jr.,
J . Chem. Phys., 5,617 (1937).
40. J. K. G. Watson, in VibrationalSpectra and Structure, Vol. 6, J. R. Durig, ed. (Elsevier,
Amsterdam, 1977),pp. 1-89.
41. M. R. Aliev and J. K. G. Watson, in Molecular Spectroscopy: Modern Research, Vol. 111,
K. Narahari Rao, ed. (Academic Press, New York, 1985),pp. 1-67.
42. E. Hirota, High-Resolution Spectroscopy of Transient Molecules (Springer-Verlag,
Berlin, 1985).
43. E. U. Condon and G. H. Shortley. Theory of Atomic Spectra (Cambridge University
Press, Cambridge, England, 1935).
44. H. Hijnl and F. London, Z. Phys., 33, 803 (1925).
45. Reference [27], pp. 86-93.
46. See W. H. Flygare, Molecular Structure and Dynamics (Prentice-Hall, Englewood
Cliffs, NJ, 1978).
47. E. U. Condon, Am. J. Phys., 15,365 (1947); Phys. Rev., 32,858 (1928).
48. C. Noda and R. N. Zare, J. Molec. Spectrosc., 95,254 (1982).
49. For example, an electronic transition is said to be pure electric dipole forbidden if
(E'V'R'I ~ g ( p g )IE"V"R1')
o vanishes. However, if by retaining additional terms, the
electric dipole moment matrix element is nonzero, the transition is said to be vibronically
induced. See G. Herzberg, Electronic Spectra of Polyatomic Molecules (Van Nostrand,
New York, 1966).
294 6 ENERGY-LEVEL STRUCTURE AND WAVE FUNCTIONS OF A RIGID ROTOR

PROBLEM SET 4

1. Let fi be a unit vector pointing along the angular velocity vector w so that

The kinetic energy of the rigid body in the center-of-mass frame is given by [see
Eq. (6.18)l

but [see Eq. (6.1I)]

Hence

where I is a scalar quantity given by

It is traditional to call I the moment of inertia about the axis of rotation.

A. Evaluate fi . .fi and show that Eq. (6.140) results.
B. The distance of the particle i from the axis of rotation fi is ri sin Bi, where Bi
is the angle included between r, and fi. Thus this distance is the same as the
magnitude of the vector ri x fi. The customary definition of the moment of
inertia about an axis of rotation is the mass times the square of the distance
of the object from the rotation axis. Hence for a rigid body

Show that Eq. (6.141) is equivalent to Eq. (6.140).

The value of I depends on the direction of w . In general, w changes its direction
with respect to the center-of-mass frame as a function of time, causing J also to
change its direction. At any given moment
PROBLEM SET 4 295

2. Microwave spectrum of methylene fluoride.

A. Calculate the principal moments of inertia (expressed in m u A2) and rota-
tion constants A, B , C (expressed in MHz and in cm-') for the methylene
fluoride molecule, l2 CH2F2. Use the following parameters:

TCH = 1.092 A, TCF = 1.358 A

LHCH = 111•‹52', LFCF = 108O17'
m H = 1.007825 m u , mc = 12.000000 m u , mF = 18.99840 m u

A handy conversion factor is

B. List the frequencies (in MHz) of the allowed rotational transitions for methy-
lene fluoride with J 5 2.
C. Discuss whether the observation of the above transitions could be used to
determine the bond angles and bond lengths in methylene fluoride. Are any
assumptions or additional information needed?

3. In terms of the rotational constantsA, B, C , show that the asymmetrictop energy

levels for J = 3 have the form:

Prolate Limit
General Formula A>B=C

Hint: An inspection of the symmetric form of the solutions should save you
some time; it is not necessary to find the roots of each secular determinant.

4. Consider a symmetric triatomic molecule AB2 that belongs to the point

group C2,. Let r = TAB denote the length of the AB bond and 0 = ~ B A Bdenote
the included angle. The masses are designated by MA and MB.
 6 ENERGY-LEVEL STRUCTURE AND WAVE FUNCTIONS OF A RIGID ROTOR

Show that the three moments of inertia are given by

where 1 is the principal axis in the plane of the molecule that bisects the BBAB
angle, 2 is the principal axis in the plane of the molecule that is perpendicular
to 1, and 3 is the principal axis perpendicular to the plane of the molecule
(i.e., perpendicular to axes 1 and 2 ) .
We can identify I,, with 13.In order for II to equal 12, it is necessary that

For 8 > = I1 while for 8 < d i m , I,, = IIand I b b = I 2 .

,I,, = 1 2 and Ibb
Show that the asymmetry parameter n depends only on the mass ratio

;
Make a plot of n versus BBAB for r = (corresponding to MA = M B ) ,
r = $ (corresponding to MA = 2 M B ) ,and r = $ (corresponding to MA =
0 . 5 MB). The only case for which the AB2 molecule is exactly a symmetric
top is when the AB2 molecule is linear (OBm = 180'). Examination of
the graph of n versus BBAB will explain why most AB2 triatomic molecules
are near prolate symmetric tops, although eBABoften differs markedly from
180".
We define the inertia defect A as

Note that A = 0 for a symmetric triatomic molecule. Prove that A = 0 for

any planar rigid rotor.
Suggest how a real "planar" molecule, such as benzene or ozone, can have
a nonvanishing inertia defect.
For a general discussion of the inertia defect, see J. E. Wollrab, Rotational
Spectra and Molecular Structure (Academic Press, New York, 1967).
APPLICATION 15 INTRODUCTION TO DIATOMIC MOLECULES 297

APPLICATION 15
INTRODUCTION TO DIATOMIC MOLECULES

The rotational energy levels, wave functions, and line strengthsof diatomic molecules
may be worked out in a manner similar to those presented for asymmetric tops. The
following treatment is intended only as a brief introduction to this rather extensive
topic[l-211. We concentrate our attention on 2~ and 217 electronic states because
these are the simplest and most commonly encountered examples of open-shell
diatomics.
We begin by presenting our cast of angular momentum characters. We ignore
nuclear spin, a tiny shy fellow, because hyperfine structure is seldom resolved in
optical spectroscopy, and the effects of nuclear spin can be readily taken into account,
if needed[22]. We define

L -electronic orbital angular momentum

S -electronic spin angular momentum
J -total angular momentum
N -total angular momentum excluding electronic spin (i.e., N = J - S)
R -nuclear rotational angular momentum (i.e., R = N - L).

We also introduce the Hund's case (a) coupling scheme (see Figure 1). In Figure 1
L and S are both "tied" to the intemuclear axis (the z axis in the molecule-fixed frame)
making the signed projections A and Z ,respectively, while R is at right angles to the
intemuclear axis. Thus the total angular momentum J makes the projection M on
the space-fixed Z axis and the projection f 2 = A + C on the molecule-fixed z axis.
Consequently, in case (a) coupling the rotational wave function is the symmetric top
wave function IJ f 2 M ) .
A linear molecule has only two rotational degrees of freedom. Hence the Euler an-
gle x is redundant and its value can be fixed arbitrarily. We follow the convention[l4]
that x = 0 and write for the rotational wave function

where the normalization factor has been altered to reflect the fact that integration over
the solid angle element no longer includes X. The choice of x = 0 implies that the
molecule-fixed y axis coincides with the line of nodes, that is, lies in the plane of
rotation, while the molecule-fixed x axis is perpendicular to the plane of rotation,
for M= J.
298 6 ENERGY-LEVELSTRUCTURE AND WAVE FUNCTIONS OF A RIGID ROTOR

FIGURE 1 Vector diagram for Hund's case

(a) coupling.

which is the direction J points in the high J limit. The alternative choice[ll, 201,
x = 7r/2, places the molecule-fixed y axis perpendicular to the plane of rotation for M= J.
The case (a) wave functions may be written as a product of an electronic orbital
part Id) , an electronic spin part ISZ ), a vibrational part Iv), and a rotational
part 1 JSL M):

The separation IdS Z ) = I I

d ) S Z ) is strictly valid for only one- and two-valence
electron molecules[l3]. It is retained here because we are interested primarily in
the symmetry nature of the electronic part of the wave function, which we shall also
write as In S+lAn) . These basis set functions prove to be a convenient starting point
because the operators L,, 5' ,J,, s2,and S, are all diagonal in this coupling scheme.

lZ States

Therotational Hamiltonian is simply

APPLICATION 15 INTRODUCTIONTO DIATOMIC MOLECULES

where

Thus in the vibrational level v the effective Harniltonian becomes

and the energy levels are given by

since J = N = R when L = S = 0 . Equation (6) is recognized as the familiar

expression for the rotational energy levels of a linear rigid rotor. The mechanical
interpretation of B, as proportional to (vl l / r 2 Iv) assumes that mixing with other
electronic or vibrational levels is negligible.

A. Use the Van Vleck transformation to calculate the first centrifugal distortion
correction to the rotational energy levels of a 'C state. Show that

where

Ill States

In Ill electronic states IA I = 1, and we must introduce the effects of L. The

rotational Hamiltonian is

that is,

The second term in Eq. (10) has no J dependence, and its expectation value is
incorporated into the band origin (which is a polite way of saying that it is neglected).
300 6 ENERGY-LEVEL STRUCTURE AND WAVE FUNCTIONS OF A RIGID ROTOR

The third term in Eq. (10) has no nonvanishing expectation value; instead, as we shall
see, it couples states that differ by one unit in A , that is, couples l7 states to C and
A states. The first term in Eq. (10) does have an expectation value, and we readily
find for the rotational energy of a 'll state ( [AI = 1) :

where for 'll states the lowest J value is J = 1. Again because S = 0 we could also
replace J by N in Eq. (1l), if desired. To this approximation, the rotational energy
level structure of a 'l7 electronic state looks just like a 'C state, except the J = 0
level is missing. Note that each level is doubly degenerate as A = +1 and A = - 1
give the same energy; that is, it does not matter whether L precesses clockwise or
counterclockwise about the internuclear axis. This is called A doubling or A-type
doubling. It is the diatomic molecule analog of K-type doubling in a symmetric top.

2C States

1na2C state,^ = 0 b u t s = $ sot hat^ = +$or-$ accordingtowhetherC = +;

9.
or - Rather than evaluate a 2 x 2 secular determinant, we rewrite the molecular
wave function as an appropriate linear combination of the case (a) wave functions in
the same spirit as the Wang transformation

and for the special case A = 0 , C = 0

where pf denotes the sign of the linear combination. As we shall see, pf is related to
the parity of the wave function. The rotational part of the Hamiltonian is once again
B( r) R' = B( r ) ( J - S) 2 , but we must also include the phenomenological spin-
rotation coupling term 7 ( r ) N . S (which is usually caused almost exclusively by
spin-orbit interaction that couples 2C and 'll states[23]). Hence for the vibrational
level v the rotational Hamiltonian may be written for C' states as
APPLICATION 15 INTRODUCTION TO DIATOMIC MOLECULES

where B, is the rotational constant and 7, the spin-rotation constant.

B. Show that

where J = N + $,and

where J = N - i. Hint: Evaluate (I,!I(~*) 1x2; ,",)I,!I(p*)).Use the phase

conventions (see Section 3.4):

Note carefully the similarities and differences between Eqs. (19) and (22).

Equations (15) and (16) show that (even in the absence of perturbations[20]) the
rotational levels of a 2C state are split by an amount

where the magnitude of the spin-rotation splitting increases linearly with J (in the
absence of centrifugal distortion corrections to 7,). Thus each rotational level except
the J = $ level of p+ parity occurs in closely spaced pairs. The levels with
J = N + $ are called F l , and those with J = N - $ are called F2. Hence
302 6 ENERGY-LEVEL STRUCTURE AND WAVE FUNCTIONS OF A RIGID ROTOR

FI( N ) - F 2 ( N ) = i Y v ( 2N + 1) [see Eqs. (15) and (l6)]. The spin-rotation

constant 7, may be positive or negative.

211States

4 i.
In a 211state, A = f1 and S = so that Z = f Hence B can take on the
values -$, - i, i, and $. Except for the J = $ level (which can be exclusively
associated with 10I = i ) , each J level has four rotational levels associated with it,
two of each parity. We use as basis set wave functions

(24)
and

(25)
The contribution to the energy caused by nuclear rotation is expressed as

= B ( r ) [ ( J - S) - L ] ~

In the last line of Eq. (26) the first two terms contribute to the first-order energy,
the third term connects states differing by one unit in A and the last term has no J
dependence and once again contributes only an overall shift to the total energy. Hence
we write

'~122
= BB,[J2- 2 J,S, + S2 - 2 (J, - S,) L, + L:] -Bv [ J+S-+ J-S+] (27)

We also must introduce the spin-orbit interaction which we approximate[24] as

APPLICATION 15 INTRODUCTION TO DIATOMIC MOLECULES 303

Only the A( r ) LzSz term contributes to the energy in first order so that

7igJ = AuLzSz (29)

C. The rotational energy levels of a 'l7 state are the same for each parity block,
7 = E(217;v Jp-) . Hence they occur in doubly degenerate
that is, ~ ( ~;vlJp+)
pairs. Show that for J > i
) B, [ ( J -
E ( ' ~ ; v J= i)( J + 9)& $XI (30)
where

X = [4 ( J - 5) ( J + f ) + ( Y - 2)'lt

and

Hint:Show that the 2 x 2 secular determinant is

and carry out its evaluation,using the identity J ( J + 1) - $ = ( J - 9) ( J + f ).

D. The energy levels associated with the plus sign in Eq. (30)are called Fz ; those
with the minus sign Fl . Show that

where
304 6 ENERGY-LEVEL STRUCTURE AND WAVE FUNCTIONS OF A RIGID ROTOR

FIGURE 2 Vector diagram for Hund's case

(b) coupling.

For large rotational values, Fl( J ) is the term series that behaves as B, N( N + 1)
i
with J = N + while F2( J ) is the term series with J = N - i.

It is informative to compare the value of ( Y - 2)2 to 4 (J - i) (J + %) =

4 J ( J + 1) - 3. When the former dominates, bJ m 0 and we can assign the F!
and 4 levels as belonging to separate 211r and 2H; states. It is for this situation that
the molecular behavior follows closely the case (a) coupling scheme. However, with
increasing rotation the B,( J - L) . S = BUN. S coupling term from the rotational
part of the Hamiltonian grows in importance compared to the A, L . S spin-orbit
coupling term. This causes S to uncouple from the nuclear axis and recouple to N to
give the resultant total angular momentum J. This coupling scheme, called Hund's
case (b) is shown in Figure 2.
In the case (b) limit Eqs. (36) and (37) show that a~ = bJ = (2)-); that is, the
wave functions for the Fl and Fz energy levels become an equally weighted linear
combination of the case (a) wave functions[25]. In general, the coupling in 211 states
is intermediate between case (a) and case (b), going away from case (a) and toward
case (b) with increasing rotation. If A, > 0 , the energy ordering is said to be regular,
while for A, < 0 , the energy ordering is said to be inverted.

E. Construct a plot of E(211; v J ) versus Y for B, = 1 cm-', -20 5 Y 2 20 and

J= g, i, i.
and Optional: Include the level J = in your plot.

Most diatomic molecules are well described by intermediate behavior between the
limits of Hund's case (a) and Hund's case (b) coupling. However, there are two other
APPLICATION 15 INTRODUCTION TO DIATOMIC MOLECULES

FIGURE 3 Vector diagram for Hund's case (c) cou-

pling.

coupling schemes, introduced by Hund[2], which also are of importance. For heavy
nuclei, the atomic spin-orbit interaction often becomes large and the interaction
between L and S may exceed that between L and the internuclear axis. In this case,
called Hund's case (c) coupling, L and S add vectorially to form the resultant J,
which then couples to the internuclear axis to make the projection R. The nuclear
rotational angular momentum R then adds vectorially to R to form the total angular
momentum J (see Figure 3). The R = 0 states are nondegenerate and we speak of O+
and 0- states in strict analogy to C+ and C- states. The R # 0 states would be
degenerate for a nonrotating molecule but split under rotation into two components.
This is called R -type doubling in analogy with A-type doubling.
An opposite extreme is described by Hund's case (d) coupling (see Figure 4)
in which the coupling between L and R is much larger than that between L and
the internuclear axis. Hund's case (d) may be regarded as what happens to Hund's
case @) molecules in the limit of very large rotation (L-uncoupling). Hund's case (d)
coupling also serves as an excellent starting point in describing the electronic states
of many Rydberg molecules where the Rydberg electron is atomic-like and weakly
interacts with the molecular core.

Parity Classification of Rotational Levels

Each wave function associated with an energy level may be classified as even or odd
according to whether it remains unchanged or changes sign on inversion through the
origin of the spatial coordinates of all particles. Since the molecular Hamiltonian
is invariant under this symmetry operation, only states of the same parity have
nonvanishing matrix elements connecting them-a fact we already employed to
306 6 ENERGY-LEVELSTRUCTURE AND WAVE FUNCTIONS OF A RIGID ROTOR

FIGURE 4 Vector diagram for Hund's case (d) coupling.

advantage when we used a case (a) parity basis set to factor the secular determinant
to find the rotational energy levels of 'C ,'II ,2Z,and 217 electronic states. Moreover,
because the electric dipole operator changes sign under inversion of the spatial
coordinates of all particles, electric dipole transitions can only connect levels of
opposite parity. Therefore, it is of great importance to establish the parity of different
rotational levels. This is accomplished by considering how the inversion operator i,,
acts on the case (a) wave functions, a subtle topic that one might want to skim on first
reading[26].
It can be shown from simple geometric arguments that the inversion of the spatial
coordinates is equivalent to reflection of the molecule-fixed electronic coordinates
followed by rotation of the molecular frame by 180" about an axis through the origin
and perpendicular to the reflection plane. In the language of group theory we can
write is, = C2( y) q,(XZ) or iSp= C2( x)cr,( y z ) . We choose the latter as it is more
convenient[27] to carry out these operations with our choice of x = 0. It may be
shown[ll, 13, 16, 191 that

and

where s = 1 for C- states, s = 0 for all other states. We outline how Eqs. (38)
and (39) come about.
For A # 0 states, I d ) transforms as YLA( 0, X) (L unspecified) and under the
reflection in the yz plane, x -+ 7r -X so that

For A = 0 states, we already denote whether the orbital part of the electronic wave
function is unchanged or changes signs under reflection in a plane containing the
APPLICATION 15 INTRODUCTIONTO DIATOMIC MOLECULES 307

internuclear axis by calling the former Z+ states and the latter C- states. Hence
s = 0 except for Z - states in which case s = 1.
Under the action of o,( yz), x -t -x, which is the same as an inversion in the
molecular frame x -+ -x, y --+ -y, z -+ -z followed by a rotation by .~rabout
I t)
the x axis, Cz(x) . Under inversion the spin-up wave function a = f , and the
I i)
spin-down wave function ,8 = f , - are unchanged, whereas under C2 (x) ,

The spin state ISC) is a determinantid wave function of n indistinguishable

electrons. However, we are only interested in the symmetry properties, so we shall
take some liberties. The ISZ ) state may be regarded as formed from a superposition
of coupled states (ISl, sn)SX), where IS,l C-1) may also be regarded as a
superposition of the coupled states I(Sn4 sn-1 ) Sn-1 Cn-1 ), and so on. Hence the
uncoupled component Isl a1) Isz02) . . . Isnan) occurs in ISC ) with a coefficient
whose dependence on the spin projections q is contained in the product of Clebsch-
Gordan coefficients, (S-1, &-I, snonIS2 )zS,( ,En-2 , sn-i 0-1 ISn-1 En-1 ) x
- .x (sl 01 , s2 a2 IS2C2). On reflection we interchange a and L,3 in the uncoupled state
so that Z -+ -X , and we pick up a phase factor (-I)=, where x equals the number
of electrons divided by two. From the product of Clebsch-Gordan coefficients we
also pick up a second phase factor ( - 1) Y where

because the magnetic quantum numbers of each Clebsch-Gordan coefficient have

been reversed [see Eq. (2.26)]. Hence

a,( yz) ISZ) = (-1)'" IS -Z) = (-1)' IS -C) (45)

308 6 ENERGY-LEVEL STRUCTURE AND WAVE FUNCTIONS OF A RIGID ROTOR

Finally we turn to the behavior of IJQ M ) under spatial inversion. The Euler
angles ( 4,O, X) relate the space-fixed frame F = X Y Z to the molecule-fixed frame
g = x yz according to

~4 C O CX-SC$ s x - c $ C O S X - S $ c x ~4 S O
s x - c 4 c O s x + c ~ c xs + s O
se sx cO
(46)
where sin a and cos a have been abbreviated s a and c a, respectively. Under spatial
inversion X -+ -X, Y -+ -Y, and Z -+ -2. In order to preserve Eq. (46), it is
readilyseenthatx -+ -x, y +y,andz -+zaswellas4 -+ .rr+ 4,O -+.rr-8,and
x -+ - X. As Table 6.3 shows, this corresponds to the group operation C2(x) on
the Euler angles. From Eq. (6.60) it follows that

The vibrational wave function Iv) depends solely on the internuclear distance r
Therefore, Iv) is unchanged by spatial inversion.
Putting together all the bits, we have

Thus the levels of the pf parity block have the parity f(-1) J-S+8 under spatial
inversion. We follow the convention[28] of designating the parity levels as e or f
according to the following. For integral J values, e levels have parity ( - 1) and f
levels ( - 1)J+l; for half-integral J values, e levels have parity ( - 1) J - f r and f levels
(-l)J+k

Orientation of the Electronic Charge Cloud

In the case (a) limit the electronic charge distribution has strict cylindrical symmetry
about the internuclear axis as both L and S precess uniformly about this axis.
However, for A # 0 states, molecular rotation causes the wave function associated
with the p+ and p- A components of the same J to be such that one is symmetric (A')
while the other is antisymmetric (A") with respect to reflection in the plane of rotation
of the molecule[29]. One measure of this is the electron charge cloud asymmetry[30],
given by
A = (f2 - y2) = (cos2 x - sin2 X) = (cos 2 ~ ) (49)
APPLICATION 15 INTRODUCTION TO DIATOMIC MOLECULES 309

where x is the azimuthal angle of the electron charge distribution in the molecular
frame measured from the line of nodes (which we have taken to coincide with the y
axis). The dependenceof the electronic orbital angularmomentum part of the case (a)
wave function on x has the form Id) = ( 2 n)-) exp( iA )o as we have already
noted.

F. Show that

by explicit evaluation of A using the Fl and F2 wave functions. Hint:Present

arguments that

while

We conclude that the A value for each pair of A components of fi and F2 is

i,
oppositely signed[31]. For J = [A I = 0 , and the electronic charge distribution is
cylindrically symmetric. As J increases, IA I 4 $; the charge distribution points for
one A component along P (along J in the high J limit[32]) the other along 9 (in the
plane of rotation); both are perpendicular to the internuclear axis. The importance of
this is that a preferential population of one of the two A components in a particle,
surface, or photon collision process serves to tell us about the dynamics of this
process in terms of the asymmetry produced in the electronic charge distribution of
the product[33].

G. Find the values of IA ( versus J for J in the case of the two 2nground
state radicals, OH and NO, given that Y = -7.4 for OH x2nv = 0 and
Y = 72.9 for NO x2IIv = 0 . This comparison explains why it is more likely
to observe preferential population of the A components in OH x2n than in
NO x2n for low to moderate J values in a "collision process."

In addition to characterizing the two A-doublet levels by the label elf to denote
the total parity, it is useful in the high J limit to label A-doublet levels by A (A')
and A(Al1) depending on whether the electronic wave function is symmetric or
antisymmetric with respect to reflection in the plane of rotation of the molecule
of the spatial coordinate of all electrons[29]. Even if the A doublets are nearly
310 6 ENERGY-LEVEL STRUCTURE AND WAVE FUNCTIONS OF A RIGID ROTOR

degenerate in energy, electric-dipole-allowed transitions access separately the two

A-doublet levels of opposite parity, making the population differences, if present,
readily observable. For example, irrespectiveof multiplicity, the main-branch Q lines
of a II - Z+ transition probe the l l ( A " ) levels in the high J limit, while the main-
branch P and R lines probe the l l ( A ' ) levels in the high J limit. For a l l - Z -
transition this is reversed: Q lines probe l 3 ( A 1 )levels while P and R lines probe
l l ( A " ) levels[29].

A-Type Doubling

In the foregoing we have not treated explicitly terms in the molecular Hamiltonian
that couple different electronic states. We consider here interactions between Z and
II states of the same multiplicity by means of the Van Vleck transformation. This
causes additional adjustable parameters to appear in the block of levels having the
same parity, J, S, and A. In particular, for 'll and 211 electronic states this removes
the degeneracy of the A components and causes them to differ in energy.

H. For a 'll state the terms in the Hamiltonian responsible for causing interaction
between the Ill state and other 'C+ and 'Z - states are

[see Eq. (lo)]. Use the Van Vleck transformation to show that for a v , J level
of P+ parity
~ g A l , ~ ==qlv ( I X ' ) J ( J + 1) (52)

and for a v , J level of p- parity

where

(nllllvl B(r)L+In' 'X$vl) (n' lX$v11 B ( r ) L - lnlIIlv)

qv (lZh) =
dv'
E( n Ill1 v ) - E(nl lZ; v l )

Thus to a good first approximation the A-type splitting for a 'II state is given by

E ( ' ~ l v ~ e ) - E ( ' l l l v ~ f ) = [ ~ , ( ' Z + ) - ~ ~ ( ' ~ - ) ] J(55)

(J+l)
APPLICATION 15 INTRODUCTION TO DIATOMIC MOLECULES 311

that is, increases quadratically with rotation. Please also note that 'C+ states can
perturb only the A components of e symmetry and 'C - states those of f symmetry.
Which A component has the lower energy depends on the relative size and sign of
q, (I);+) and q, ('C-), and not on some mechanical model involving the moment
of inertia difference of the electron charge distributions associated with the two
A components (a common misconception).
The treatment of A doubling in 211 states proceeds along similar lines, but the
splitting does not have a simple analytical form. We introduce three A-doubling
parameters[5]:

and consider the correction Hamiltonian to be

The sum in Eq. (56) is over all vibrational levels of all 2X+ and 2C - electronic states.
The Van Vleck transformation shows that we can replace Eq. (57) by an effective
A-doubling Hamiltonian for the 21'1 state in level v of the form[34]

However, the A-doubling parameter o, cannot be determined independently from

spectroscopic data (its variation is totally correlated with variation in the location
of the vibrational level origin and in the value of the spin-orbit constant A, for the
211 state). Consequently, we omit it from further consideration. For the 2 x 2 block of
p+ parity H:) contributesthe additionaloff-diagonal terms to the secular determinant
312 6 ENERGY-LEVEL STRUCTURE AND WAVE FUNCTIONSOF A RIGID ROTOR

and for the 2 x 2 block of p- parity

For 212 states well approximated by case (a) coupling, that is, states for which
( Y - 2)2 >> 4 ( J - $) ( J + $) , the A-type doubling for the ht component
varies linearly with J , whereas for the 21Tg component it varies with the third power
of J and for low J values is very small compared to that of 2J3;.

Rotational Line Strength

The wave function associated with each rotational level of a diatomic molecule can
be expressed as a linear combination of case (a) wave functions. We write

where In 2 S + 1 ~ nis) the electronic part, Iv) is the vibrational part, and I J Q M) is the
rotational part. In Eq. (65) the sum extends over all possible values of $2 = A + Z
from -A - S to A + S so that there are in general ( 2 - S0,h) ( 2 S + 1) different
anza(pk) coefficients. The anZa(pf) are just the columns of the unitary matrix
(eigenvectors) that diagonalizes the secular matrix of the molecular Hamiltonian for
APPLICATION 15 INTRODUCTIONTO DIATOMIC MOLECULES

TABLE 1 -
Rotational Line Strengths for a 2C 2C Transition.@

Branch S( J'; J )

@Theseexpressions were first derived by R. S. Mulliken, Phys. Rev.,30,138 (1927).

J' , t-+
parity block p*. The rotational line strength factor for the 2 s 8 ' ~ & 2 S + ' ~ nJ
transition is calculated[35] in exact analogy to Eq. (6.123):

S(J'; J) = ( 2 J 1 + 1 ) ( 2 J + 1)

where S( J'; J ) vanishes unless the upper and lower levels are of opposite parity.
'
For ' A'J' t+ 'A J transitions ('Z - 'Z &, 'C - 'I7 , 'll - 'C ', etc.), Eq.(66)
reduces to the Honl-London factors, where A takes the place of K" in Table 6.6. For
electric-dipole-allowedtransitions involving electronic states of higher multiplicity,
it is convenient in general to evaluate Eq. (66) with a simple computer program[35,
361. Analytical expressions can be derived in some special cases, but these expres-
sions become increasingly complex as the multiplicity increases and the coupling in
the 2 S + 1 ~state deviates from the case (a) and case (b) limiting forms[9,35,36].
For reference purposes we list the rotational line strength of a 2Z - 2C transition
in Table 1 and those for 2C - 211 and 211 - 2C transitions in Table 2. The branches
are designated P , Q, or R according to whether A J = -1,0, or +l. In addition,
subscripts 1 and 2 are added to denote the Fl and F2 levels, first the upper level,
then the lower one. If both subscripts are the same, it is traditional to omit one (e.g.,
Pll = P I ,P22 = P2,etc.). For a 2C state the Fl levels are those for which J = N +
i.
and the F2 levels are those for which J = N - For a 'I7 state the Fl and F2 levels
are as defined in Eqs. (34) and (35), that is, in the case (a) limit FI is associated
with 211 and f i with 211+ for Y >> 0, the reverse for Y << 0. In applying the
rotational line strengths to molecular problems, one must take care in many cases to
distinguish whether the line transition refers to a specific pair of A components of a
A J transition or involves a sum over two (often unresolved) A component pairs of
a A J transition. The line strengths are chosen to satisfy the sum rule[37]

CS(J1,J") = (2 - 6 0 , A ~ 6 0( ,2~S~+~1)) ( 2 J + 1) (67)

314 6 ENERGY-LEVELSTRUCTURE AND WAVE FUNCTIONS OF A RIGID ROTOR

TABLE 2 Rotational Line Strengths for 2x - 2nand 2n- 2x Transitionsa

'These expressions were first derived by L. T. Earls, Phys. Rev.,48,423 (1935).

"= [r(l-+)(,+$)+iu-s'] +.
NOTES AND REFERENCES 315

where the summation is over all allowed transitions from or to the group of
( 2 - SO,*)( 2 S + 1) J levels with the same value of J' or J". The sum rule is sym-
metric in J' and J"; however, Eq. (67) is not valid for the first few rotational levels
where the full spin multiplicity is not present.

I. Show that S ( J 1 ;J ) = ( 2 J + 1)/[4 J ( J + I)] for the Q12( J ) and Q z l ( J )

branch members of a 2C - 2C transition (second entry in Table 1). Hint : Do not
forget that only levels of opposite parity can be connected in an electric-dipole-
allowed transition.

By associating A of the diatomic molecule with K of an asymmetric top molecule,

we can classify electronic dipole transitions in diatomic molecules as parallel or
perpendicular transitions. In a parallel transition, M = 0 (e.g., C - C ,II - Il ,etc.)
and the transition dipole moment lies along the internuclear axis (the z axis)[38].
In the high J limit, only P and R branches have intensity. In a perpendicular
transition M = f1 (e.g., I; - ll , ll - C ,etc.) and the transition dipole moment is
perpendicular to the internuclear axis. In the high J limit, P and R branch members
have intensity only when the transition dipole moment lies in the plane of rotation
(perpendicular to the internuclear axis), and the Q branch members have intensity
only when the transition dipole moment lies along J. These simple considerations
are useful in determining how a diatomic molecule will interact with a beam of plane
polarized (or unpolarized) light.

NOTES AND REFERENCES

1. G.Hemberg, Spectra of Diatomic Molecules, Van Nostrand, New York, 1950.

2. E Hund, Z.Phys., 42,93(1927);63,719(1930);Handb. d . Phys., 24, I , 561 (1933).
3. R. de L. Kronig, Z. Phys., 46,814(1928);50,347(1928);Band Spectra and Molecular
Structure, Cambridge University Press, Cambridge, England, 1930.
4. E. L.Hill and J. H. Van Vleck, Phys. Rev., 32,250(1928);J . H. Van Vleck, Phys. Rev.,
33,467(1929);40,544(1932);Rev. Mod. Phys., 23,213 (1951).
5. R. S. Mulliken and A. Christy, Phys. Rev., 38,87(1930);R. S. Mulliken, Rev. Mod. Phys.,
4,51,(1932).
6. Y.-N. Chiu, J.Chem. Phys., 42,2671 (1965);45,2969 (1966);57,4056(1972).
7. K. F. Freed, J. Chem. Phys., 45,4214(1966).
8. R. T.Pack and J. Hirschfelder,J. Chem. Phys., 49,4009(1968).
9. I. KovaCs, Rotational Structure in the Spectra of Diatomic Molecules, American Elsevier,
New York, 1969.
10. A. Carrington, D. H. Levy, and T. A. Miller, Adv. Chem. Phys., 18,149 (1970).
11. J. T. Hougen, "The Calculation of Rotational Energy Levels and Rotational Intensities in
Diatomic Molecules," N.B.S. Monograph 115,Washington, DC, 1970.
316 6 ENERGY-LEVELSTRUCTURE AND WAVE FUNCTIONS OF A RIGID ROTOR

12. L. Veseth, Theor. Chim. Acta, 18, 368 (1970);J. Phys. B, 3, 1677 (1970);J . Phys. B , 4 ,
20 (1971);J . Molec. Spectrosc., 38,228 (1971).
13. I. Roeggen, Theor. Chim. Acta, 21,398 (1971).
14. R. N. Zare, A. L. Schmeltekopf, W. J. Harrop, and D. L. Albritton, J. Molec. Spectrosc.,
46,37 (1973).
15. M. Mizushima, The Theory of Rotating Diatomic Molecules, Wiley, New York, 1975.
16. B. R. Jucld, Angular Momentum Theory for Diatomic Molecules, Academic Press, New
York, 1975.
17. J. M. Brown, M. Kaise, C. M. L. Kerr, and D. J. Milton, Molec. Phys., 36, 553 (1978);
J. M. Brown, E. A. Colbourn, J. K. G. Watson, and F. D. Wayne, J. Molec. Spectrosc., 74,
294 (1979).
18. P. R. Bunker, Molecular Symmetry and Spectroscopy, Academic Press, New York, 1979.
19. M. Larsson, Physica Scripta, 23,835 (1981).
20. H. Lefebvre-Brion and R. W. Field, Perturbations in the Spectra of Diatomic Molecules,
Academic Press, Orlando, FL, 1986.
21. P. D. A. Mills, C. M. Western. and B. J. Howard, J. Phys. Chem., 90,3331 (1986).
22. However, one must not forget about nuclear spin statistics, which can cause every other
level to be missing in homonuclear diatomic molecules composed of zero spin nuclei.
For a review of hyperfine structure in the electronic spectra of diatomic molecules, see
T. M. Dunn, in Molecular Spectroscopy, Modern Research, K. N. Rao and C. W. Mathews,
eds., Academic Press, New York, 1972, pp. 231-257.
23. S . Green and R. N. Zare, J. Molec. Spectrosc., 64,217 (1977).
24. This is the phenomenological spin-orbit Hamiltonian, which can only connect electronic
states of the same multiplicity. To include singlet-triplet, doublet-quartet (etc.)
interactions, we must replace Xso by the microscopic spin-orbit Hamiltonian. See
P. R. Fontana, Phys. Rev., 125,220 (1962);K. Kayama and J. C. Baird, J. Chem. Phys.,
46,2604 (1967);L. Veseth, Theor. Chim. Acta, 18,368 (1970).
25. The 2X electronic state follows case @) coupling, and its wave function is also an equally
1
weighted linear combination of the two case (a) wave functions Id)1st) J : M ) and
IS IJ
I d ) -$) - + M ) .
26. The arguments present many chances for error. After considering this problem at some
length, Chiu[6] writes in footnote 25 of the 1966 paper: "Thus we see that nature, noble
and true, retains her supreme symmetry in spite of our sometimes misleading, humble
effort of description which has yet to strive for perfection." Indeed, this problem did not
get fully sorted out until the work of Larsson[l9] in 1981.
27. For the choice x = 7r/2, the operations i,, = C2( y ) o;(xz) are the more convenient.
However, in either case the same result [Eq. (48)]is obtained.
28. J. M. Brown, J. T. Hougen, K.-P.Huber, J. W. C. Johns, I. Kopp, H. Lefebvre-Brion,
A. J. Merer, D. A. Ramsay, J. Rostas, andR. N. Zare, J. Molec. Spectrosc., 55,500 (1975).
29. M. H. Alexander and P. J. Dagdigian, J. Chem. Phys., 80,4325 (1984). For 'n states
the e levels are symmetric with respect to reflection in the plane of rotation and the
f levels are antisymmetric. For 2n states the Fl e and Fz f levels are symmetric and
the Fl f and F2 e levels are antisymmetric. The ( A ' ) and A ( A " ) nomenclature to
describe the A-doublet levels in rotating linear molecules is proposed and discussed by
M. H. Alexander et al., J. Chem. Phys., 89, 1749 (1988).
NOTES AND REFERENCES 317

30. W. D. Gwinn, B. E. Turner, W. M. Goss, and G. L. Blackman, Astrophys. J., 179,789

(1973).
3 1. As Alexander and Dagdigian [29] show, the electron distribution in the FleA-doublet level
is oriented preferentially in the plane of rotation and in the Fl f A-doublet level oriented
preferentially perpendicular to the plane of rotation along J for a single filled orbital. The
opposite applies to Fz e A-doublet levels. For a .rr configuration, the preferences are
reversed. The treatment outlined in F is an approximation in which the integration over
only x is considered. A more complete treatment in which the integration over 6 and $ are
included modifies Eq. (50).
32. S. Green and R. N. Zare, Chem. Phys., 7,62 (1975).
33. For example, see R. Vasudev, R. N. Zare, and R. N. Dixon, J. Chem. Phys., 80, 4863
(1984). and P. Andersen and E. W. Rothe, J. Chem. Phys., 82,3634 (1985).
34. J. M. Brown, D. J. Milton, J. K. G. Watson, R. N. Zare, D. L. Albritton, M. Horani, and
J. Rostas, J. Molec. Spectrosc., 90, 139 (1981).
35. R. N. Zare, in Molecular Spectroscopy: Modern Research,K . N . Rao and C. W. Mathews,
eds., Academic Press, New York, 1972, pp. 207-221.
36. E. E. Whiting, J. A. Paterson, I. KovaCs, and R. W. Nicholls, J. Molec. Spectrosc.,47,84
(1973).
37. E. E. Whiting, A. Schadee, J. B. Tatum, J. T. Hougen, and R. W. Nicholls, J. Molec.
Spectrosc.. 80,249 (1980); E. E. Whiting and R. W. Nicholls. Astrophys. J. (Suppl. 235)
27, 1 (1974).
38. It might be thought that the parallel transition classification also applies to 1152 = 0
transitions. However, for = o fi = transitions, this need not be the case; see
L. Brus, J. Molec. Spectrosc.,64,376 (1977).
318 6 ENERGY-LEVEL STRUCTURE AND WAVE FUNCTIONS OF A RIGID ROTOR

APPLICATION 16
MOLECULAR REORIENTATION IN LIQUIDS

We consider here the nature of molecular rotation in liquids. In the limit that the
motion of the molecule can be described as a random walk over small angular
orientations, the Debye small-angle rotational diffusion model[l, 21 is an appro-
priate description. We follow closely Favro's treatment[3] (see also Freed[4] and
Huntress[S]). Let P ( Q ,t) & denote the probability of finding the molecule point-
ing into the solid angle element & = dd sin Bdedx at time t. The orientational
probability at some later time t + A t can be found from P ( Q ,t) dQ by including the
contributions due to all possible rotations AQ that occur in the time increment At:

where f& is the orientation that when rotated by AR gives Q ,and p(A!2, At) is the
probability density of the rotation by AQ taking place during At. The integration in
Eq. (1) can be taken over or AQ, since these two variables are not independent
but are related by Q = AQ + a. It follows that the rotation R(AQ) carries
P(&,t)& into P ( Q , t ) & , that is,

where 8 is the angle conjugate to L which brings to Q.

where

is the second moment, $ ( 8 e ) / ~ t ,or so-called diffusion tensor. Here (. . .)

denotes an ensemble average. Hint: Substitute Eqs. (2) and (3) into Eq. (I),
expand the exponential in the integrand, develop P ( Q ,t + At) in a Taylor series
APPLICATION 16 MOLECULAR REORIENTATION IN LIQUIDS 319

in At, and retain only terms first order in At and second order in 8. Use the fact
that the first moment

vanishes for a random walk.

In terms of its Cartesian components, Eq. (4) becomes

where

Except for a missing factor of i / h on the right-hand side, Eq. (7) is the same as the
Schrodinger equation for a rigid body! Thus all the results already developed for the
general motion of an asymmetric top can be applied to the anisotropic rotation of
molecules in liquids (in the diffusion model limit).
In particular, let us work in the coordinate frame that diagonalizes the diffusion
tensor. Then

and the principal diffusionconstants Di play the same role as 1/(2 I;), where I, is the
ith principal moment of inertia. For a completely anisotropic diffusor Dl # D2 #
D3; for a symmetric diffusor Dl = D2 = Dl and D3 = Dll(DL # Dll);and for a
spherical diffusor Dl = D2 = D3 = D.
The solution to Eq. (7) may be written as

where P ( !&) is the probability density that at some initial time t = 0 the diffusor was
at some initial orientation !& and G(!& IQ ,t) is the conditional probability density
! then at time t it will be found at Q . This latter
that if the diffusor was initially at &
function is a Green's function, which describes the evolution of the motion of the
diffusor from its original orientation; it is subject to the initial condition that

Because P ( Q ,t) satisfies the diffusion equation, so must G(!& IQ ,t). Let &(Q)
be the eigenfunctionsof H with eigenvalues En.Then G(!& IQ ,t) can be expanded
in the & ( R ) since the latter constitute a complete set.
320 6 ENERGY-LEVEL STRUCTURE AND WAVE FUNCTIONS OF A RIGID ROTOR

B. Show that
G(% la,t) = C$'(%)$n(Q) exp(-Ent) (12)
n

Use this result to express the Green's function for a symmetric d i e o r as

Ast -tO,G(%IQ,t) ~ ~ ~ O ~ ~ ~ C ~ , ~ , ~ D ~ K ( ~= O , ~ O
h3(Q - 4 ) and Eq.(11) is satisfied. The Green's function for an asymmetric
diffusor has a more complex form and will not be pursued here[3-51.

The description of many physical properties of molecules in liquids, from light

scattering[6] to fluorescence depolarization[7] to electron spin[4] and nuclear mag-
netic relaxation[5], is most conveniently formulated in terms of time-correlationfunc-
tions (A*[Q(t)] A[Q (0)]) where A is some dynarnical variable that depends on the
molecular orientation and the angular brackets (. . .) denote an ensemble average (see
Application 10). Explicitly

(A*(t + 7)A(t)) = lim

It is always possible to expand A in terms of the symmetric top wave func-

tions rlLKM(Q).Thus we are interested in general in the time-correlation functions
of the form

where Q (t) specifies the orientation of the molecule at time t. These correlation
functions can be evaluated from G ( 4 IQ ,t) according to

C. Show that for a symmetric diffusor having an isotropic initial distribution

NOTES AND REFERENCES 321

As t + oo,the F i K ( t ) approach zero except for F$, ( t) , which approaches

( 1/8r 2 ); that is, the symmetric diffusers become uniformly distributed. The
symmetric diffusor correlation functions FhK(t) for L # 0 and K f 0 are
characterizedby two lifetimes (relaxation times), one equal to [ L( L + 1)Dl] -'
and the other to [ K' ( DI1- DL)]-'. In the most general case of an asymmetric
diffusor, the rotational relaxation is more complex and the fluorescence depo-
larization and light scattering correlation function are both characterized by a
superposition of five exponentials[6,7].

NOTES AND REFERENCES

1. P. Debye, Polar Molecules, Reinhold, New York, 1929.

2. P. S. Hubbard, Phys. Rev., 131, 1155 (1963).
3. L. D. Favro, Fluctation Phenomena in Solids, R. E. Burgess, ed., Academic Press, New
York, 1965, pp. 79-101; Phys. Rev., ll9.53 (1960).
4. J. H. Freed, J. Chem. Phys., 41,2077 (1964).
5. W. T. Huntress, Jr., Adv. Magn. Res., 4, 1 (1970);J. Chem. Phys., 48,3524 (1968).
6. B. J. Beme and R. Pecora, Dynamic Light Scattering, Wiley, New York, 1976.
7. G. G. Belford, R. L. Belford and G. Weber, Proc. Natl. Acad. Sci. USA, 69, 1392 (1972);
T.J. Chuang and K. B. Eisenthal, J . Chem. Phys., 57,5094 (1972); M . Ehrenberg and
R. Rigler, Chem. Phys. Lett., 14,539 (1972); R. Rigler and M. Ehrenberg, Quart. Rev.
Biophys., 6,139 (1973). See also K. Razi Naqvi, Chem. Phys. Lett., 136,407 (1987).
 Appendix

COMPUTER PROGRAMS FOR

3-j, 6-j, AND 9-j SYMBOLS

Simple (inelegant and unoptimized) FORTRAN function subprogram listings are pre-
sented for calculating 3 - j , 6 - j , and 9 - j coefficients[l] using (1) twice the arguments
as integers, (2) real arguments, and (3) integer arguments. The arguments for the 6 - j
and 9 -j coefficients are referred to by matrix element notation. In particular:

RJll RJ12 RJ13

RJ31 RJ32 RJ33

 A COMPUTER PROGRAMS FOR 3 - j , 6-j,AND 9 - j SYMBOLS

JJll

RJll RJ12 RJ13

= {RJ21 RJ22 RJ23 1

In a similar manner we also have

THRJ (JJl,JJ2, JJ3, MM1, MM2, MM3)

A COMPUTER PROGRAMS FOR 3 - j , 6 - j , AND 9 - j SYMBOLS

and

Subroutine SETUP must be called once prior to the first usage of any of these
functions. SETUP calculates the factorials and stores them in a common block used
by the other function subprograms. The size of the arguments for the 3 - j , 6 - j , and
9 -jcoefficients is limited by the size of the array in the common block and by round
off error. Arguments less than approximately 75 are satisfactory.
The Wigner 3 -jsymbol is evaluated using the expression[2]

A ( a , b, C) =
a+b-c)!(a-b+c)!(-a+b+c)!
(a+ b+ c + l)!
]' (A-2)

(-1)''
u!(jl + j2 - j3 - u)!(jl - ml - u)!(j2 + m2 - u)!

and the summation is over all positive integer values of u such that no factorial
in Eq. (A-3) is negative. The Wigner 6 - j symbol is calculated by evaluating the
expression[2]
326 A COMPUTER PROGRAMS FOR 3 - j , 6 - j , AND 9 - j SYMBOLS

where the A functions are defined in Eq. (A-2),

and again the summation in Eq. (A-5) is over all positive integer values of u such that
none of the factorials have negative arguments. The Wigner 9 - j symbol is calculated
by evaluating the sum over the triple product of Wigner 6 - j symbols using Eq. (4.24).
An alternative algorithm for the calculation of the Wigner 9 - j symbol is the numerical
evaluation of its algebraic expression[3]:

where

d ( a ,b, c) =
a - b+ c ) ! ( a +b - c ) ! ( a + b + c + I ) !
(b+ c- a)! ]'
T
(-4-7)
NOTESANDREFERENCES 327

and the summation in Eq. (A-8) is restricted to positive integer values of 7,p , u
for which no factorial has a negative argument. However, there appears to be no
significant gain in computation time using Eq. (A-8)[4].
These function subprograms have been extensively tested using orthonormality
relations, and they have brought us many hours of pleasure.

NOTES AND REFERENCES

Attention is called to the existence of algorithms for evaluating 3-j and 6- j coefficients
[K. Schulten and R. G. Gordon, J. Math. Phys., 16, 1961 (1975)l as well as 9 - j
coefficients [K. Schulten and R. G. Gordon, J. Chem. Phys., 64, 2918 (1976)l based
on the exact solution of recursion relations in a particular order designed to ensure
numerical stability even for large angular momentum arguments. These algorithms
are particularly useful in the frequently occurring case, such as in molecular scattering
calculations, where the evaluation of a whole set of related coefficients is needed. A
computer program for carrying out this task is available elsewhere [K. Schulten and
R. G. Gordon, Computer Phys. Comrn., ll,269(1976)l.
G. Racah, Phys. Rev., 62,438(1942).
A. P. Jucys and A. A. Bandzaitis, Theory ofAngular Momentum in Quantum Mechanics,
(Mokslas, Vilnius, 1977).
Daqing Zhao and R. N. Zare, Mol. Phys. 65, 1263 (1988)
328 A COMPUTER PROGRAMS FOR 3 - j , 6 - j , AND 9 - j SYMBOLS

LIBRARY FUNCTIONS FOR 3-J,6-J,9-J SYMBOLS USED BY THE ZARELAB

FROM CHRIS G R E W (1/5/83 VERSION), JOYCE GUEST (9/27/85 VERSION),
AND ANDREW KUMMEL(8/9/86 VERSION) .
SUBROUTINE SETUP MUST BE CALLED PRIOR TO THE FIRST USAGE OF ANY
OF THESE FUNCTIONS. SETUP CALCULATES THE FACTORIALS AND STORES
THEM IN A C-N BLOCK USED BY THE OTHER FUNCTION SUBPROGRAMS.

FUNCTION XNINEJ(JJll,JJl2,JJl3,JJ21,JJ22,JJ23,JJ31,JJ32,JJ33)
*** CALCULATES 9-J VALUE, EACH JII IS 2X ACTUAL J
IMPLICIT REAL*8 (A-H,0-Z)
=*I RJll,RJ12,rn3,RJ2l,RJ22,RJ23,RJ31,RJ32,RJ33
Jll = JJll
512 = 5512
513 = 5513
521 = 5521
522 = JJ22
523 = 5523
531 = JJ31
532 = 5532
533 = 5533
GOT0 50

ENTRY REALPJ(RJll,RJ12,RJ13,RJ21,RJ22,RJ23,RJ31,RJ32,RJ33)
*** CALCULATES 9-J VALUE, EACH JII IS REAL
Jll = INT (2.*RJ11)
J12 = INT(2.*RJ12)
513 = INT(2 .*RJl3)
521 = INT(2.*RJ21)
522 = INT(2.*RJ22)
523 = INT(2.*RJ23)
J31 = INT (2.*FW31)
532 = INT(2.*RJ32)
J33 = INT(2.*RJ33)
GOT0 50

ENTRY VINT9J(IJ11, IJ12,IJ13,IJ21, IJ22,IJ23,IJ31,1532,1533)

*** CALCULATES 9-J VALUE, EACH JII IS AN INTEGER
Jll = 2*IJll
512 = 2*IJ12
513 = 2*IJ13
521 = 2*IJ21
522 = 2*1J22
523 = 2*IJ23
531 = 2*IJ31
532 = 2*IJ32
533 = 2*IJ33
A COMPUTER PROGRAMS FOR 3 - j , 6 - j , AND 9 - j SYMBOLS

KMINl = KMINl+ 1
KMAXl= IWAXlt 1
XNINEJ = 0 .DO
IF (KMIN1.GT .KMAXl) GOT0 1000
DO 100 K1 = KMINl,KMAX1,2
K-K1-1
S1 = SIXJ(Jll,J2l,J31,J32, J33,K)
52 = SIXJ(J12,522,J32,J21,K, 523)
53 = SIXJ(J13,J23, J33,K, Jll, 512)
P = DFLOAT ( (Ktl)* ( (-1)**K))
XNINEJ = XNINEJ + P*Sl*S2*S3
CONTINUE
CONTIMJE
VINT9J = XNINEJ
REAL9J = XNINEJ
RETURN
END

FUNCTION SIXJ(JJll,JJl2,JJl3,JJ2l,J522,JJ23)
*** CALCULATES 6-5 VALUE, EACH JII IS 2X ACTUAL J
IblPLICIT REAL*8 (A-H,0-2)
RJll,RJ12,Rn3,RJ21,RJ22,RJ23
DIMENSION FL (400)
COMMON/FACTOR/FL
J1D = JJll
J2D = 5512
J3D = 5513
J4D = 5521
J5D = JJ22
J6D = 5523
GOT0 50

ENTRY REAL6J(RJl1,RJ12,RJl3,RJ2lIRJ22,RJ23)
*** CALCULATh:S 6-5 VALVE, EACE J I I IS REAI,
JlD = INT (2.*-I)
J2D = INT(2 .*RJl2)
J3D = INT (2.*RJl3)
J4D = INT(2.*RJ21)
J5D = INT(2 .*RJ22)
J6D = INT (2.*RJ23)
GOT0 50

ENTRY VINT6J(IJ11,IJ12, IJl3,IJ21,IJ22,IJ23)

*** CALCVL&TES 6-5 VALUE, EACE JII IS AN INTEGER
JlD = 2*IJll
J2D = 2*IJl2
J3D = 2*IJl3
J4D = 2*IJ21
J5D = 2*IJ22
J6D = 2*IJ23
 A COMPUTER PROGRAMS FOR 3-j, 6 - j , AND 9 - j SYMBOLS

JM = JMl
IF(iRd2.GT.JM) JM = JM2
IF(JM3.GT.JM) JM = JM3
IF(JM4.GT.m) JM = JM4
JX = JX1
IF(JX2.LT.Z) JX = JX2
IF(JX3.LT.JX) JX = JX3
IW=JM+l
u x = m + i
IF(IW.GT.KX) GOT0 9998
TERM1 = E'RTSJL(JlD,J2D,J3D, J4D, J5D, J6D)
SIXJ = O.DO
DO 10 I1 = m , K x
I = I l - 1
TERM2 = FL (1+2)-FL (I+1-JMl)-FL (I+l-JM2)-FL (Itl-JM3)
T E W = TERMS-FL (I+l-JM4)-FL (JX1-I+1)-FL (JX2-I+1)
TERM2 = TERM2-FL(JX3-I+1)
TERM = O.DO
T E N = DEXP (TENl+TERM2)*DFLOAT ( (-1)**I)
SIXJ = SIXJtTERM
CONTINUE

GOT0 9999
CONTINUE
SIXJ = O.DO
CONTINUE
VINT6J = SIXJ
REAL6J = SIXJ
RETURN
END

FUNCTION THRJ(JJl,JJ2, JJ3,MMl,MM2,MM3)

*** CALCULATES 3-5 VALUE, EACH JI/MI IS 2X ACTUAL JI/MI
IMPLICIT REAL*8 (A-A,0-2)
REAL*4 RJ1,RJ2,RJ3,RM1,WIRM3
DIMENSION FL (400)
COMMON/FACTOR/FL
J1D = JJ1
J2D = JJ2
J3D = JJ3
rnD = mdl
M2D = MK?
M3D = MM3
GOT0 50

ENTRY REAL3J(RJl,RJ2,RJ3,RMl,W,RM3)
A COMPUTER PROGRAMS FOR 3-j, 6-j, AND 9 - j SYMBOLS

*** CALCULATES 3-5 VALUE, EACH JI/MI IS REXG

JlD = INT(2.*RJl)
J2D = INT (2.* W 2 )
J3D = INT (2.*RJ3)
MID = INT (2.*Rbll)
M2D = INT (2.*RM2)
M3D = INT (2.*RM3)
GOT0 50

ENTRY VINT3J(IJl,IJ2,IJ3,IMl,IM2,IM3)
*** CALCULATES 3-5 VALUE, EACH JI/MI IS AN INTEGER
J1D = 2*IJl
J2D = 2*IJ2
J3D = 2*IJ3
W D = 2*1W
M2D = 2*IM2
M3D = 2*IM3

Xl = DFLOAT (JlD)/2.DO
X2 = DFLOAT (J2D)/2 .DO
X3 = DFLOAT(J3D)/2.DO
Y1 = DFLOAT ( W D )/2.DO
Y2 = DFLOAT(M2D)/2.DO
Y3 = DFLOAT (M3D)/2.DO

.
IF (IABS(MID) GT .J1D) GOT0 9998
IF (IABS(M2D).GT.32D) GOT0 9998
IF (IABS(M3D).GT.J3D) GOT0 9998

IF (JlD+J2D-J3D.LT. 0) GOT0 9998

IF (J2D+J3D-JlD.LT.0) GOT0 9998
IF(J3D+JlD-J2D.LT.0) GOT0 9998
IF(MlD+M2D+M3D.NE.O) GOT0 9998

KMIN = (J3D-JlD-M2D)/2
KMINl= KMIN
KMIN2 = (J3D-J2D+WD) /2
IF (KMIN2.LT .KMIN) KMIN = K M I M
KMIN = (-1)*KMIN
RMAX = IFIX(SNGL(X1 + X2 -
X3 + 0.1))
RMAXl= RMAX
IQdlX2 = IFIX(SNGL (XI -
Y1 + 0.1))
EadAX3 = IFIX(SNGL (X2 + Y2 + 0 .l))
IF(KMAXP.LT.I(MAX) RMAX = RMAX2
IF(IWAX3.LT.IWAX) RMAX = KMAX3
IF(KMIN.LT.0) KMIN = 0
IF(KMIN.GT.KMAX) GOT0 9998

JMfN = KMIN + 1
JMAX=WdAX+l
T E W = FRONTL(Xl,X2,X3,Yl,Y2,Y3)
MSIGN = (-1)** ( (JlD-J2D-M3D)/2)
SUM = O.ODO
332 A COMPUTER PROGRAMS FOR 3 - j , 6 - j , AND 9 - j SYMBOLS

1 = 1 1 - 1
TERM2 = FL (11) + FL (KMINl+Il) + FL (KMIN2 + 11)
TERM2 = TERM2+FL (-1-I+1) +FL (-2-I+%) +FL (-3-I+1)
TERM = DEXP(TEREdl-TERM2)
TERM = TERM*ldSIGN* ( (-1)**I)
SUM = SUM + TERW
CONTINUE

THRJ = SUM
GOT0 9999
TIlRJ = 0.0
CONTINUE
VINT3J = THRJ
REAL3J = THRJ
RETURN
END

FUNCTION FRONTL (XI,X2, X3,Y1,Y2, Y3)

***USED IN 35 CALCULATION
IMPLICIT REAL*8 (A-H,0-2)
DIMENSION FL (400)
COMMON/FACTOR/FL

L1 r: IFIX(SNGL(Xl+X2-X3+1.1))
L2 = IFIX (SNGL(X2+X3-X1+1.1))
L3 IFIX (SNGL(X3+X1-X2+1.1))
L4 = IFIX (SNGL(Xl+X2+X3+2.1))
L5 = IFIX (SNGL(Xl+Yl+l.1))
L6 = IFIX (SNGL(XI-Y1+1.1))
L7 = IFIX (SNGL(X2+Y2+1.1))
L8 = IFIX (SNGL(X2-Y2+1.1))
L9 = IFIX (SNGL(X3+Y3+1.1))
L10= IFIX (SNGL(X3-Y3t1.1) )
FRONTL = FL(L1) + FL(L2) + FL(L3) - FL (L4) + FL (L5) + FL(L6)
FRONTL = FRONTL + FL (L7) + FL (L8) + FL (L9) + FL (L10)
FRONTL = FRONTL/Z.DO
RETm
END
--

FUNCTION FRTSJL(JlD,J2D, J3D, J4D, JSD, J6D)

IMPLICIT REAL*8 (A-H,0-2)
***USED IN 65 CALCULATION

FRTSJL = DL(JlD, J2D, J3D) + DL(JlD, JSD, J6D)

FRTSJL = FRTSJL + DL (J4D,J2D, J6D) +
DL (J4D,J5D, J3D)
RETURN
END

FUNCTION DL (JlD,J2D,J3D)
***USED IN FRTSJL/CJ CALCULATION
IMPLICIT REAL*8 (A-H,0-2)
DIMENSION FL (400)
COMMON/FACTOR/FL
A COMPUTER PROGRAMS FOR 3 - j , 6-j, AND 9 - j SYMBOLS

L 1 = (JlD+J2D-J3D) / 2
L 2 s: ( J 2 D + J 3 D - J l D ) / 2
L 3 = ( J 3 D + J l D - J 2 D ) /2
L4 = (JlD+J2D+J3D) / 2 + l
DL = F L ( L 1 + 1 ) +FL ( L 2 + 1 ) +FL ( L 3 + 1 ) -FL ( L 4 + 1 )
DL = D L / 2 .DO
RETURN
END

SUBROUTINE SETUP
***CALCULATES AND STORES FACTORIALS FOR 3 N J PROGRAMS

I M P L I C I T REAL*8 (A-8,O-2)
DIMENSION FACL ( 4 0 0 )
C~ON/FACTOR/FACL
***N CAN BE INCREASED OR DECREASED AS DESIRED

N = 400
FACL ( 1 ) = O.DO
DO 1 0 0 I = 2 , N , 1
11 = 1-1
FACL ( I ) = FACL ( 1 - 1 ) + DLOG(DFL0AT ( 1 1 ) )
CONTINUE
RETURN
END
 INDEX

orbital, 2, 30
polarization, 226
absorption, 200
recoupling coefficient, 173
absorption matrix element, 203-204
rotational, 70
absorption oscillator, 125-127, 133
sharp, 180, 196,200
abstract vector space, 6
spin, 2
active rotation, 78, 104, 107
tensor, 226228,230-232
active transformation, 75, 104
transfer, 233-236,242
addition of angular momenta, 143-176 vector, 12
addition theorem, spherical harmonic, 95-98 wave function, 6-1 2
adjoint, 5, 185,202 angular momentum distribution, 21 1, 214,
alignment, 119,226242 226-242
analytic continuation, 38 angular momentum operator, 1-41,73-77
angle angular momentum operator representation
dihedral, 16, 19 molecule-fixed frame, 8 1-85
magic, 207 space-fixed frame, 8 1-85
polar, 18 angular velocity, 255,258-260,263
angle of closest approach, 26 angular velocity vector, 263,294
angular correlation, 64,223-225 anharmonic effect, 278
angular distribution, 117, 120-122, 191,221, anisotropic diffusor, 3 19
224 anomalous commutation relation, 83, 85,
angular frequency, 139-140 106-107, 179-180,271
angular momentum, 1 4 1 , 2 5 5 antisymmetric tensor, 216
addition of, 143-176 apex half-angle, 13
conservation of, 24,77,220,233,262 Ar, 34,138
coupling of more than two, 143-176 arrow notation, 151-153
coupling of two, 43-72 associated Legendre function, 8
eigenfunction, 6-12 asymmetric diffusor, 320-321
general, 2 asymmetric top, 105, 257, 262-263, 265,
graphical method, 172, 175-176 267-269,271-272,274-277,283,
matrix element, 3-6 287-288,291,295,297,3 19
operator, 1-41,73-77 asymmetric top symmetry species, 271
335
 INDEX

asymmetry parameter, 121,274,276,296 Breit-Wigner formula, 39,41

atom, 18, 46, 66-69, 91, 110-112, 172,
190-191,196,200,204,212-213,
218-219,223-224,232,235,238,
242-243,248-249,283,289,293 Cartesian tensor, 216-2 17
many-electron, 66, 111 cascade, 224-225
two valence electrons, 111-1 16, 196 C-CI bond, 249
atomic orbital, 66 CC14, 138,257
atomic term, 66,112-1 13, 115-1 16,196 center of mass, 254-255,258,294
atomic term energy, 198,218-219 center-of-mass frame, 23
atomic term separation, 114 central field approximation, 111-1 12
attractive potential, 26,3 1-32 central field Hamiltonian, 111-1 12
axial recoil approximation, 120, 122 central force field, 27,29
axial symmetry, 230-232, see also cylindrical central force potential, 23
symmetry centrifugal barrier, 28,40
axial vector, 184 centrifugal distortion, 70,279,281-283,299,
axis of quantization, 13 30 1
azimuthal angle, 8 centrifugal distortion Hamiltonian, 282-283
centrifugal potential, 25.3 1
CH2C12,257
CH2 F2, 295
background phase shift, 37,40 CH3 (3,249,257,262
band, 15-16 CHq, 129
band origin, 70 chaotic limit, 230
band strength factor, 289-290 character, 109-1 10,268
beam stop, 92-93 character table, 110
beam-gas scattering, 230,232 charge density, 246
bending mode, 292 charge distribution, 179, 190, 243, 245, 249,
Bessel function, 19 309
Biedenharn-Elliott sum rule, 176 CHC13, 138,257,261
bimolecular reaction, 191 circle, 27, 127,263
binomial expansion, 87 circular orbit, 15, 17
bipolar harmonic, 190-191 circular polarization, 208-21 1, 233, 235
closure condition, 191 degree of, 210-211,213,215
definition, 190 CI, 244,249-250
orthogonality, 191 Cl-, 249
block, 161, 163-164 class, 109-1 10
body-fixed frame, 77,79,84,258-259,265 classical mechanics, 1, 18, 23, 43, 84, 106,
Bohr condition, 136 253-266,29 1
Bohr frequency, 200 classical orbit, 13, 15
Bohr magneton classical orbiting, 28
electronic, 7 1,91, 133, 184 classically allowed region, 55,91-92, 174
nuclear, 184 classically forbidden region, 55, 91, 174
bond angle, 295 CICN, 243-245,248-250
bond length, 295 Clebsch-Gordan coefficient,4343,117,145,
Born-Oppenheimer approximation, 277,288 151,153-154,177,180-181,222,
boson, 179 236,307
bound state, 38 algebraic expression, 5 1,57
Breit level-crossing formula, 202,213 explicit expression, 49,64
INDEX

geometric interpretation, 51-55 contact transformation, 277, 279-283, 292,

graphical representation, 153-154 299,310-311
matrix of, 45 continuum state, 40
orthogonality relation, 45 contraction, 186-1 87, 190, 194,246
orthonormality, 144, 187,237 contragradience, 106
phase convention, 49 correlation, 172,191,212,220,223,225,311,
recursion relation, 4 8 4 9 320-321
symmetry relation, 49-50 more than two vectors, 191
two arguments vanish, 57, 154 two-vector, 191
Clebsch-Gordan series, 98-103, 117, 177, correlation function, 136-141, 321
180, 187 unit vector, 137
geometric interpretation, 99-100 correspondence, 12, 13, 18,65
inverse, 100, 1 17 counterclockwiserotation, 77, 84
clockwise rotation, 84 coupled and uncoupled representations, 144
closed diagram, 157-158, 161, 164 coupled representation, 44, 46-47, 70, 100,
closed shell, 29, 112,249 113, 115,144, 186,196,240
cluster, 266 coupled state, 52
CN, 249 geometric interpretation, 52
CO, 137-138 coupling scheme, 244,249,298
C02,257 Hund's case (a), 297-298,3 15
coherence, 200,203,229 Hund's case (b), 304.3 15
coincidence measurement, 191 Hund's case (c), 3 15
collision complex, 221 Hund's case (d), 3 15
collision process, 23 covalent, 249
combination band, 290 cross product, 188,215,255,257
commutation relation, 2-3,77,83, 179 cross section,
anomalous, 83, 85, 106-107, 179-180, differential, 27-3 1,34,36-37, 191,220,
27 1 222
commutation rule, 2-3,6,43,84, 107 partial, 40
total, 27.32-34, 37, 121, 191,233-234
normal, 82
crossed beam, 230,232
commutator, 2, 18,43, 1 12
crossed polarizer-analyzer pair, 94, 108
complex plane, 38
crystal field splitting, 110
complex variable, 38
Cs, 34
component, 185
current loop, 184,198
compound spherical tensor, 196
cyanogen chloride, 243
Condon approximation, 289,293
cylindrical symmetry, 230,232,238, see also
cone, 13,260-263
axial symmetry
configuration, 111-1 16, 198-199,218
configuration interaction, 116
conjugation operator, 106-107
connected line summation convention, 155- d orbital, 179
158 D2 group, 268-27 1,273-274,288
conservation law, 73,75 damping factor, 201
conservation of angular momentum, 24, 77, Darstellung, 108
220,233,262 de Broglie wavelength, 35,37
conservation of energy, 24,75,262 deflection angle, 23,26-28,33,36-37
conservation of linear momentum, 76 deflection function, 28
constant of the motion, 24 deflection pattern, 27
 INDEX

degenerate bending motion, 292 directed line, 151

degenerate electronic state, 292 direction correlation, 220
degenerate vibrational mode, 292 direction cosine, 79-81, 130-13 1,259
degree of circular polarization, 2 10-2 11,213, direction cosine matrix, 80, 128
215 direction cosine matrix element, 79-81, 98,
degree of linear polarization, 125, 134, 141, 125-127,178,269
203,207-208,210,213 direction cosine transformation, 82
delta function directional correlation, 220,223
Dirac, 136,320 direction-direction correlation, 223-225
A K = 0 transition, 286 displacement operator, 76
A K = f1 transition, 286 dot product, 188,215,255,257
A h = 0 transition, 3 15 generalized, 188
A h = f1 transition, 3 15 double Stern-Gerlach experiment, 92-93
A m = 0 transition, 224 double-headed arrow, 226
A m = f1 transition, 224 dynamical information, 141, 180, 191, 230,
A R = 0 transition, 317 309
+
A vk = 1 transition, 290 dynamics, 180, 191,230,236,266,309
density matrix, 21 1, 214, 225-242
depolarization, 240,320
depolarization ratio, 129
determinantal wave function, 111,307 e level, 308,311,316-317
diagram, 154, 157, 159,161,163-164, 175 effective Hamiltonian, 111,28 1,299
diatomic molecule, 46, 66-67, 70-72, 107, effective potential, 25,40
120, 124, 141,212,266,297-317 eigenfunction, 3-4, 11,266
homonuclear, 6 9 , 3 16 eigenvalue, 3-7,230,266
diatomic molecule dissociation, 66, 120-122 Einstein coefficient for spontaneous emission,
differential cross section, 27-3 1, 34, 36-37, 284
191,220,222 elastic scattering, 23, 26, 29
differentiation electric dipole, 189
chain rule for, 7 electric dipole moment, 186185,246
differentiation of a definite integral, 36 electric dipole moment matrix element, 293
diffraction, 33 electric dipole operator, 184, 201, 205, 284,
diffusion constant, 3 19 289,306
diffusion equation, 3 19 electric dipole photon, 223-225
diffusion model, 3 18-3 19 electric dipole radiation, 102
diffusion tensor, 3 18-3 19 electric dipole transition, 120, 185, 200,233,
diffuser, 3 19 238-239,243,283-284,288,290,
dihedral angle, 16, 19,53-54, 173,222 306
dilute absorber approximation, 130-131 electric dipole transition moment, 125
dipole, 226,232 electric dipole transition operator, 132
dipole antenna pattern, 224 electric field, 108, 123-124, 128, 190, 201-
dipole correlation function, 137-141 202,204,209,243
Fourier transform of, 137 electric field gradient, 190,243,246247,249
dipole moment, 128, 184-185, 190,290 electric field gradient coupling constant, 247-
dipole moment matrix element, 233 248
dipole moment operator, 184-1 85 electric field gradient tensor, 246
dipole operator, 178 electric hexapole focusing, 124
Dirac delta function, 136,320 electric moment, 184
direct product, 64, 177, 180, 187 electric quadrupole moment, 243,246247
 INDEX

electric quadrupole transition, 185 factorial, 86,325-327,333

electric vector, 125, 128, 133,223,230,284 FCI, 249
electrical anharmonicity, 290 fermion, 64, 179
electric-dipole-allowed transition, 3 13, 3 15 figure axis, 77, 106,260-261, 266,282,286
electron charge cloud asymmetry, 308 finite point group, 108, 110,295
electron charge density, 245 first moment, 3 19
electron charge distribution, 246-248, 308- first-order perturbation theory, 7 1-72, 123,
309,3 11 199,219,248
electron density, 249 fluorescence, 200,223
electronic Bohr magneton, 7 1,91, 133, 184 fluorescence depolarization, 321
electronic configuration, 111 fluorescence matrix element, 203, 205
electronic orbital angular momentum, 297, flux, 27,30,32,36
309 football, 257
electronic spin angular momentum, 297 forbidden transition, 293
electronic state, 66, 69 forward scattering, 28-29
electronic term, 66 forward-backward symmetry, 22 1-222
electronic wave function, 69, 288 Fourier transform pair, 141
ellipse, 264 four-photon process, 131
ellipsoid, 263-264 Franck-Condon factor, 289,293
ellipsoid of inertia, 258 free electron, 72
elliptical polarization, 209 free line, 155, 163-164
emission, 200, 223 free radical, 70-72
emission oscillator, 125-127, 133 frequency, 284
energy, 24,38 frisbee, 257
conservation of, 24 fundamental transition, 290
kinetic, 24
potential, 24
energy difference measured in half-widths, 40
ensemble average, 137,230,232,318, 320 g, 69
entropy, 230 g factor, 9 1, 184
equator, 16,221,266 7-ray, 223,289
equilibrium configuration, 277,290 Gaussian, 141
equivalent electrons, 113, 199 general angular momentum, 2
equivalent representations, 109 generalized dot product, 188
Euler angle, 77-80, 82, 91, 104, 120, 269, generator of an infinitesimal rotation, 73-77,
297,308 81
even moment, 246 generator of an infinitesimal transformation,
even multipoles, 226 74-75
excitation-detection geometry, 128, 133-1 34, geometric interpretation, 10,54,91-95, 173
200,20&207,2 10 ($, 0 , ~91-95
) ~
expectation value, 10, 20 d&,,(0), 91-95
exponential of an operator, 74 gerade, 69
external line, 157, 161-162 glory scattering, 28.34
glory undulation, 34
good quantum number, 2,66,250,268
gradient, 91,94
f level, 308,311,316 gradient of the electric field, 190
FI level, 301,303-304,309,3 13,316 gradient operator, 1
FZ level, 301,303-304,309,3 I 3 , 3 I 6 graphical method, 151-168, 176, 194, 213
 INDEX

graphical representation of Clebsch4ordan identity operator, 195

coefficient, 153-154 impact parameter, 24-28,30-33,37,220
graphical representation of 9 -jsymbol, 160- impenetrable sphere, 27
161 inelastic scattering, 226
graphical representation of 6 - j symbol, 158 inequivalent electrons, 114
graphical representation of 3 -j symbol, 151 inertia defect, 296
great orthogonality theorem, 108 inertia tensor, 256
Green's function, 3 19-320 infinitesimal rotation, 73-77
group theory, 6445,107-108,110,127,177, infinitesimal transformation, 74
187,212,225,268,288,292,306 infrared spectrum, 141
inhomogeneous magnetic field, 91-92,94
instantaneous axis of rotation, 255
intensity, 200, 206, 216, 3 15
H, 69 interference, 3 1-33,40, 131,202-203
half-width at half-maximum, 39 internal angular momentum, 107,292
Harniltonian, 70,75, 190,241,246,278 internal diffraction, 289-290.293
Hanle effect, 133-1 35,202 internal line, 157
harmonic internuclear axis, 66-67, 122, 248-249, 297,
bipolar, 191 300,307,3 15
n-polar , 191 internuclear distance, 308
harmonic motion, 277 invariable line, 262
harmonic oscillator wave function, 278,290 invariable plane, 262-263
Heisenberg uncertainty principle, 18 inverse Clebsch-Gordan series, 100, 117
helicity, 208, 210,223,226 inversion, 67-69, 112, 184,305-308
Hermitian, 3.5, 12, 1 8 , 7 4 7 5 , 229,256,279 ionic, 249
Hermitian conjugate, 185 irreducible representation, 64, 10&110, 177,
herpolhode, 263 180
Hertzian dipole oscillator, 125, 128, 133 isotope, 243
hexadecapole, 226
Hg, 34
Honl-London factor, 285-286,293,3 13
homogeneous magnetic field, 92 Jahn-Teller effect, 292
homonuclear diatomic, 69,3 16 Ijm), geometric interpretation, 13
Hund's case (a) coupling scheme, 297-298, Ij m ) representation, 3
304,312,315
Hund's case (b) coupling scheme, 304-305,
315
Hund's case (c) coupling scheme, 305,3 15 K, 34
Hund's case (d) coupling scheme, 305-306, K stack, 283
315 K-I K I label, 272
hydrogen atom, 258 KCI, 249
hyperfine splitting, 241,248 kinetic energy, 24
hyperfine structure, 200, 208, 240, 244, 249, Kr, 34
297.3 16 Kronecker delta, 5
K-type doubling, 267,300

ICI, 249
identity matrix, 73-74,280 ladder operator, 4
INDEX

A component, 3 11.3 13 liquid, 3 18-32 1

A doubling, 300 long-lived complex, 220-221
2S+1Ag. 69 longitude, 16
s+lA,, $69 loop, 158
A-doublet, 3 17 Lorentzian shape, 135
A-doubling parameter, 3 11 lowering operator, 3-4, 107
A-type doubling, 67,300,305,310-312
A-type splitting, 3 10
Land6 factor, 133
Land6 interval rule, 199 m quantum number, 3
Larmor precession, 133, 135 magic angle, 207
Larmor precession frequency, 133-1 35 magnetic dipole, 7 1
latitude, 16 magnetic dipole moment, 184-1 85
law of cosines, 53,55, 174
magnetic dipole operator, 184
left circular polarization, 209-2 11, 233
magnetic dipole transition, 185
Legendre function, 222
magnetic field, 71-72,91-92, 133-135,202
associated, 8
magnetic moment, 91, 133, 184-185
Legendre polynomial, 9-10, 14, 30, 32, 101,
magnetic quadrupole transition, 185
114, 123,220
magnetic relaxation, 320
Legendre series, 221,223
magnetic resonance, 93-95, 108,320-321
level crossing, 202
magnetic resonance spectroscopy, 95
lifetime, 40,321
light scattering, 141,320-321 many-electron atom, 66
line mass spectrometer, 92
directed, 151 matrix representation, 10,20,73, 108-1 09
external, 157, 161-162 mechanical anharmonicity, 290
free, 155, 163-164 metastable state, 40
internal, 157 methylene fluoride, 295
undirected, 151 microwave spectroscopy, 250-251,277,292
zero, 155, 161-163 microwave spectrum, 243,295
line of nodes, 77-78,80,258,260,297,309 mixed state, 21 1,228,230
line profile index, 40 modified spherical harmonic, 183, 190, 218,
line strength, 297 245
line strength factor, 283-290.3 12-315 molecular beam magnetic resonance, 91
linear, 257 molecular beam magnetic resonance spec-
linear molecule, 124, 139, 141,249,266,292, trometer, 92-93
297 molecular beam scattering, 2 11
linear momentum, 1,28,33,76,84,255,282 molecular polarizability tensor, 128
linear momentum transfer, 28 molecular relaxation, 141
linear polarization, 120, 125-1 26, 128-1 29, molecular reorientation, 3 18-321
133,200,206-208,211,230,233, molecule, 69-72, 106, 124, 128, 136-141,
235 190-191,200,204,211,223,232,
degree of, 125,134,141,203,207-208, 235, 238-239, 243, 245, 248-
210,213 251,253,257-258,265,277,283,
linear top, 257 28&289,292-293,295-321
linear transformation, 73 molecule-fixedframe, 78-79,81-85,98, 107,
linear triatomic molecule, 243 120, 128, 179-180,248,253,284,
lineshape factor, 136 297,308
lineshape function, 136 moment, permanent, 184
342 INDEX

moment of inertia, 70, 139, 253-258, 265, node, 151,157,159-161,164-165,224

282,294-296,3 11,319 nonlinear molecule, 292
moment of inertia ellipsoid, 263,268 nonrigid behavior, 70,277-283
moment of inertia tensor, 256 nonrigid top, 266
momentum normal commutation relation, 179
angular, 1 normal commutation rule, 82, 106
angular operator, 2 normal diagram, 163-164
linear, 1 normal form, 161,163-164
momentum transfer, 28 normal mode, 110,278,289-290
monopole, 226 n-polar harmonic, 191
most asymmetric top, 265 nuclear Bohr magneton, 184
multiphoton process, 131 nuclear charge density, 245
multiplet splitting, 102 nuclear charge distribution, 245,247
multiplication, 186-1 87 nuclear quadrupole interaction, 243-25 1
multiplicity, 66 nuclear quadrupole moment, 243-25 1
multipole expansion, 178,218,226 nuclear quadrupole tensor, 246
multipole moment, 179, 184-185, 225-226, nuclear rotational angular momentum, 297
230-232,237,239-241 nuclear spin, 244,247,297,316
multipole moment expansion, 245 nuclear spin angular momentum, 239
nuclear spin statistics, 316
nucleus, 106, 111-112, 123, 185, 212, 223,
225,243-247,249,277,289,292,
N, 244,250 305
N atom, 249
1 4 ~ 269
,
I4N I5N, 69
natural excitation, 284 02,69
natural lifetime, 200 oblate symmetric top, 257, 266-268, 272,
natural linewidth, 202-203 274,288
n-C7HI6,138 octahedral symmetry, 266
near-oblate top, 258,269 octupole, 226
near-prolate symmetric top, 296 octupole moment, 190
near-prolate top, 257-258,269 odd multipoles, 226
near-symmetric top, 258 OH, 69,309
9 - j coefficient, 189 0 component, 3 17
9-3 symbol, 143-150, 161, 189, 194195, 0 = 0 state, 305
240 0 = ++ 0 = transition, 317
algebraic expression, 326 R -type doubling, 305
computer program, 150 one-electron operator, 196, 198
graphical representation, 160-16 1 open shell, 297
numerical calculatios, 323-333 operator
one argument vanishes, 150, 166-167 ladder, 4
orthonormality, 149,240 lowering, 4
orthonormality relation, 327 mean value, 228
recursion relation, 327 raising, 4
relation to the 3 -j and 6 - j symbols, 150 shift, 4
symmetry relation, 149, 161 step-down, 4
n-j symbol, 150 step-up, 4
NO, 309 optical pumping, 200,213
INDEX

optical spectroscopy, 297 resonance, 37

orbit photodissociation, 120,211,226,231
circular, 15, 17 photoejection process, 191
classical, 13, 15 photoelectron, 232-233
orbital, 111 photofragments, 226.23 1
orbital angular momentum, 2, 6, 30,66, 181, photoionization, 226,231-233,235-236
220 photoionization cross section, 233
order of the group, 109 photoselection, 131-132
orientation, 123,226-242 n configuration, 3 17
orientational probability, 3 18 n orbital, 3 17
orthogonality relations, 81 n - n transition, 315
overtone transition, 290 rI - Z transition, 3 15
l3 state, 66,300,3 10
TI state, 299-300,306,3 I 0,3 16
P branch, 127, 141,285-287,313,315
'n - 'C transition, 3 13-3 14
'I3 state, 297,300,302,304,306,309-3 13
p character, 249
inverted, 304
parallel-type transition, 120, 122, 286, 3 15,
regular, 304
317 211+state, 304,312-3 13
paramagnetic, 7 1
parity, 67-68, 112, 114, 116, 184-185, 'n; state, 304,3 12-3 13
198,273,300-302,305-306,308, 7r3 configuration, 317
310-311,313,315 planar molecule, 296
parity basis set, 306 Planck's constant, 284
parity block, 303,308,3 10,313 plane of reflection, 23 1
parity violation, 185 plane of rotation, 297-298,3 15-3 17
partial cross section, 40 plane of symmetry, 258
partial wave, 30,32-33,37 plane polarization, 108, 1 17, 1 19-120, 125-
Paschen-Back effect, 72 126,128-129,133-134,200,231,
passive rotation, 104, 107 315
passive transformation, 75, 104 plane wave, 30
Pauli exclusion principle, 64, 111, 113 Poinsot construction, 262
Pauli spin matrix, 11 Poinsot ellipsoid, 263
permanent moment, 184 point group, 108,212,269,292,295
permutation, 15 1 point of closest approach, 24
perpendicular-type transition, 120, 122, 286, polar angle, 8, 18
315 polar coordinate, 24
perturbation, 92,94, 123,278-279,28 1,301, polar vector, 184
316 polarizability tensor, 128
perturbation theory, 123,225,253,279 anisotropic part, 129
first-order, 7 1-72, 123 isotropic part, 129
second-order, 124,279,281 principal axes, 129
phase, 9 polarization, 191, 200-201, 203, 206-208,
phase convention, 5, 9, 18, 49, 64, 90, 108, 212-213,223,233,236
181,301 degree of, 125,134,141,203,207-208,
phase information, 230 210-211,213,215
phase relation, 117 polarization anisotropy, 125, 141,207
phase shift, 31-32,3438 polarization tensor, 204,206, 209
background, 37 polarization vector, 209
 INDEX

polarized light, 125-132 quasi-degenerate, 28 1

pole, 16,38-39,221,264
polhode, 263
position operator, 178
positive rotation, 76,78 R branch, 127,141,285-287,313,315
potential Racah coefficient, 145
attractive, 31-32 radial wave function, 111
long-range, 29,33-34 radiationless transition, 291
repulsive, 3 1-32 radiative cascade process, 224-225
short-range, 33 radiative lifetime, 127, 133,200,202
potential energy, 24 radiative transition, 185,238, 253
potential energy function, 292 radical, 309
power series expansion, 74-75, 86 radiofrequency field, 92,94
pressure broadening, 244 rainbow deflection angle, 28
principal axis, 258-259,262, 268-269, 296 rainbow scattering, 28
principal axis of inertia, 264 raising and lowering angular momentum
principal inertial axis system, 248 operator, 84, 179
principal moment of inertia, 256-257 raising and lowering operator, 3 4 8 4 , 107
principal-axis system, 257 raising operator, 3-4
probability, 10, 15-16,27,52,55,91-92.99- Raman effect, 128
100, 117, 120, 140, 173, 191,225, Raman scattering, 125-1 32,292
229,238,3 18 Raman spectrum, 141
probability amplitude, 51-52,91, 100, 117 random walk, 3 18-3 19
probability density, 10, 16-17,54, 174,318- rank, 177, 185, 188,216, 231
319 Rayleigh scattering, 128-129
product of inertia, 256 RE surface, 265-266
projection operator, 204-205 reactive scattering, 220,226
prolate symmetric top, 247, 257, 265-268, recoil distribution, 121
272,274,28 1-283,288 recoupling coefficient, 143-176, 189
propagation vector, 30 recurrence time, 291
pseudohalogen, 249 reduced matrix element, 181-182, 194-195,
pseudovector, 184 197,205,231,240
pure rotation spectrum, 243, 249, 288,296 reducible representation, 64, 109, 177
pure state, 21 1,228-230 reflection, 306-308,3 16
reflection basis set, 67
reflection plane, 67
reflection symmetry, 67
Q branch, 127,285-287,313,315 region
quadrupole, 226,232 classically allowed, 16,55
quadrupole coupling constant, 248 classically forbidden, 16,55
quadrupole moment, 179, 184-185, 190 relative velocity, 220
quadrupole moment operator, 184-1 85 relative velocity vector, 230
quadrupole splitting, 248 relativistic interaction, 198
quantum beat, relaxation time, 291,321
rotational, 21 3 Renner-Teller effect, 292
Zeeman, 133-135,202 representation, 20,64,73, 108-109
quantum mechanical tunneling, 40 coupled, 4 4 , 4 6 4 7
quantum mechanics, 2, 18,29,35,41,43,84, irreducible, 64
106-107 reducible, 64
INDEX

uncoupled, 44,46-47 rotational kinetic energy, 257,259,262

representation basis, 108 rotational line strength, 102, 107, 283-290,
representation dimensionality, 108 312-315
repulsive potential, 26, 3 1-32 rotational partition function, 140
resonance, 3 7 4 0 rotational quantum beats, 213
resonance condition, 92 rotational quantum number, 70,278
resonance fluorescence, 125-132, 199-21 1, rotational relaxation, 321
215-216 rotational tilt, 191
magnetic depolarization, 133-135 rotational transformation, 76
resonance phase shift, 37,40 rotational wave function, 287-288,297
resonance scattering, 37 rotation-vibration band, 137
reversed angular momentum method, 107- Russell-Saunders coupling, 102, 196
108 Rydberg state, 305
rf field, 92-94
rf frequency, 92
right circular polarization, 209-2 1 1,233
right-handed screw, 76 S term, 199
rigid body, 104-105,253-321 scalar, 161,178, 181,191,204
rigid linear rotator, 105 scalar contraction, 188-190, 194,204,246
rigid rotor, 70, 117,243,247,253-321 scalar operator, 178, 188- 190, 194-195
rotation, 91-92, 177,307 scalar product, 144, 181, 188, 190
rotation matrix, 85-9 1, 132, 180 scalar quantity, 96
algebraic expression, 88-89 scattered wave, 3 1
explicit expression, 86-88 scattering amplitude, 30-3 1
geometric interpretation, 91-95, 106 scattering center, 26,30,35
integrals over products, 98-103 scattering plane, 23.25
rigid body rotational wave function, scattering resonance, 38-40
104-106 scattering theory, 2 3 4 0
sum rule, 90 Schrodinger equation, 3 19,266
symmetry relation, 88-90 second-order perturbation theory, 124, 279,
unitary, 179 28 1
rotation operator, 77,85 second-rank Cartesian tensor, 128, 130, 132,
rotational angular momentum, 70,239,243 216217,256
rotational branch, 207, 210-21 1, 285, 313- secular determinant, 67, 71-72, 248-250,
315 268,27 1,273-274,295,300,303,
rotational constant, 70, 124, 243, 258, 281, 306,311
295,301 secular equation, 67
rotational degeneracy factor, 140 secular matrix, 198,312
rotational degrees of freedom, 297 segment, 15-16
rotational diffusion constant, 3 19 selection rule, 136, 180, 185, 198,288
rotational diffusion equation, 3 19 self-adjoint, 256
rotational diffusion model, 3 18-3 19 semiclassical limit, 16
rotational diffusion tensor, 319 semiclassical phase shift, 35-36
rotational energy, 139 semiclassical treatment, 13-1 5, 18-19, 34-
rotational energy level, 70-72,253,3 15 40.65, 108, 174,222-225,277
rotational energy surface, 265 sharp angular momentum, 180, 196,200
rotational Hamiltonian, 266, 275, 281, 298- shift operator, 4
299 C - l7 transition, 3 15
rotational invariance, 96 C - transition, 3 15
346 INDEX

x state, 66-67,266,297,300,302,307,310, modified, 183,245

312 spherical harmonic addition theorem, 95-98,
C - state, 67,69,305-307 114, 190-191,221-223
C + state, 67,69,305,307 spherical irreducible tensor, 132, 177-25 1
'x state, 70,298,300,306 spherical polar coordinates, 6,77,95
'x- state, 124, 310-311 spherical tensor, 116, 177-25 1
'x+ state, 124,310-31 1 composite, 192
'x 'n
* - transition, 3 13 compound, 196
'x * - 'I;* transition, 3 13 spherical tensor adjoint, 185
2x 2n- transition, 3 13-3 14 spherical tensor contraction, 186-1 87
2Z - 'C transition, 313,315 spherical tensor matrix element, 21 3
2Z state, 70-72,297,300-301,306,3 13,316 spherical tensor multiplication, 186-1 87
2C state, 3 11 spherical tensor product, 186-191
2Z state, 3 11
+ matrix element, 192-2 14
single-headed arrow, 226 spherical top, 257-258,266
6 - j symbol, 143-150, 161, 174-175,222 spherical trigonometry, 19
algebraic expression, 148,325 spherical wave, 30
Biedenharn-Elliott sum rule, 176 spin, 2,66,226
computer program, 149 spin angular momentum, 2,6,66, 181
geometric interpretation, 173-1 76 spindown, 11, I l l , 307
graphical representation, 158 spin eigenfunction, 10
numerical calculation, 323-333 spin function, 86-87, 111, 114
one argument vanishes, 148, 166 spin projection, 307
orthonormality, 146, 167-168 spin space, 11
orthonormality relation, 327 spinup, 11, 111,307
recursion relation, 176,327 spinor, 11-1 2
relation to 3 - j symbol, 146-148 spin-orbit constant, 199, 3 11
symmetry relation, 145-146, 158-160 spin-orbit coupling term, 304
Slater determinant, 111 spin-orbit Hamiltonian, 196, 198, 219
Slater-Condon parameter, 115, 218 microscopic, 3 16
small-angle scattering, 28-29 phenomenological, 3 16
solid angle element, 10,27, 117,220,318 spin-orbit interaction, 196-199, 219, 300,
space-fixed frame, 77-79, 81-85, 98, 107, 302,305
120,128,179-1 80,253,259,265, spin-orbit parameter, 2 19
284,308 spin-orbit splitting, 198
spatial average, 20 spin-orbital, 111, 115, 198
spectral density, 141 spin-rotation constant, 70,301-302
spectroscopy, 136 spin-rotation coupling term, 300
speed of light, 284 spin-rotation splitting, 7 1, 30 1
sphere, 10, 16,35, 264 spontaneous emission, 201
impenetrable, 27 spontaneous transition, 201,238-239
spherical basis, 98, 188,216-217 stable rotation, 291
spherical basis set, 178,209 Stark effect, 124-125, 184, 243
spherical diffuser, 3 19 state multipole, 236-237
spherical harmonic, 8-10,66,68,70,96, 105, static moment, 184-185
111,114,117,121,177-179,183- statistical weight, 228
186, 190,221,237,245,306 step-down operator, 4
geometric interpretation, 13-17 step-up operator, 4
integral of the triple product, 102, 183 Stem-Gerlach experiment, 91
INDEX

double, 92-93 time reversal transformation, 84

triple, 95, 108 time-dependent Schriidinger equation, 84
Stem-Gerlach magnet, 92-94 top
subbranch, 288 asymmetric, 257, 262-263, 265, 267-
sub-Doppler spectroscopy, 203 269,271-272,274-277,283,287-
sum of powers of integers and half-integers, 288,291,295,297,319
20-22,47-48 classical motion, 258-266
sum rule, 176,285,287.3 13,315 linear, 257
super hyperfine splitting, 249 most asymmetric, 265
superfine structure splitting, 266 near-oblate, 258,269
surface, 2 11,309 near-prolate, 257-258,269
symmetric diffuser, 320-32 1 near-prolate symmetric, 296
symmetric top, 105-106, 123-1 25,258-259, near-symmetric, 258
263, 266-268, 279, 284, 286, nonrigid, 266
296297,320 oblate symmetric, 257, 266-268, 272,
symmetric top wave function, 105-106, 123, 274,288
278,287 prolate symmetric, 257, 265-268, 272,
symmetric traceless tensor, 216 274,281-283,288
symmetry operation, 108-1 10 spherical, 257-258,266
symmetry relations, 194 symmetric, 258-259, 263, 266-268,
Taylor's series expansion, 75,3 18 279,284,286,296297,320
temperature, 139 top flop, 291
tetrahedral symmetry, 266 top motion, 262-266,29 1
tetrahedron, 173 stable, 264-265, 291
three-dimensional rotation group, 108, 177- unstable, 264-265.291
178 top type, 253
3 - j symbol, 48-63, 145-147, 149-151,153- total angular momentum, 244,255,259,279,
154, 157, 164, 176, 194 297
algebraic expression, 5 1,62,65,325 total angular momentum excluding electronic
algorithm for evaluation, 5 1,64 spin, 297
computer program, 5 1 total cross section, 27.32-34,37, 121
geometric interpretation, 5 1-55 velocity dependence, 33-34
graphical representation, 15 1 anisotropy, 191
numerical calculation, 323-333 trace, 109.2 16,229-230
orthogonality relation, 157-1 58 trajectory, 23-25,33,35,92, 265-266
orthonormality, 147 transformation
orthonormality relation, 327 active, 75
recursion relation, 327 infinitesimal, 74
symmetry relation, 50-5 1, 151 linear, 73
two arguments vanish, 62, 154 passive, 75
three-photon process, 131 rotational, 73-141
time correlation function, 136-141, 320 time reversal, 84
time development operator, 76 unitary, 44,73-75,78,80-81, 108-109
time displacement operator, 75 transition
time evolution operator, 76,240 forbidden, 293
time invariance of the system, 75 vibronically induced, 293
time reversal, 84, 106 transition dipole, 288
time reversal operator, 84, 106-107 transition dipole moment, 120, 289, 315
time reversal symmetry, 85 transition 224
 INDEX

triangle, 54, 158-159, 166-167, 175 vibrational angular momentum, 292

area of, 54 vibrational coordinate, 278
triangle condition, 45, 146, 158, 198 vibrational degrees of freedom, 277
triatomic molecule, 292,295-296 vibrational energy, 278
triple Stem4erlach experiment, 95, 108 vibrational frequency, 277
tunneling, 40,266 vibrational overlap integral, 289
turning point, 26 vibrational quantum number, 278,289-290
two-electron atom, 111-1 16, 196 vibrational state, 70
two-photon absorption process, 130
vibrational wave function, 278,28&290,308
two-photon emission process, 130
vibrational-electronic interaction, 292
two-photon process, 130, 132
vibrational-rotational spectrum, 288,293
two-vector correlation, 191
vibration-rotation Harniltonian, 293
vibration-rotation interaction, 253, 265, 277,
283,291
u, 69 vibration-rotation spectrum, 290
uncertainty, 12,33 vibration-rotation term, 70
uncertainty principle, 13, 18,33 vibronic band strength, 289
uncoupled representation, 44, 4 6 4 7 , 67, 71, vibronically induced transition, 293
100, 115, 144, 186 Viergruppe, 268-27 1,273-274,288
uncoupled state, 5 1-53 virtual state, 40
geometric interpretation, 52-53 volume of an irregular tetrahedron, 173
undirected line, 151
ungerade, 69
unit matrix, 73
unitary, 73 W coefficient, 145
unitary matrix, 8 1, 85 w3j, 325
unitary transformation, 4473-75,78,80-8 1, w6j, 325
108-109, 149,279,281 w9j, 326
unpaired electron, 70, 72 Wang transformation, 268,300
unpolarized light, 21 1,230-23 1,235, 3 15 wave
unstable rotation, 244-265.291
partial, 30
plane, 30
spherical, 30
V group, 268-27 1,273-274,288 wave number, 38
valence electron, 111-1 12,218-219,249,298 weak force, 185
van der Waals potential, 29 width, 40
Van Vleck transformation, 277, 279-283, Wigner coefficient, 44, 117, 151
292,299,3 10-3 11 Wigner rotation matrix, 132, 177
variance, 12-13 Wigner 3 - j symbol, 50
vector addition coefficient, 44 Wigner-Eckart theorem, 164,180-186, 192-
vector coupling coefficient, 44 193,204,213,215,231,234,237,
vector model, 12-18, 51-53, 55-56, 91-94, 240
99, 173,232,240 Wigner-Witmer Rules, 66-69
vector operator, 178, 187-189, 215
velocity, 27,33,84,255
velocity-selected beam, 34
vibrating oscillator, 278 Xe, 34
INDEX

Zeeman effect, 7 1, 184

Zeeman quantum beats, 133-135,202
Zeeman splitting, 202
zero line, 155, 161-163
zero-field level crossing, 203
zero-order Bessel function, 19
0 state, 305
0 state, 305
+

zone, 16,31-33
 Ground Level One Quantum

Large J Transition Moment

--

orbi id den Transition Excited State

With permission from Professor Henry A. Bent, Department of Chemistry, North Carolina
State University.
Chapter 7

ERRATA TO 2ND PRINTING OF

ANGULAR MOMENTUM
194 CHAPTER 7. ERRATA TO 2ND PRINTING OF ANGULAR MOMENTUM

The following errata have been collected from many sources to which we express our sincere gratitude. Special
thanks go to Melissa A. Hines, Alexei Buchachenko, Shan Tao Lai, Andrew J. Orr-Ewing, and Yo Fujimura who
communicated numerous corrections.

pg. 10 In Eq. (1.57) for the expression for Y2,±1 (θ, φ) change 16π to 8π in the square root sign.

pg. 36 In Eq. (45) change − sign to + sign in front of 1b ( π2 ).

pg. 36 In Eq. (46) replace f [b(x), y] by f [x, b(x)] and replace f [a(x), y] by f [x, a(x)].

pg. 36 In the two lines below Eq. (46) replace b/k by 1/k.

pg. 38 Replace twice k by κ in the expression in the line below Eq. (58).

pg. 39 In the ordinate of Fig. 5 replace E0 − E by (E0 − E).

pg. 43 In the third line below Eq. (2.1) replace Eq. (1.16) by Eq. (1.8).

pg. 66 In line 9 replace “total angular momentum” by “total orbital angular momentum.”

pg. 90 In both Eq. (3.77) and Eq. (3.78) replace superscript † by superscript ∗, and replace d JM M 0 (θ) by dJM 0 M (θ).

pg. 91 In line 1 replace † by * and omit “transpose.”

pg. 99 Fig. 3.6 is misdrawn; J2 should connect to J1 and J3 .

pg. 99 In the second to last line on the page change “Figures 1.1, 1.2, 1.3, and 2.2” to read “Figures 1.1, 2.1, 2.2, and
2.3.”

pg. 101 In the exponent of (−1) in Eq. (3.111) change J1 − M1 to J1 + M1 .

pg. 101 In the exponent of (−1) in Eq. (3.112) change J1 − M10 to J1 + M10 .

pg. 117 In the left side of Eq. (2) replace PJM (θ) by PJM (θ)dΩ.

pg. 144 In Eq. (4.3) replace hj12 j3 j | j1 j23 j 0 i by hj12 j3 j 0 | j1 j23 ji.

pg. 149 In the second row of the 9-j symbol appearing once in Eq. (4.21) and twice in Eq. (4.22) interchange j 3 and
j4 .

pg. 152 Rewrite rules 1 and 2 to read:

1. To add an arrow pointing toward or to drop an arrow pointing away from a particular jm pair, multiply the
diagram by (−1)j+m and change the sign of m in the diagram.
2. To add an arrow pointing away or drop an arrow pointing toward a particular jm pair, multiply the diagram
by (−1)j−m and change the sign of m in the diagram.

pg. 164 In the graphical diagram at the bottom of the page omit the central line j 1234 = 0 and its label and add two
lines connecting the + and − nodes; label these lines by j12 and j34 .

pg. 167 In Eq. (4.58) in the graph in the first line at the top of the page change from + to − the node on the right and
the node on the bottom. Then in the graph in the second line of this equation, change from + to − the node on
the top and the central node. Finally, remove the phrase starting with “where . . . ”

pg. 174 In Eq. (4.65) change j2 to j.

CHAPTER 7. ERRATA TO 2ND PRINTING OF ANGULAR MOMENTUM 195

pg. 174 In the first line below Eq. (4.67) change j23 to |j23 |.

pg. 174 In Eq. (4.67) add a minus sign before

d
(j23 · j).
dt

pg. 174 In Eq. (4.69) change (j1 × j) to (j12 × j).

pg. 176 In Eq. (4.71) change j9 to j7 in the last 6-j symbol on the right side of this equation.

pg. 178 In the third line from the bottom, change (see Application 4) to (see Eq. (14) of Application 4).

pg. 179 In three lines below Eq. (5.8) change Eq. (5.9) to Eq. (5.8).
0
pg. 181 In the line below Eq. (5.14) insert the factor (−1)k+j−j in front of (2j + 1)1/2 .

pg. 183 In the second line of Eq. (5.25) replace h3 || L || 3i by h3 || L (1) || 3i.
P P
pg. 187 In Eq. (5.38) change q0 to q0 ,q0 .
1

P P
pg. 188 In Eq. (5.46) in the first and second lines replace q by q,q 0 .

pg. 189 In the line above Eq. (5.50) replace Eq. (4.21) by Eq. (4.20).

pg. 190 In the last line of the first paragraph replace “the gradient of the electric field” by “the gradient of the gradient
of the electric field.”

pg. 194 In Eq. (5.67) add +2j1 to the exponent of (−1).

pg. 194 In Eq. (5.69) add δjj 0 to the last line on the right side.

pg. 195 In Eq. (5.70) add on the right side δjj 0 .

pg. 201 In Eq. (5.85) replace ψ(αe Je Me ; t = 0 by ψ(αe Je ; t = 0).

pg. 201 In Eqs. (5.86), (5.87), and (5.88) replace ψ(α e Je Me ; t) by ψ(αe Je ; t).

pg. 204 In the last line of this page remove the parentheses about êa and ê∗a .

pg. 209 In Eq. (5.118) insert a minus sign before the right side.

pg. 209 In two lines below Eq. (5.118) insert a minus sign before e(1, −1).

pg. 216 In Eq. (5.133) replace † by ∗.

pg. 217 In Eq. (5.134) and in Eq. (5.136) change † to ∗ twice in each equation.

pg. 222 In Eq. (13) second line from bottom of page change j 0 to j.

pg. 224 In right side of Eq. (17) change −1 to 1.

196 CHAPTER 7. ERRATA TO 2ND PRINTING OF ANGULAR MOMENTUM

pg. 228 Below Eq. (23) it should read: In the general case our system is in a mixed state which is represented by a
density operator ρ that is an incoherent superposition of a number of pure states |ψ (i) i with statistical weights
W (i) .
X  ED 
 
ρ= W (i) ψ (i) ψ (i)  (24)
i

where
X
W (i) = 1. (25)
i

pg. 228 In the line below Eq. (25) change “in” to “In.”

pg. 229 In Eq. (27) insert “hJM | ρ |JM 0 i =” before the expression on the right side of this equation.
(i)
pg. 229 In the sixth line from the bottom of the page, change a M to aM .

pg. 230 Change the sentence in line 13 to read: A pure state represents a completely ordered ensemble, whereas the
mixed state that is uniform is in a state of maximum disorder.

pg. 230 Change the sentence in line 17 to read: Then for a pure state S = 0, whereas for a state of maximum disorder
S = ln(2J + 1).

pg. 232 Add “cosine of the” before “angle” in part J.

pg. 237 In the exponent of (−1) in Eqs. (62) and (63) change M 0 to M .
(1) (2)
pg. 238 In Eqs. (67) and (69) insert 3 in front of O0 (Ji ) and after A0 insert the factor
 
5Ji (Ji + 1)
.
(2Ji + 3)(2Ji − 1)

pg. 238 In Eq. (68) change = to ∝.

(2) (2)
pg. 239 In the right side of Eq. (71) change A0 (Jf ) to A0 (Ji ).

pg. 239 In the right side of Eq. (72) it should read:

 1/2
1
1− .
(Jf + 1)2

pg. 239 In the right side of Eq. (73) change (2Jf + 3) to (2Jf + 1).
h i1/2
pg. 239 In the right side of Eq. (76) make it read: 1 − J12 .
f

pg. 239 In the right side of Eq. (77) change (2Jf − 1) to (2Jf + 1).

pg. 240 In the Eqs. (79) and (80) drop the subscript q on J and the subscript 0 on I.

pg. 240 In the right side of Eq. (81) insert the factor [(2F 0 + 1)(2F + 1)]1/2 inside the summation sign.
CHAPTER 7. ERRATA TO 2ND PRINTING OF ANGULAR MOMENTUM 197

pg. 241 In Eq. (84) add a minus sign in front of the factor i(EF 0 − EF )t/h̄ that appears in the first and third lines of
this equation.

pg. 250 In the fourth line of Eq. (22) add a minus sign in front of (5) 1/2 .

pg. 271 Replace the paragraph starting with “Eq. (6.63) is . . . ” by “Note that Eq. (6.58), (6.59), and (6.60) are valid
for integer and half-integer J and K, whereas Eq. (6.63) holds true only for integral J and K.”

pg. 280 In Eq. (6.99) insert i in front of (Ej − Ek ).

pg. 281 In line 2 change Eq. (6.97) to Eq. (6.98).

pg. 281 In Eq. (6.103) in the first line for hj| G3 |ki change (Ek − Eα ) to (Ek − Eβ ).

pg. 282 In Eq. (6.105) in its first line, replace twice HR by H1 .

pg. 284 In Eq. (6.112) replace π by π 4 .

pg. 285 Eq. (6.117) should read:

S(J 0 K 0 ; J 00 K 00 )
 2
X  2J 0 + 1   2J 00 + 1 1/2 Z 

J0 1∗ J 00 ∗
= 3  2 2
D M K0 0 D 0
0K −K 00 D M K00 00 dΩ 
 8π 8π 
M 0 ,M 00
 
X  [(2J 0 + 1)(2J 00 + 1)]1/2 Z  ∗ 2

J0∗ 1 J 00
= 3  2
DM 0 K 0 D0K 0 −K 00 DM 00 K 00 dΩ 
 8π 
M 0 ,M 00
 2
X  2J 00 + 1  
= 3  hJ 00
M 00
, 10|J 0
M 0
i hJ 00 00
K , 1K 0
− K 00 0
|J M 0 
i
 2J 0 + 1 
0M ,M00

00
2J + 1 00 00 X
0 00 0 0 2 2
= 3 hJ K , 1K − K |J K i hJ 00 M 00 , 10|J 0 M 0 i
2J 0 + 1
M 0 ,M 00
00
2J + 1 00 00 X  2J 0 + 1 
0 00 0 0 2 2
= 3 0 hJ K , 1K − K |J K i hJ 00 M 00 , J 0 − M |10i
2J + 1 3
M
2
= (2J 00 + 1) hJ 00 K 00 , 1K 0 − K 0 |J 0 K 0 i
 00 2
J 1 J0
= (2J 0 + 1)(2J 00 + 1)
K 00 K 0 − K 00 −K 0

0
−1+K 00
pg. 287 In Eq. (6.123) insert the phase factor (−1)J in front of the 3-j symbol inside the double summation.

pg. 294 In Eq. (6.142) change ω to ω in the middle term.

pg. 297 In last line insert after rotation: for M = J.

pg. 298 In line 2 insert after rotation: for M = J.

198 CHAPTER 7. ERRATA TO 2ND PRINTING OF ANGULAR MOMENTUM

pg. 303 Change Eq. (34) to read:

 E  E
 
|ψ(F2 )i = aJ 2 Π 32 vJ − bJ 2 Π 12 vJ .

pg. 303 Change Eq. (35) to read:

 E  E
 
|ψ(F1 )i = bJ 2 Π 23 vJ + aJ 2 Π 21 vJ .

pg. 304 In the second sentence change F1 and F2 to F2 and F1 .

pg. 306 In line 13 change former to latter.

pg. 306 Change Eq. (38) to read:

σv (yz) |n Λi = (−1)s |n −Λi .

pg. 306 Change Eq. (39) to read:

σv (yz) |S Σi = (−1)s |S −Σi .

pg. 306 In the first line above Eq. (40) change xz to yz and change χ → −χ to χ → π − χ.

pg. 306 Change Eq. (40) to read:

∗
YLΛ (θ, χ) → YLΛ (θ, π − χ) = (−1)Λ YLΛ (θ, χ) = YL−Λ (θ, χ).

pg. 307 In line 3 change σv (xz), y → −y to σv (yz), x → −x.

pg. 307 In line 5 change y to x and change C2 (y) to C2 (x).

pg. 307 In line 6 change C2 (y) to C2 (x).

 
pg. 307 Change Eqs. (41) and (42) by replacing
 (0, π, 0) by (π, π, 0); the right side of Eq. (41) is e iπ/2  21 , − 21 ; and

the right side of Eq. (42) is eiπ/2  12 , 12 .

pg. 307 Change Eq. (43) to read:

   
1 
1/2  1

σv (yz)  , σ = (−1)  , −σ .
2 2

pg. 307 In line 9 of the second paragraph the sentence should read: On reflection we interchange α and β in the
uncoupled state so that Σ → −Σ, and we pick up a phase factor (−1) x , where x equals the number of electrons
divided by two, i.e., x = n/2. Then omit the sentence beginning with “Since. . . ”

pg. 307 In Eq. (45) change σv (xz) to σv (yz), and change (−1)S−Σ to (−1)S .
pg. 308 In the third line below Eq. (46) change x → x, y → −y, and z → z to read: x → −x, y → y, and z → z.

pg. 308 In the fourth line below Eq. (46) change π − χ to −χ and change C 2 (y) to C2 (x).

pg. 308 In the fifth line below Eq. (46) change Eq. (6.59) to Eq. (6.60).
CHAPTER 7. ERRATA TO 2ND PRINTING OF ANGULAR MOMENTUM 199

pg. 308 In Eq. (47) change J − Ω to −J.

pg. 308 In Eq. (48) in the second line change Λ + s to s, S − Σ to S, and J − Ω to −J.
pg. 312 Replace anΩ by anΣΩ in Eq. (65) and twice in the second line from the bottom.
pg. 312 In the third line from the bottom of the page, change δ 0Λ to δ0,Λ .
pg. 313 Change Eq. (66) to read:

S(J 0 ; J) = (2J 0 + 1)(2J + 1)

  2
X X
 0± ± J 0 −1+Ω J 1 J0 
× an0 Σ0 Ω0 (p )anΣΩ (p )δΣ,Σ0 (−1)  .
 0 Ω Ω0 − Ω −Ω0 
Ω Ω

pg. 316 In reference 27, change C2 (x)σv (yz) to C2 (y)σv (xz).

pg. 316 Update the last line of reference 29 to read: M. H. Alexander et al. J. Chem. Phys. 89, 1749 (1988).
pg. 317 Reference 31 should be rewritten to read:
As Alexander and Dagdigian [29] show, the electron distribution in the F 1 e Λ-doublet level is oriented preferen-
tially in the plane of rotation and in the F1 f Λ-doublet level oriented preferentially perpendicular to the plane of
rotation along J for a single filled π orbital. The opposite applies to F2 Λ-doublet levels. For a π 3 configuration,
the preferences are reversed. The treatment outlined in F is an approximation in which the integration over only
χ is considered. A more complete treatment in which the integration over θ and φ are included modifies Eq.
(50).
pg. 320 In Eq. 16 replace ψLKM (Ω0 ) by ψLKM (Ω) and replace ψLKM (Ω) by ψLKM (Ω0 ).
pg. 325 Eq. (A-3) should have the signs in front of ν in the parentheses in the denominator −, −, −, +, + so that it
agrees with Eq. (2.25).