I[

If

0;

f~

GROUP THEORY
THB APPLICATION TO QUANTUM MECHANICS

PAUL H. E. MEIJER
17Ie Clltllollc UniNriity 0/ AIIterlt:G, Wash'IIgton D.C.

EDMOND BAUBR
Laboratoire de Chi",. Physlqw, Pari3

1962

NOR.TH-HOLLAND PUBLISHING COMPANY - AMSTERDAM

No part 0/ this book 1IItly be reproduced In any form by print. plwioprint,

microfilm or any othe, means wlthollt written permission from the publisher

PUBLISHERS :

NORTH-HOLLAND PUBLISHING CO. - AMSTERDAM

SOLE DISTlUBUTORS FOR U.S.A.:

a division of
INC. - NEW YORK

1NIEUCmNCE PUB USHERS,

JOHN WILEY &

SONS~

PRINT.ED IN THE NE.THERLANDS

BY N.V. DIJKSTRA'S DRUKKERU

v/H

BOEKDRUKKERU GEBROEDERS HOITSHMA

GJtQN1N9EN

PREFACE
Seldom has an application of so-called pure mathematics to mathematical
physics had more appeal than the use of group theory to quantum mechanics. Almost every student in this subject, after going through the necessary
theorems, felt the satisfaction of overlooking a broad field,) Inastering it in
its complete generality, as an award to his efforts.
In recent years the availability of tables of coefficients has increased the
applicability of many ideas introduced one or two decades ago and the number of papers applying the results of representation theory has been steadily
increasing.
The application of group theory to problems in Physics can be classified
in two types. As an example of the first type we mention the considerati<;lns
based on the symmetry of a crystal used to reduce the 6 by 6 stressstrain.:matrix (the generalization of Hooke's law)~ An example of the same
type is the heat conduction tensor in a crystal. Instead of nine components the
number of different elements is reduced by the symmetry of the crystal and
still further reduced by the Onsager relations (which are based on microscopiq time reversal). Still another example of the first type is the set of piezoelectric constants .
T~e second type of problems are those eigenvalue problems where the
differential equation or the boundary are of such a geometric nature that
certain rotations or translations leave the. problem unchanged. In this case it
may happen that the eigenvalue connected with the solution of the problem
is degenerated; that is, more than one eigenfunction belongs to the same
eigenvalue.
The central problem of the book is the study of this second type through the
transformation properties of these eigenfunctions. In the first cases the application of "group theory" is hardly more than the application of symmetry
considerations. In the second case the application of group theory, or actually the application of representation theory, is a much tn<?re essential
matter. The general idea of the transformation induced in function space
by a rotat.ion (or translation) in configuration space as explained in Chapter 4 is not only useful in Qlla.nt.wn. mechanics but in any other eigenvalue
\

PREFACE

vi

problem s.uch as vibrating systems (molecules or lattices) or waveguides as wen.

The crucial point in the developing of representation theory is Schur's
Lemma. The proof has been illustrated with a symbolic diagram and in
subsequent sections the theory is developed on the basis of this lemma.
Great value is attached to represent the ideas in a geometrical fashion:
For instance the similarity transformation is ~escribed as a rotation in multidimensional function space and the reduction is described in terms of
mutual orth0ional spaces..
Although there are many books written on group theory as w~n as on theconnection between physics and group theory, the number of books of in.
troductory nature are only very few. The general references are chosen with
'emphasis on clarity and readability and are mainly mentioned for further
study in this field.
The book is based on a French monograph entitled "Introduction it. la
theorie des groupes et ses applications en physique quantiqueU which appeared in 1933 in the Annales de I'Institut Henri Poincare. Chapters 1 through
S are a translation with addition of the monograph. The subjects treated in
the additional chapters deal mainly with developments since then. The bas.i~
ideas. of the application of group theory to quantum mechanics have not
~onsiderably changed and hence a fairly literal translation of the material
of the first edition is still valuable today.

the

I would like to quote from the introduction to tl;le first edition the following s"4&tement: 'The pages that follow are simple introduction. having aa
main~8oal to familiarize the reader with- some new concepts and to permit
him to read the oriainal papers orWell and Wigner with less difficulty."
I ,_ouId like to thank Mrs. Peretti, Drs. Morrison and Barry and Mrs.
Mielczarek for their cooperation.
J

CONTENTS

CoNTBNTS ....

CHAPTER. 1.
1.

vii

VECTOR SPACES - UNITARY GEOliE"J.1lY

n-dimensional vector or aff1De spaces

1.1. n-dimensional vector spaces ...

1.2.
1.3.
1.4.
1.S.

2.

. .. .

Tranaformation of coordinates..
.. ~ .
Linear mapping of the space at. on itself (linear transformation)
Composition of mappinp. Multiplication of matrices ....
fA

.......

Inverse matrix . .. . . . . . , .
~
1.6. Transformation or a mapping or projection-matrix by a chanp
dinates (similarity transfonnation)..
. . . . ~ " ~ .
Euclidean and unitary spaces. .
. . ,. . .
2.1. UQitary metric . . ~ . . .
~ ~
2.2. Unitary transformation . . \'; *
2.3. "Bilinear forma and Hermitian matrices
. . .
2.4. Goneralization of unitary spac-e into bra... and ket spaces ~
R~uction to main axes . . . . " ~ , ~
3.1. Dialonalizability. .
c

'I

..

..

..

..

..

01 coor4
6
6
8
8

...

..

4t

3.

..

..

..

9
11
11

3. 2. Determination of diqonal elements.. . . .

3.3. loint dialOnalizatioD of a set of matriCX\S

.0

3.4. Invarianco of a secular equati9n . . . .

.
3.S. Trace or spur of a matrix " . . . . .
Function spa.co. Complete lOti of ortbolonal (unctions . . . . .. . .
4.1. Function space . . . . ~ . . " . . . . . . " .
4.2. Scalar produc:t; norm . . .
4.3. Fourier series . Complete sets of ortholOnal futlCtions ... ""
Operators "...
S.I. Traalformation of tile function space into itself by a linear operator .
S,,2. BDiaear forms. Hermitian operators .
~
~
~
5.3. Reduction of an Hermitian operator to iu main Ue5 e ~ "

4"

I.

..

..

..

fl

'"

..

..

..

..

"

...

<!

..

:ot

:I;

,.

"

..

CHAPTBlt 2..
1.

..

,.

1
1
1
2
3
4

12
14
14
IS

l'lJ
16
17
19

19

21
21

THE PRINCIPLES OF QUANTtlM MECHANICS

Wa"v'es ~ . .. = ~
1.1. Classical waves . . . - . . . .
1.2. Quantum mechanical waves
.....
1.3. The free particle . . , .
<

vii

29
29
30

31

viii

2.
3.

4.

s.

6.
7.

CONTENTS
The Schrodinger equation .
2.1. 8.chrodinger equation .
2.2. The n-particle problem
Angular momentum . . .
3.1. Operator . ~ . . . . .
3.2. Operators and groups .
3.3. Commutaiion relations
The postulates of quantum mechanics.
..
Time dependence of a state and of a physical
5.1. General theory
I. ~ . . . . .
5.2. Heisenberg representation . . . . . . .
Transition probabilities and radiation theory.
Perturbation theory . . . . . .
. .
7.1. Formulation of the problem . . . . . .
.
7.2. Non-degenerate problems
7 . 3. Degeneracy . . . . . . . .
7.4. Quasi degeneracy . . .
7.5. Application: diatomic molecule . . . . .
b

CHAPTER 3.
1.

2.

3.

s.

6.

32
33
33
34
34

35
36
37

. ., .
observable.. . .

40

40
42

.
., .
. . . . . . . . .
. . . .

45
47

47
48
49
52
56

GROUP THEORY

The role of group theory in quantum mechanics. .

Examples. . . . . . . . .
2.1. General considerations.. . . .
2.2. Group postulates
..... .
2.3. Further examples of groups . . . .
2.4. Group table
Subgroups . . . . . . . . . . . . . . . . . . .
3.1. Definition . . . . . . . .
3.2. Cosets or complexes associated with a subgroup. . .
Conjugated elements. Classes
.....
. .
4.1. The case of linear substitutions
. . . . .
4 . 2. Generalizations. Invariant subgroups
4.3. Factor group . . . . . . . . . . . . . . . ,
4.4~ Abelian groups . . . . . . . . . . . . . . .
Some properties of the group of permutations of n objects
group) . . . . . . . . . .
5.1. Notation Vy'ith cycles , . . . . . .
5.2. Conjugated permutations . . . . . . .
5.3 .. Alternating group d. of n variables
Isomorphism and homomorphism .
6.1. Definition . . . . . . . . . . . . . . . .
6..2. General theorems . . . . . .
6.3. Representations of a group
6.4. Equivalent representations .

58/

. . .
. . . . .
. .

59
59
61
62
63
64,
64
65

66
66
67

68
69
9'. (symmetric
70
70
71
72

72
72
72
74
74

CONTENTS

ix

7.

Reducibility of representations . . . . . .
7.1. Invariant subspace . . . . . . . . . . .
7.2. Complete reduction or decomposition
... .
1.3. Reduction of the unitary matrices of a group into their irreducible parts
7.4. Ex.a.mple . . . . . . . . . ~ . . . . . . . . .
7.5,. Finite gI"OUpS
............................ ~ .
8. Uniqueness theorem. The decomposition of a given representation (II from
a group ~ into irreducible constituents is only possible in one way.
.....
9. Schur's lemma and related theorems
10. Characters of a representation . . . . . . . . .
10.1. Definition . . . . . . . . . . . . . . . . . . . . . . . . .
10.2. The number of irreducible representations of a finite group. . .
10.3. Regular representation of a group
. . . . .
11. Orthogonality relations (finite groups). . . . . . . . . . . . . .
11.1. General formulas . . . . . . . . . . . . . . . . . . . .
11.2. Application to the characters of irreducible representations
11.3. Class..Space . . . . . . . . . . . . . .
12. Sum of a class; projection operators . . . . . . . . . . . .
12.1. Definition of the sum of a class; structure coefficients.
12.2. Character tables . . . . . . . " . . . . .
12.3. Projection operators (idempotent elements) .
13. Repre~enta~ons of the permutation group.
13.1. Young-tableaux. . . . . . . . .
13 . 2. The P" Q-operations; idempotents . .
13.3. Irreducible representations . . . . .
Appendix 3.1. . . . . . . . . . . . . . . .

CHAPTER 4.

..

..

75
75

76
77
77

80
81
82
85
85

86
87
89
89
92
93

94

94
95
97
100
100
102
104
106

GENERAL APPLICATIONS TO QUANTUM MECHANICS; WIGNER'S

THEOREM

Invariant properties of the Schr5dinger equation . . . . . . . . . . . .

1.1. The two groups of the SchrOdinger equation . . . . . . . . . . . .
1.2. Transformations induced in function space by the transformations in configuration space . . . . . . . .
1.3. Expression of the invariance of H . . . . . . . . .
1.4. Constants of the motion .
2. Wigner's theorem . . . " . . . . . . .
2.1. Theorem . . . . . . . . . . . . . . . .
2.2. General solution by successive reductions
2.3. Equivalent description . ~ .
3.
Abelian groups
~.1. Permutations of two objects
. . . . . . . .
3.2. Plane rotations (around a fixed axis) (group 9,.). . .
4.
Non-abelian groups. Rotations and retlections in a plane
Appendix 4.1. " . . . . . . . . . . . . . . . . . . . . . .

110
110
110

111
112
113
113
114
120

121
122
124
126

CONTENTS

CHAPTml 5.

!'I,

ROTATIONS IN 3-DIMBNSIONAL SPACE: GROUP

Spherical harmonics and representation of the rotation aroup ... .

128
2.
R.otation group and two-dimenaioDal unitary poup . . . . . . .
131
2.1. Relation betwoon the rotation lfOuP and the unitary aroup . . . . . 131
2.2. Tho representations or the II'OUP .,. u reproae;ntations or tho aroqp It. 135
3. Infinitesimal traDSformations and aqul. 1IlOmontum. .. .. .. 137
3. t. Infinitesimal traBsformations of a continuous &rOUp .. 137
3.2. Linear substitutions . . . .. . _ . . .. . _ 140
3.3. Ilepresentations of the rota~on group_ Matrices for the aaau1ar momeata. 142
3.4. Pauli matrices . . .
..
1.3
3.S. An,war momentum in the state J. ...... _ ...
146
4. Transition from tho aroup 94, to tho aubpoup fl .. .. ..
147
4.1 .zeeman etrect ................... .
141
s. Product of two representations. Reduction formula. . .

149
5.1. Kinematic coupling of two systems with sphorical sJllbDOtrY.
14'
S. 2. Product of two representations . . . . . . . . . . . . . . . . . .
.. 150
5.3. Reduction or the diJect product of two ropresentations. Group _. I I
an example. Clebsch-Gordan formula . . . .
~
1'2
5.4. Total angular momentum . .
. .. . .
153
5.5.. Helium atom without spin
. " . .. . . . . . . .
1~
6. The electron spin .
156
6.1" UhJenbeck and Goudsmit hypothesis
" " .. .
J"
15.
6.2* Translation in quantum theory (pauli) " . . . . . .
6.3. Applications . . . .
. .. .
. .. . . . . . . . . .. . IS9
6.4. Complex atoms
..."........"....
. . . .. .. . . 160
7. Selection rules. . . . .. . . . . . . . . . . .
. .
162
8.
Parity or reftection character. Approximate selection rules ........." . 164
8.1. Parity; Tho rule of Laporte . . . . . . . .. . ....
. .. ..
164
8.2. Approximate selection tules . . . . . . . . . . . . . .
166
9,' Stark effect. Anomalous Zeeman effect. Line componetlta iDteDsity. ..LaacW
166
splittin, factor. Paschen-Back eaect ..
. .. . . . .
9.1. Goneral theory
. .
" .. . . .. 167
9.2. Intensity or components . . . . .
.. .. . . . .. . . 168
169
9 . 3. Lande-factor . .. .. . . . . . . . . .
. " .
Appendix I. The connection between a formal set of basis f\mctions aad the spherical
harm.onics . . . . . . . . . .
172Appendix II.. Construction of the irreducible representation of the 1l'01Ip _ . . 173
Appendix Ill. Proof of the formula (5.34a) . " " , . .. . . . " . .. . ... 171
1.

&

..

..

..

..

..

.......
...

fI

..

..

&

..

..

.....

..

..

"

It

..

Ii

CHAPTER

1.
2.
3.
4.

6.

CONTINUATION OP THE nmoRY 01' THE It.OTAnoN GllOtJP

Irreducible tensor operators. .. . .. . . . .. . . . .

R.epresentation of tensor operators . . . . . . . .
Wisner-Eckhardt theorem, reduced matrix elements
Racah coefficients .
~
>

"

..

..

"

. ..
. . .. .. ..
..
.

184
187

189
193

CONTENTS

CBAPTml7 . SPACE GROUPS

1.

23.
4.
S..

6.

7.
8.
9.

2.

3.
4.

5.

It

199
202

. .

.
.

198

206
208

..

..

"

.. ..

."

210
213
217
220

220
223

FINITB GROUPS

Rotational crystal symmetry . . . . . . .. .

Crystal field theory . . . " . . . . . . " . . .
2.1. Angular wave functions UDder finite rotational
2.2. Explicit calculation or wave functions .
.
Double groups . . . " . . . . " . .
Operator Hamiltonians ........
. .
4.1. Van Vleck perturbation theory
. " . .
4.2. Tonaor operator ,~"..."...
KramOl'l' theorem and time reversal. . . .
S.1. Kramers' theorem.
.s .2. Time reversal ....
JahnTeller effect . . . . . .
6.1. Introduction. examples
6.. 2. Normal coordinates " . . . ,
6.3. General description .
4

6.

.. ..

CHAPTER 8.

I.

. ..
.
.

Outline . . . .. .. ..
..
.. .. .. . .. .. . .. .
Crystalloaraphic point groups versus pnal point groups . ..
, ..
..
Space Iloups .
.. .. .. . .. .,

Stntcturo ot tho space IfOOP

.. .
.. ..
The quantum m~anics of IOlid state.

..
.. ..
Pure traaslatiollS " ..
" ..
" "
.. .. ..
Bloch theorem .. .. .. . . .
. .. "
Reduced wave vectors "
"
" "
"
"
.
.
..
Little groups, W.B.S. method"
"
".. .. . .. .
9.1. Little a;roup theory .. . ..
.

.. ..

9.2 Applications or tho theory of the little group . "

hOBI . . . .

. " " . . . . . . "

228

. . . "

233
233

." .

symmetry
II

...

II

"

237

......

. . . .
. . .

. .
~

. . . .

. . . .
. . . . . . . .

244
245
249
~3

. .

2'3
256

" ..

2S9

241

259
261
264

. ..

261
273

217
SYlTDlATIC BmLIOOIlAPHY

279

CHAPTER 1

VECTOR SPACES -

UNITARY GEOMETRY

1. n-dimeDSionai Vector or Affine Spaces

1.1. n-DIMENSIONAL VECfOR SPACES

They are obtained by generalizing the properties of ordinary space which

is an abstraction made from the concept of measuring lengths. Let us remember that in affine geometry the vectors x, " .... are defined only with
regard to the operations of addition x+ y (commutative and associative),
subtraction x- y, and multiplication by a complex number (distributive and
associative). The n vectors, e l ' e2, ., en, are linearly independent if one
cannot satisfy the equation
Xlel+x2e2+ ... Xne,.

= 0,

unless all the numbers Xn are equal to zero.

An n-dimensional vector space is defined by n linearly independent basis
veFtors. It consists of the set of vectors, obtained by taking all possible
linear combinations from these n basis vectors

(1.1)
In ordinary space, n = 3, the el are unit vectors laid out in three arbitrary
directions, the coordinate axes, and Xi' the components of the vector x, are
real. In quantum theory the quantities considered are in general complex;
thus one is led to take complex

x,.

1.2. TRANSFORMATION OF COORDINATES

If at,. is a vector space spanned by n basis vectors ef, let us select n other
independent vectors ei. From (1.1) the e~ will be linear combinations of the
ei (as all vectors of 8lra are). Thus we will have

(1.2)
where S

(Sik) is the transformation matrix With

1

11

rows and columns.

,lEcrOR SPACES -- UNITARY

A fixed vector x 'Will have the coordinates

X,

GEO~iETRY

and

[eh. I, 1

Xl in these two systems.

Equations (1.1) and (1.2) give:

%

"

=:; ' -

Xi~! ;: ~"
Xk e"

L
It

'
=v
L Xkeisik,
u:

from which
(1.3)

or
(1.3a)
Notice the difference between (1.3a), the transformation formula for the
components of the vector, and (1.2), the transformation of the unit vectors.
One traDsro~tion is the transpoaed inoerse of the other.
The notion of a vector can be genera1ized to tensors, quantities that ~ave

"t components instead of n components. k is the rank or order

of the

tensor. By definition these components transform like the products of vector

components. For instance a tensor of the second rank has n2 components

that have the transformation property

(1.3b)

The coefficients may have a symmetry relation Xu == XJI among them.

Suph a symmetric tensor of the second rank has only !n(n + 1) independent
coinponents. We shall not make use of this concept till Chapter 5, 2.2.
1.3. LINEAR MAPPING OF THE SPACE It.. ON ITSEIJF (LINEAR TRANSFOR, MATION)

To each vector x of It,. let there correspond a vector of the same space
., == Ax defined in the same coordinate system by its components;
Yi =

Lk

Qlkx/C;

Ax =

L sJafkxk'

(1.4)

i. It

A == (aile) is the mapping matrix or projection matrix.

Its order or degree is n (the number of rows or columns). There exists
a matrix of the order n that makes the vector x correspond with the
vector x itself that is it leaves the space 8l,. unchanged. According to
(1.4) all the elements are zero except those of the main diagonal and these
are equal to one. This is the unit nlatrix:

c, = 1.
h

tt... DIl\1ENSIO"NAL

Ch. I, Ij

VECTOR

SPACE~

One can map' It,. on another space &lift where m may be different from 11 and
in this case, the matri~ A is rectangular.
The mapping, A, of Ut n' on itself transforms a vector which is originally
identified with one of the basis vectors ek; into 'Ik == Aek and since ek has the
components D,. lone finds in accordance with (1" 1),.

(1.4a)
Note the identity in form of the equations (2.1) and (l.4a),' A change of
coordinates is a type of mapping limited to the basis vectors only.
The mappings are transformation operations on the n-dimensional space.
Their symbols A are operators acting on the vectors of this space in order to
transform. them into other vectors. These operators are analytically expressed
by matrices of the degree n. The mathematical problem presented by quantum mechanics is to generalize the properties of these operators to the case
where n is infinite.
1.4.

CO~fPOSITION

OF MAPPINGS.

~fULTIPLICATION

OP MATlUCES

Let us make two successive mappings of 8l on itself, first A then 8:

z = By = BAx ==-

ex .

The successive operations are carried out in the proper order by reading
from the right to the left in this equation. This rule will always be applied in
the: calculus of operatorso Equation (1.4) gives us
z,

=:

L b"YI = L b.,(L DIkXJ == L c",Xt ,

I

'"

that is

(1.S)
The matrix C is obtained by the usual multiplication rule of determinants.
The same result is also found when the mappings are carried out from one
space to another provided that the number of dimensions lend themselVes'
to it, as we shall see beiow.
It is convenient to represent a vector :t by a matrix X, where all the components are arranged hI a single column (the first one for example) and where
lOne has obviously

VECfOR SPACES -

UNITARY GEOMETRY

[eb. 1, 1

all other places are filled with zeros. Without difficulty one sees, from (1.4)
and (1.5) that the formula y = Ax becomes

Y = AX.

(1.4b)

A still simpler scheme is to consider vectors as a special case of rectangular

(i.e. non-square) matrices. We recall that in the case of a product of two
rectangular matrices A and B the width of A has to match the height of B
(see Fig. 1.1).
..
I
..
nt

...

{C]

t-

II

1
Pig. 1.1.

As in equation (1.4b) A stands in front of X. This implies that X has to be a

one-column matrix.
All our operations are thus reduced to the multiplication of matrices.. In
particular, 'the change of coordinates (1 . 3) becomes
X = SX'.

(1.3a)

1.S. QNERSB MATRIX

I

In ~order to insure that the mapping A will be reversible the determinant

IAI == taal has to be different from zero; the equations. (1.4) are then solvable
for ~ and one has

xl=La;1Yk or

X=A~lY=A-IAX.

So the inverse matrix A -1 of A is defined by

AA- 1

= A -lA =

I =

(t?ik),

(1.6)

that is
(1.6a)
1.6. TRANSFORMATION OF A MAPPING OR PROJECTION MATRIX BY A ,
CHANGE OF COORDINATES (SIMILARITY TRANSFORlvIATION)

Let e, be the initial system of basis vectors in which we will carry out the
transformation x -+ y = Ax. Now let us change the basis alld let e~ be the

Ch. 1, 1]

n-DIMENSIONAL VECTOR SPACES

y;

new basis vectors and xi, the new set of components of the vectors % and
y. The mapping or projection A, that is to say the correspondence between
the vectors % and y, is expressed in the second coordinate system by a
matrix A' such that
(1.4c)

" -= A'x'.

The transformation that converts A into A' is called a simllarit)' trans/or",.

lion or canonical transformation.
The form of A' can be calculated without trouble. Let us indicate the change
of coordinates e -+ " by S and the inverse e' -+ e by S-1, then we have
1
(compare 1.2) % == 5%' and x' == 5- 1 that is
The mapping
A' can be done in two di1ferent ways, either directly in the system .', or
indirectly by aoing from the axes e' to the axes It using S-1 and them carry.ina
out the mapping A in the system _ and finally going back to the system .'
by means of S. These two procedures are equivalent. The second is the
series of transformations

x,

l' =

x; = Las; x.

S-1"., == A%,

== S.x',

combJnin, these three pves

y'

== S-1 AS;x.' == A' x',

that is
(1.7)

Twb transformations A and A' are said to be equivalent if they are related
by an equation of the type (1.7), where S is au arbitrary transformation,
having an inverse. They go over into each other by a chanse of axes.
, As a simple example let us consider the operation in which every vector of
a th.reo-d:imeasional vector space tR3 is rotated throup an anaJe 8 .. 45
around a given axis II coinciding with e3- This operation can be expressed by
y == Ax where the matrix A is given by (compare Fig. 1.2&):

= (-: :

'~)

001

(8

= iJ2).

If we now perform a mapping of Ul3 on itself with the matrix:

S=

(~0 1~ -i),
0

VECTOR SPACES -

UNITARY GEOMETRY

[Cb. 1, 2

we obtain the new basis vectors el, e~ and e; indicated in Fig. 1.2b. The
4S-opera:tion described above is now given by y' == A' x' where
A' =

(~

o
1
o

-B)0 ,
B

and one easily verifies the relationship (1.7).

\ ...............

",

:~
.... -_I

\
\

.......,

\:

\ I

\~

.
;
.

,-------6 ,........
, ..........

.,/

45-

... ....

':/5-

\ J

\,

"

Fig. 1.2a
Fig. 1.2b
Fig_ 1.2. Tho . operation "rotate every vector of the 3-dimensional space . . through, an
&nile 8 1=: 45 0 around a given axis a" performed in two different systems of reference.

2. Eodidean and Unitary Spaces

2.1~

UNITARY METRIC

The preceding definitions are sufficient to determine the affine'properties

o~!a vector space. In order to. complete the analogy with ordinary usual
space, it is necessary to attribute additional properties to it (the so-called
properties) and for this it is sufficient to give an invariant and positive
definite form which is a function of the components of an arbitrary vector
and which defines the sqIIaI'e of the length or the norm of that vector,

_,ric

x2 =

i: 9a x ,x" 1.
ik

1 A metric in which. go Is non-diagonal is not very often used, at least in the c:ase of
CODStant coe.IIi.deDtB. A non-diagonal metric geometrically interpreted means that we have

an oblique coordinate system. In this case tho projections perpendicular and parallel to
the other coordinate axes are tile co- and the contravariant components. If tho vectors are
functiOns (compare 1.4). a DOD...iialonaI metric meaDS a set of non-orthogonal wave
functions, which are used in some exceptional cases, for instance the Orthogonali.zed
Plane Wave method in solid state physics.,
In general gQ may be functions o'fthe coordinatss Xl,~ Here alr~~dy in the diagonal case

Ch. 1, 2]

EtJCLIDEAN

<l~~ND

UNITARY SPACES

An affine space with a metric in a real domain is Euclidean if one can

reduce (g ik) to a unit matrix by a change of a,.'tes S:

In quantum mechanics one deals in general with complex quantities, but

the norm (or fundamental metric invariant) of an arbitrary vector has necessarily to be real in order to llave a meaning in physics. "fhe space is unitary'
because this invariant can be brought into the following form by a change
of axes s:
n

xl. ;: (x x)

== L, XiXi'

(1.8)

i= 1

where xi is the complex conjugate of Xi.

The coordinate system is thus unitary. In such a system one can even define
a scalar product of t\VO vectors y and x:
(x y)

= L, XiYi

= (y

x)*.

(1.8a)

These equations show that the axis of a unitary coordinate system are
orthogonal tmit vectors; their norm. is equal to one and the scalar products
of any pair are zero because the components ell and ek~ of' e,and e.. are 6" and
~'d respectively. 11lus we have:
(1.8b)

One verifies at the same time that the component Xi is equal to the scalar
product of the vector x and the unit vector e,:
(l.8e)
because the components of , are real.
In 1.4 we mentioned that vectors may be considered as special (rectangular) matrices. In order to obtain invariants, i. e. one-by-one dimenst9nal
matrices, like the scalar product (1.8) and (I.Sa) one matrix has to be of the
one-column type, the other matrix of the one-row type. If we write equation
one must distinguish between. co- and contravariant components (MOBIB aDd FIIHBACB
(1953) pp. 30 and 44) .. A simple example ora variable orthogonal metric is a...,Jar c0ordinate ~~. The most sencral fOmi, where 6,. are functions and aon-diaaonal, is.,aod in
the gcmera1 theory or relativity. (Compare "I. H. WBYL [1922) sections 2. 3. 4.and 11.)
y.7~ would like to remind thO reader that there are linear spaces used in mathematicS in
wpiqt 114.' metric is introduced, for instance,. the group-algebra mentioned in Chapter 3,
10,'.

VEcrOR SPACES -

UNITARY GEOMETRY

reh. 1, 2

(1.8) in matrix notation, the left-hand side of the product becomes the
transpose of the complex conju,ate of the right-hand side.
2.2. UNITARY TRANSFORMATION

A unitary transformation is a change in coordinate system, it is a transformation from one unitary system e, to another ~;. If x, and x; are the
components of an arbitrary vector x in the two systems, one must have

x x == Lxix,
, == L, x;*x; ,
hence this requires that the coefficients in (1.3) obey the following restriction,

Lxix,
=: L (SIkX~)StlX; == L x~X;(L S:Sn),
,
i. k. ,
k, I

that is to say
or using

(1~6a),
..
Slit

==

-1
S,d

The transpose of the complex conjugate of S, is called the adjoint matrix S,

that is
t :::2 Ski.
*
S,t
Using this definition the condition which unitary transformations have to

satisfy .becomes

5ts = I.
Sf ::::: S- 1 ,I.e.

(1 ,9)

The t1I;litary transformations, considered as projections of the space a"

'onto ~self, are those which leave the scalar products of vectors invariant
and yihich consequently conserve their orthogonal relations. This definition
leads again to (1.9).
2.3. BILINEAR FORMS AND HERMITIAN MATRICES

Let us consider the scalar product,

x Ay ==

L x:aIkY~.

(1.10)

i, k

This is a bilinear for", of the variables

X,

and Yt . 1

Prom its/Very definition this form is invariant, that is to say it does Dot cbanaeunder a
unlttll7coordiDate transformation. Hence if ODe considers theJt, as covariant components
aDd the x: u contravariant components, the matrix A is a mixed ten80r of the second order
in unitary space.
1

Ch. 1, 2]

EUCLIDEAN AND UNITARY SPACES

It is Hermitian if the projection A can operate just as well on the first as on

the second vector, that is if,

xAy::Ax"

(1.11)

thus

and finally

ait -- a*
lei'

A = At.

(l.lla)

One can verify without any trouble that the quadratic form Ax % is real if
and only if A is Hermitian. The expression x % is always real by virtue of its
definition (1.8) and obviously ;c. 1 is Hermitian.
2.44. GBNERALIZATION OF UNITARY SPACE INTO BRA AND KET SPACES

To establish a scalar product of quantities that have complex components

we could follow a different line of thought. Instead of using one space, the
unitary space, and the attribution of a certain metric to this space, we will
proceed as follows.
Take two independent spaces It and .. of the same dimensionality. To
every vector % in the first apace corresponds one and only one vector i in the
second space. This one-to..one correspondence should be chosen such that
to every transformation x -+ y == Ax, there corresponds a transformation
i -+ ., = Ai, in the ii-space, and also that there is a. one-to-one correspond. .
ence between the transformations A and A. Finally instead of defining a
metric we postulate that in order to obtain a number one baa to take the
components of % and multiply each one with the correspondina component
of ;. This is possible since i-th component Xi is the coefficient of the unit
vector _, and this unit vector is related to a definite unit vector ., in
As a result a component (which can be a complex number) of vector r
corresponds to a dofinite component of1 and one can perform the sum over
all these products
If we take this scalar-product type of sum of a
vector and its own adjoint in ii-space, we call it the uiength" of the vector.
Instead of indicating the unit vector of a certain baisis bye, or if we use,
following DDlAC [1958], only the inwces. A unit vector in space 8l is desisnated by Ii) a ket..unit vector, and in .. by (II a bra-unit vector. These strange
names stem from the fact that the length of a vector is always the product of
a bra and a ket and only these bra-kets are ordinary numbers (or in case of
application to quantum mechanics; only these are observable quantities).

V,;,.

10

,lEcr'OR SPACES -

UNITARY GEOMETRY

[Cli. 1, 2

Th.e notation of a length is as follows. A vector in It or ket space is

I
the adjoint vector in

it

>= LXIii);
t

or bra space is

<I=: LXi(il,
i

the result is the following bracket:

<I >= (Li (ilx,)(EJ xjli) = Li

XiXi

since the brackets (iii) are all equal to one as Ii) was a unit ket and (If was a
unit bra and we assume for simplicity that the scalar product of two different

bra's and ket's is zero

<iIi> = 0,

i:p j.

An operator that establishes a projection in ket space can be fully characterized by the mappings of all the unit vectors,
Aji) =

L aljli),
j

and if we are interested in the scalar product of an arbitrary bra <il and the
projected ket Ali) we obtain a set of numbers.

<iIAli)

aijvli) = au

which are representative for this particular operation A. Similarly if we study

the same operation in bra space the transform of (jJ will be

<iIA = aJi(il
(we write

A to the left

m, bra space for reasons of symmetry in the final

result, one has to remember, however, that the action of this operator is
from right to left). The scalar product with an arbitrary vector in ket space
gives again a set of numbers that characterize the operation completely.

(jfAli)

:=

ail-

We can calculate in a similar way as 1.6 the matrix elements of A with

respect to a new set of basis kets Iv) and bras (pl.

A'

== S-1AS .... (pIA'lv) = (Pli) (itAlj) vlv)-

The same index in the "opposite" positions means a sum over that particular

Ch. 1, 2, 3]

11

REDUcnON TO MAIN AXES

index and is called a contraction. The relation between the coefficients of

transformation in bra space and the coefficients of transformation in ket
space is
(Illi) = (i!f.l).

Finally we want tf) mention a quantity which consists of a ket and a bra:
li)</JI . This is not a number, but an operator. Its matrix elements between
the bra (il and the ket Iv) are

0Ii><JLI11) ==

~ij~/tV"

It is sometimes called a projection operator since this special operator picks

out a particular ket, i.e. tu), and projects it upon one and only one other ket:
Ii) .. This quantity is comparable to the direction cosines in an ordinary
change of axes. One could call it the dyadic product of a ket and a bra.
All this is so far only a different language and a different notation. Nothing
new is added that could not have been described by a unitary ~pace with a
certain metric, if we had considered the bra vectors as the complex conjugate
of the ket vectors. It is, however, possible to generalize the preceding picture
quite easily in cases where the bra vectors have a different relation to the
ket vectors, and this is used in relativistic quantum mechanics.
3. Reduction to Main Axes
We will admit, without specifying its proof, the following theorem which
is the generalization of theorems on the equations in S and the reduction of
quadratic forms' to a sum of squares.
3.1. DIAGONALIZABILITY

In unitary space every Hermitian form may be written in the form:

by a convenient choice of orthogonal coordinate axea; fI, being real nrIIIIIMIn.
Or, that is: Every Hermitian matrix A can be brought in diagonal/orm'by Il
unitary transformation:
(Xl 0 .... 0
o (11 0

. . . . . ,
o 0 ... IX,.

A similar statement holda

if the matrix is unitary.

sts

:8

I.

(1.13)

VECfOR SPACES -

12

UNITARY GEOMETRY

[Ch. 1, 3

We would like to end this section with a short note on the diagonalization
of matrices in general. The question "Which class of matrices can be fully
diagonalized7", is usually not treated in physics, hence the mathematical
techniques in physics sometimes create the misunderstanding that all matrices can be diagonalized.
, A normal matrix N is a matrix that commutes with its Hermitian conjugate:
NNt - Nt N = O.
(1.12)
An arbitrary matrix can always be brought into a form in which lutlf the
off-diagonal elements, Le. at one side of the diagonal only~ are zero 1 ~
If we have a normal matrix, the relation

(~1

;:2:: :)(~~f~ .. !~ :::) (:.~~ )~ :)(~ ~~~

=

..

::

..

: : :)

gives a set of relations:

..tiA,l +IN1212+IN1312+ ~. == A1A~

hence all NlJ(i #= j) are zero.
Conversely if a matrix is diagonal (1 .. 12) holds and since this equation is
invariant it holds if the matrix is not diagonal.
If we diagonalize the normal matrix its eigenvalues are complex. If they
are all real the matrix is Hermitian and vice versa. In the unitary case" the
eigenvalues have absolute value 1.
3.. 2i DETBRMINATION OF DIAGONAL ELBMENTS

The problem of determining the diagonal elements arises in geometry,

in mechanics, and in the physics of crystals. Consider for instance an anisotropic dielectric. The distribution of dielectric constants in different directions
in space is represented by a quadratic form: the ellipsoid of specific dieleetric
constants. To bring this ellipsoid on its axis means to find the "principal
directions" of the crystal, where the vector of the dielectric displacement D,
is paiallel to the "Vector of the electric field strength '8. One knows that
there are in" general three directions 2 and that they are orthogonal. Henee
1 It takes more effort to show that one c~ simplify an ~.rbittary matt-be to a form
consi.ting of a set of diaaonal elements and off-diagonal elements Oi+l ~ 1t the Jonian
t

'

oormal form .
. Except the case in which the ellipsoid or dielectric constanta is an ellipsoid of revc>lu..
tion~

Then the two main values of the dieloctric constant are equal (a degeneracy)" Thus
only one main direction is determiDed; it is normal to a plane in which the two othor
axes can bo chosen arbitrarily!

REDUcnON TO MAIN AXES

Ch. 1. 3J

13

there are three rectangular axes for which one has

Di = ai E "
(i == 1, 2, 3)
the Bl are the main dielectric constants of the medium. Simllarly in the general
case in which we are actually interested, we have to search for a unitary
system of axes ei which brings the matrix A' (the transformed of ,4..) into the
diagonal form (1.13). This means solving the following problem: find n
directions ei, e~, . . , e~ such that every vector x parallel to one of the
directions ,; will be transformed by A into a vector y 'which is parallel to %!

y == Ax = ax,

(1~14)

where (X is a constant.
Writing the "components" of this vector equationt we obtain n linear
homogeneous equations of the form

a'lxl + ... (an-)x,+ . ,. ,. a",x" := 0

(i == 1, 2, ... , n),
(1.14a)
which are only compatible with each other if the undetermined constant Of
is a root of the secular equation 1

(all-A)
Det IA-A.It

I
(a22 -A)
I = o.
(a.-A.) I
at2

.. ..

ala

42,.

==
..

..

II>

..

..

..

II>

..

..

(1.15)

..

This equation has in general n roots.l ::: (Xl II . . . . ,. (which may be distinct
or not) to which correspond the 11 directions of the axes given by' equation
(1.141.). We know only the directions of the vectors :I but ftot their'magnitude
becaUse (1.144) determines omy the ratios between the oomponents. The
roots' , of (I. IS) are the eigenvalues, propervalues, or characteristic con..
stallts of the matrix A. They are real in the' Hermitian case; in the unitary
case 'their absolute value is 1. The are the eiaenvector. or principal direetiollSa~ . When (1.1S) represents a multiple roOt of order P(<<l - 2 :::e ~

er

This name oriainates from a PrQblom in astroD9Jll)', where a siluilar "Iuation deter. .

mines the perturbation over long time interVals.

The p~ oftllele theorebllaabe made bJd100sinaa root CIt, 0(1.15), detem1iDiog
the correapondiDa eiaenvoctor..~ . _complotfq~tho,~>with (II-I) vecto.l'J whidt
have to be ortholonal to the first
In order to fonn'a unitary system or axes. Asa
result of the symmetry proportiea ot the HermitiaD':and:'1Ulitary lIIatrices the coefficients
all
at". list Q".l, are all zero and the matrix A. takes the form:

one

0
..
:
0)
o au .
() .. ..::: .

<<1 }~

t.-

~:

a", .

~~

VEcrOR SPACES -

14

UNITARY GEOMETRY

[eh. 1, 3

ap = tX), the vectors which have the property 1 == Ax ==

form a subspace
of p dimensions in which the direction of the axes is not determined. (See
(X%

the second note on page 12).

3.3. JOINT DIAOONALIZATION OF A SET OF MATRICES

In order that all the matrices of an Hermitian or unitary system can be

reduced at the same time 10 their principal axes, it is necessary and sufficient
that they all commute with each other.
First we "ill show that this condition is necessary: let A and 8 be two

matrices. By a transformation of axes S we reduce them simultaneously to

diagonal form A' and Sf. Then obviously they commute with each other:
A'B'

= B'A'.

Hence S-lASS-las = S-lABS == S-l BAS i.e. AB == BA.

Now we will show that the condition is sufficient: let us suppose that AS =
BA and let us make a transformation of coordinates such that 8 is diagonal:
We have

If
{J, ~

p",

QUe

== O.

So ~ fiIld that A is a Itep-'WiIe matrlxt of which all the terms are zero,
cxce~ those which are", situated on the main diagonal, or those inside certain
squar~ that share the diagonal; they correspond with the case in which

p, -"ll"e These squares are eaGh related to a subspace \ll in which the
~pal directions of the matrix B are undetermined (the matrix posseasoa a
cirtuiar, spherical, or hyperspherical symmetry)_ One may finaDy choose
these undetermined axes in such a way that A will be completely diagonal
3.4. INVAlUANCE OF A SBCULAll EQUATION

A traDsformatiOll. of coordiiudes does not chaDse She form of the secular

equation (leIS). Let US carryout the tra1l1formation of coordiaates 5:

A

-+

A' .. S-lAS;

The rule of multiplication of determiDants gives us

IS-1AS-lII =- IS- 1 1-IA-lIIISI

== IS- 11 lSI IA- AII == fA -;. II.

(1.16)

Ch. 1, I 3, 4]

15

FUNCI10N SPACE

3.S. TRACE OR. SPUR OF A MATRIX

Let A = (aft) be a matrix, the trace is then the sum of all the diagonal
elements

,.

Tr A ==

,=2: ail

(1.17)

To prove that the trace is invariant we shall write equation (I. t S) in the form:

(-l)-+ (_l)'l-l(al1 +a22+ .. a,.,.)+ ... = (-~r+ (_A)',-t Tr A + ....

Because this equation keeps its form under a transformation of coordinates
S" all the coefficients are invariant, in particular the second:
Tr A == invariant.

(1.17a)

4. Function Space. Complete Sets of Orthogonal FDDetlo..

Every function of continuous variables and particularly the wave function

in SchrOdinger mecbanics can be represented by vectors in a /UllCtion space
in which the "umber of dimensions is infinite. The operators acting on the
wave runctions ~l transform these functions into other functions producing
a transformation of this space irlto itself. On the other hand quantum mechanics can also be formulated with the help of certain relations among matrices. The matrices in this so-called matrix mechanics are matrices with an
infinite number of rows and columns. They can again be considered as
operators that transform a space with an infinite number of dimensions into
its~f. This analogy, brought forward by Hilbert, made it possible for
Schradinger and later for Dirac to show the equivalence between wave
mechanics and matrix mechanics.
There exists between the two spaces just mentioned a difference which
appears to be essential: the matrices operate on a space in which the number
of dimensions it denumerably infinite; on the contrary the Dumber of dimen..
sions in function space is of the order of a continuum. We will see, in a
moment that this difference is more apparent than real (see 4.3).
4.1. FUNCTION SPACE

"'(x) defines a function of one variable x. The simplest case is that of a

discontinuous function where the value is given only for a finite number of
~alues/or the variable x : Xl, %2, , x., i.c. the domaia of the variable x
consists of a set of discrete points. The " correspondina values of the
function "'1' tit2 t/J. ("', == ';<xc can be considered as components of a
vector in a sp~ of n dimensions. To each different function tp(x) defined
f

16

VECTOR SPACES -

reb.

UNITARY GBOMETRY

1, 4

for the same values of x there corresponds a distinct vector which is the
geometrical representation of that particular function.
Normally the functions in which we are interested depend on continuous
variables. So we have to go over to the limit in which the number n of dimensions of the function space becomes infinite. In such a case one may
consider the value ~ of the variable x as an index and. say that J/J(e) is the
component of a veetor '" along an axis characterized by the index ~4I For
functions of several variable., such as "'(x~ Yt z) the set x, y, z constitutes one
single index and the function space shows the same properties as those with
a single variable.
In the next section we will reason maiDly by analolY and be satisfied by
indicating now and then the mathematical difficulties. We will always assume
that conditions of convergence are realized.
4.2.. SCALAR. PR.ODUCT'; NORM

In IJ-dimensional space, the scalar product of two vectors

been defined by

X, = '-1l:x:Yill

and y has

fJ

(I.Sa)

By anal6gy in the case of two functions of a discontinuous variable, which

are defined for the values Xl' X2' II, XII of the variable, we will put:

(t/I tp)

=tL- l ",*{Xt)!p(x,,).

If the functions cp(x) and ';(x) are defined in a continuous domain

of the variable x the only possibly generalization is

(1/1 .) =

.,,*(x)cp(x}dx.

(1.18)

Thia iatearal is the _ar product of two /unctiDNI ';(X) ' and 'P{x). The
wave function in quantum mechanics depends on the coordinates of the
particles Xl , Yl, %1, X2' , z,., the domain D is the, finite or infinite configuration space in which we will indicate the element by dt == dx 1 ., dYl
dZ l dx2
II

dzp and we have

(t/I cp)

-=

Iv",.dt.

(L1Sa)

In particu1arthe norm of the function t/I, which corresponds to the square

of the length of the representing vector can be written

('" "') ... J.,","JjJ d'C

( 1.19)

17

FUNCflON SPACE

Ch. i, 4]

We will assume that all the functions which we will consider make the
integrals converge, whatever the domain D is, even if it is infinite. These
are called square integrable functions.
4.3. FOURIER SER.IES. COMPLBTE SETS OF ORTHOGONAL FUNCTIONS

Every function l/I(x) satisfyin, certain broad conditions about contiDuity

inside a domain contained between the values x == 1t and x == -1 can be
developed into a convergent trigonometric series, a series of fi,lntIamental
basis /unctioIU sin nx and cos nx (n = 0, 1 .... co). If :x is an angle, this domain is a circle with unit radius. This is the wen-known theorem by Fo~~r.
It will be convenient to use complex variables. Then the basis functions
the exponentials ei"~
exp (inx) where n takes on all inteFrs between
- 00 and + co. The possibility of these power. .series developments is related
to two essential properties of the fundamental functions:
1. III the domain - n, + 1t the basis ftw.ctions multiplied by a convenient
"normalization factor", which happens to be equal to (11 J2i). are orthogonal and normal or orthonormal which means that they satisfy the wellknown
and easy-to-check equations

are

;III:

(tp" P,.)
tp,.

.
= -1 f+- e-u'~ei~dx
== 6

ma ;

2" _.

= (21t)"~*ehU:;

((i",

= (2n)-iet m.x.

2.j These functions form in this domain a complete or closed system.

We will give this proposition the following meaning: in the domain considered
we Can develop every function that interests us into a convergent series of
basis functions t for instance continuous functions which have a sufficient
number of derivatives.
We know that the Fourier development can be extended witho~t trouble
i

Por reasons or rlsor, matbeDlatJciaD8 Jive in general a more precise de8nJtion or a

complote,system: a system of fucdamental fimctions ,,(x) is complete it ODt-caD ftnd for
each continuous fUnction .,(x) coefficients PI/. such that
'1

Jim

Jt-+ co

J
D

f-,(x)- EJl..(x>f adx .. 0,

Ie

(1 ..21&)
_

I it means the absolute square. So we deaJ with -an unlimited approximation In the mean
on the domain D or C(Jnv~'gence in tit, lMaR. It is Dot necessary that ~)Plltp" is converscnt.
The ronnulas (1.22) and (1.23)' can result just as
trorn- tbis detlnition as from our
restricted definition. AD the calculations, including those in perturbation theory in
Chapter 2. can be pttIonnod by makin, use of the ri,OtoUl deAnition(1.21a). We have
preferred to sacriftce pnerality ip. order to simplify the ar~ents.

wen

18

VECTOR SPACES -

UNITAR,\' GEOMETRY

(eh. 1, 4

to an arbitrary number of variables. We know also that the limits of the

domain of validity of the trigonometric series can be modified by a change of
variables. We will return to this point in a moment.
Spher:Dl fUDcUons, Hermite functions and Laguerre functions which one
encounters in quantum theory, possess the same properties. They form, each
in their domain (i.e. -1, + 1 for the polynomials of Legendre, the surface of
a unit sphere for the spherical harmonics, - co, + CX) for those of Hermite and
0, + <Xl for those of Laguerre) a compl.ete set of orthogo~al functions.
There are many other sets of functions which possess these properties.
Therefore we will consider in a general wayan infinite sequence of functions
1/11 "'2 of a certain number of variables and we will say that they constitute a complete system of orthogonal and normal functions inside a certain
domain D, if the following two conditions are fulfilled. First we demand that

("'i .;.) =:= ",r"'kd't =

~

~lA:'

(1.20)

the integrals being extended over the domain D. Second that all the continuous
functions which occur in the applications in physics can be expressed in this
domain by a series development of the form
(1.21)
k being an index which is allowed to talce all integer values between 0 and 00
or ~tween - 00 and + co (as in the case of a series of expon~ntials). If one
cotnpares (1.21) with (1.1) and (1.20) with (1.8b) one sees that these equatioris can be expressed in the following geometrical language:
~. A complete system of orthogonal functions establishes a unitary system of
coordinates which spans completely the function space. The p" or the
Fourier coefficients are the components of the vector t/I in this system. These
unitary axes establish, by their very definition, a denumerable set. Therefore,
they permit Feduelion of th~ properties oftbe functional space to those of a
space with an infinite denumerable number of dimensions.
If we form the scalar product of "', and", and take (1.20) into account,.
we obtain

(1.22).
a wen-known formula from Fourier which allows us to calculate the coeftl-.
dents

p,. It expresses that p, is the orthogonal projection" of l/I on the;

axis

19

OPEllATORS

Ch. 1, 4, 51

1/1,- We obtain in the same

(t/I .;) =

way the fundamental formula of Parseval

Lfpi"'!fJk,t/llr.d~ = L PiP,

(1.23)

It

If one can show that this relation is correct ror an arbitrary continuous
function, one is sure that the set of 1/1, is complete. The choice of orthogonal
functions that span the function space depends essentially on the problem
that is studied. There is an infinite number of possible choices, whether the
domain D is finite or infinite. The different complete sets can be derive4 from
each other by a change of coordinates or unitary transformation (compare
2.2).
5. Operators

'.1.. TRANSFORMATION

OF THE FUNCTION SPACE INTO ITSELF BY A

LINBAR OPERATOR

An operator is a symbol which establishes a correspondence between any

vector'" in function space and any other vector. For example, the operator"xL
acting on t/I(x) establishes a correspondence between ';(x) and.& function
tp(x) == xt/l(x); in the same way the operator d/dx acting on ';(x) gives us
""(x)

==

(d/dx)t/I(x).

The operators that interest us in quantum mechanics are linear operators.

Any Pnear operator A satisfies the following three conditions:
1. If ex is a number: A(aJ/!) == rJAt/I.
2. A(t/I+tp} == Ay,+Alp.
3. Both the functions f/I and At/! are normalizable .This implies that At/I
is defined in the same domain as '".
By definition, and by analogy with ordinary space, a mapping or tra,ns..
formatio-nof the function space on itself is a correspondence established
among its vectors by a linear operation.
Let us span the function space by a complete set
orthogonal axes
'" 1, "'2, .. Let t/I be one of its vectors with the components fJ l' /32, .,
then the development in a series of orthogonal basis functions can be
written as
(1.21)

of

The projection A makes this vector correspond to a new vector

20

VECTOR. SPACES -

from which the components

UNITARY GEOMETRY

[eb. 1, S

p;, in the same coordinate system can be ex-

pressed as a linear function (a series with constant coefficients) of the

coefficients /l" assuming of course that A was linear,

(1.24)
We write:
where At/I.J the function "" transformed by the operator A, can be developed
itself in a serieI of fundamental functions '" 1, '"2' with Fourier coefficients Ga:

(1.25)
From which we find that:
.,,' at

II, t (a", P.)'"I -= II P;';,

This implies (1.24).

The equations (1.25) are infinite in number when one deals with continuous
functions (k == 1,2 ). They are important because they define the matrix
(a'l) aa A which represents the linear operator A in the coordinate system
.p,. The matrices which we have introduced this way to associate with every
lin~ operator and which we will call representation matrices depend
essentially on the fundamental basis functions. Their components can be
obtalined, according to (1.25), by forming the scalar products

<l/I, Al/I,) -

fl/IiAl/Itd't t f
=

l/Itl/lJaJtd't

= all:;

(1.26)

(1.25) and (1.26) are constantly used in quantum theory.

The equation (1.25) may still be considered from another point of view:
as a 'CUansformatioD of axes" (compare (1.4a) and (1.2). Particularly if we
span the function space by another complete set of orthogonal functions
we will have by developing each epic in a series of functions t/I,:
p"

==

L, .;,u"'.

C{J,

(1.27)

Hence if we introduce a new set of orthogonal functions the linear operators

will undergo a similarity transformation (see (1.7)). Since the transformation
matrix U == ("It) conserves the orthogonality of the basis functions it is a
unitary matrix. One can verify that this satisfies (1.9).

21

OPERATORS

Ch. 1, 5]

The arguments of 1.6 are still valid under the condition that we give the
infinite matrices, with which we are dealing here, convenient convergence
properties. We will obtain for the expression of the transformation A in the
coordil1ate system lPk'
A' = U- 1 AU.
(1.27a)
5.2. BILINEAR FORMS. HERMITIAN OPERATORS

In n-dimensional space the scalar product (x - Ay) is a bilinear form of the

components of the two vectors x and y (compare (1.10).. We will establish
the same product (",. Alp) supposing that our functions are developed
according to (1.21):

Taking into account

(1~25)

we obtain

(1/1 A(p) = t/I*Aq>d-c =

L pil/ljl/!,alk y"d-c == L PiaUc"lh

i~ k f

..

(1.28)

i, k

which is an infinite bilinear form of Fourier coefficients of the components

of the two functions 1/1 and qJ in the basis system y, ,- Since one defines an
Hermitian operator in an n-dimensional space by the condition

(1/1 'A!f') = (Al/!' q =

L f l/!:atiP'iYIl/!,d-r: = LYka~M~

11:, i, i

(1.29)

I, t

from which, according to (1.28)

(1.29a)

thus, the corresponding matrix is Hermitian too.

5. 3. REDUCTION OF AN HERMITIAN OPERATOR TO ITS MAIN AXES

Let us reason again by analogy with the n-dimensional space: we want to

find a systenl of basis vectors '" 1 , "'2 ... to span the function space such that
eacl1 vector A ifr is "parallel" to t/I, or more precisely that the function Ay",
the transforrnation of 1/1, by the operator A, be equal to "', multiplied by a

constant:
(1.30)
The functions found in this way are the eigen/unctio1t8 or eigenvectors of
the operator .t\; the numerical values eli its eigenvalues.

22

VECTOR SPACES -

UNITARY GEOMETRY

{Ch~

1, S

The solution of this problem for a space with a finite number of dimensions
is relatively simplo ,at least in principle ( 3) and can be done by purely
algebraic methods. In function space, on the contrary, complications and
considerable difficulties appear and rigorous discussion of the equation
(1.30)1 has not been completed by the mathematicians.
The problem of finding (%, is algebraic, differential, or integral according to
whether the operator A is algebraic, differential, or integral. The Schr6dinger
equation (2.15) is a particular case: The Hamilton operator to which it refers
is a differential operator. Discussion of physical problems presented by
quantum. mechanics has lead to a better understanding of equation (1.30)
since the work of Hilbert .
Depending on the nature of the operator A and on the size of the domain
D, which restricts the variables on which the function l/I depends, we are
dealing with two different cases.
1. The easiest case arises when the square summable solutions of equation

(1.30) form a denumerable set, corresponding to a discontinuous and

denumerable set of eigenValues a;: ttl' el2, . ... There may correspond
several ei~nfunctions till to certain constants ex" but this multiplicity is
alwayS-supposed to be finite. This is expressed by saying that the spectrum
of eigenvalues of the eigenfunctions is discontinuous or discrete .
This case generalizes in a most direct way the results obtained in ndime~siona1 space. It occurs most often if the domain D is finite (as e.g. the
case ~f free particles confined to a box), but this condition appears to be
neitber necessary nor sufficient. It happens for instanc:e in the quantum
the91"Y of a harmonic oscillator that the levels form a discontinuous spectrum
although the size of the domain is infinite"
It can be shown in the case of discrete eigenvalues that the eigenfunctions
.; , of the operator A form a complete set of orthogonal functions which span
the function space completely2. In order to prove the orthogonality of
two eiaenfunctions t/I, and J/!" correspondina to two dUI'erent eigenvalues
of an Hermitian operator A, we have

Multiplying the first equation by

1

f/lt and the second by JjI f and 8ubstractiDg.

One ftDds a buic discussion in 1. v. NaUMANN

{193l]~

See, for example, ltlLBD.T and CoURANT (1930]. Chap.. Sand 6 or P. lvi. Mi.~ a.ad.
FuIlJlACH !19S3] p~ 727 and p~ ilSi>

H.

Ch. 1,

S1

OPERATORS

23

and using
we have
(ex, - rl,t)( '" i VI).)

== 0

j~e.

(t/I i l/IlIJ == 0,

"'3,

If several eigenfunctions, for example t/ll' 1/12'

belong to the same
eigenvalue eX we have degeneracy: every linear combination of these three
functions is again an eigenfunction beloIlging to the value et:, because the
three equations

after multiplication with three arbitrary constants Pi and summation win give
3

L {JiAVt,

t= 1

= A

L, PiifJ! == rJ. L, Pio/l-

Hence one can always choose three orthogonal linear combinations among
them.
In order to make use of the analogy that exists between the function space
and a n-dimensional space to the largest extent, we will consider an arbitrary
function", having components Pl, P2, .... in a coordinate system of orthogonal functions, as in equation (1.21). The scalar product (1/1 At/I) can be
a

written as a quadratic form in PI according to

(!/I ~ At/!) ==

(1.28)~

L p'taikPk ~

l,l

But if the basis functions are eigenfunctions of the operator A, (1.25) can
be replaced by (1.30) and we obtain

Thus the quadratic form (t/I At/!) is reduced to a "sum of squaresU "
2. It happens very often and particularly if the domain D is infinite t}1~t
in addition to the discontinuous spectrum we have a continuous spectrum
of eigenvalues, and it may even happen that the discontinuous spectrum is
not thereat all. The theory of atoms with a central force field furnishes an
example: the spectrum of energy levels is discontinuous up to a certain limit
(tlie ionization potential) and becomes continuous above this\> Since the
essential character of the orthogonal functions is that they fonD. a denumer-

24

VECTOR SPACES -

UNITARY GEOMETRY

[eh. 1, 5

able set, obviously, one does not have the right to apply \v:ithout modification
the language of linear vector spaces 1
The development in Fourier series (1.21) is then replaced at least partially
(compare (1 . 32) below) by integrals. These integrals can be considered as
the limits of series of orthogonal functions.. This permits us at least to a
certain extent to give a sense to the expression of orthogonal functions in the
continuous case. It will be sufficient to give two examples. Let us consider
first functions of a single variable x defined in an arbitrary finite or infinite
domain D and the operator "multiplication by x": (1.30) has now the form
xt/l(x) = exifJ(x),
an equation which has to hold everywhere in the domain D .. It is obviously
impossible to construct an analytical function which satisfies this condition,
but we will imagine a function '" which is everywhere zero except at the
point x = eX and we would consider this as an eigenfunction of the preceding
equation. In order that this has a sense one has, moreover, to demand that
the integral over the product of this function and an arbitrary function
f(x) is not zero. This makes it necessary that the 0/ becomes infinite for
x = tX. We will take this integral equal to f(a):
(J,

f(x)I/I",(x)dx = /(a).

If we permit the existence of such functions we observe that the spectrum .

of ~igenvalues and eigenfunctions of the operator x is continuous in the
domain D as the value of r:t, is arbitrary. Our function ",{x) coincides with the
Dirac delta function lJ(x - ex) . It is possible to construct analytical expressions
that have this function as their limit 2
We have obviously
LI/I",(x)I/Iix)dX = 0,
1

The label i in (1.20) has to be replaced by a parameter E that varies continuously

f '1'; (x)'PI:(X)dT -+ f'P*(X, E)1p(x. E')dT

and we will see below that the Kronecker Ott !nust be replaced by a new ki.nd of function
of the variable (E-E').
~
2 For explicit representations see HEtTLER, [1957], section 8.
This function ,vas accepted "reluctantly by mathematicians because of its non-rigorous
definition. It ,vas known, ho\\:cver, that the results obtained via the a-function could
alway be repeateu without its use. Later a French mathematician, L. Schwartz, incorporated the underlying ideas into mathematics by defining a g~neraHzed notion of the idea of
a function (LIGHTHILL [1958]).

eh. t, 5]

OPERAT()R~

.
.
SInce
one ot,....
trle f unctIons

25

b vD.C' IS
. dff~
zero Whtell tne Gtier
h erent fjrom
zero . So we Inay qualify the sym.bols !/lo..(X) ~~ o(x-o:) to a certain extent as
orthogonal and normal function~ depen.ding on a continuous: index Ct" l"'he
"norm" '} hO~Never, is in...'inite as \V{~ can see fron):

if now

Xl

1"

oeCOnlCf:~

x..?, the hnornl" is

11

i)(O}.

The second example will be given \:vitli Fourjer integrahL It \vill be useful
in the next chapter, 'Ve kn.ow that to ext.end the d01TIain uf validity of the
f'ourier series to an arbitrary interval -- tZ, +d (0 i.s rcal) it is 5ufficient to
make a change of the variable: t = ,)Lt/a~ In ih.ts \vay we obtain for functions
ljJ defined in this domain, having the periodicity 2a and satisfying certain
continuity conditjons~ the 'veH-kno\vn de v elopm.:nts'
~oc

tjJ(x) = (2a)-t

Cn

exp (inn:.x:/at

(n

JS

an

integer)~

The orthogonal basis functions are the exponentia.ls

' I",)"
'
'a
):
(1 rv
""Q} exp ()nnx/
they form a complete set. These are the eigenfunctions of the operator

--i(d/<L",')s because we have

-i(d/dx) exp (innx/a) = (n1t/a) exp (innxia)~

has to be an integer in order that the fu.uctions are periodic over a distance
2a. The normalizing factor 1/
decreases wIlen the domain D extends.
If we let a go to infinity it is convenient to introduce~

II

J2ii

Pv L1v
\Ve obtain

== cff~,i n/a.

VEcrOR SPACES -

26

and if a -+

00

t/I{x) == (2n)-+

f::

UNITARY GEOMETRY

py exp (ivx)dv, p., == (2tt)-i-

f::

[eb. 1, 5

",(x) exp (-illx)dx.

(1.32)
These are the formulas of Fourier where the index v is a continuous vari
able. The series development is replaced by an integral.
l"he formulas similar to (1.23) and (1.31) are

('" "') =

+oo

-00

",",dx

f+oo
-00

P:fJ.. dv,

and

as one can easily verify.

The pree~ding argument is not a proof but the passing to the limit <Xl can
easily be justified 1 As a result we see that the eigenvalues y of the operator
-i(dldx) form a continuous spectrum in the interval - 00, + 00. The
corresponding eigenfunctions:
Lim .J-d11/21t exp (ivx)
.4v .... O

~ve

a normalizjng factor that goes to zero if the domain D goes to infinity i.e.
in\ the limit their "amplitude" is zero.
tWe have seen in these two examples the diflculties that will occur if one
d.teDds the idea of eigenfunctions to the case of continuous spectra. It is
therefore better in physical problems to avoid if possible the continuous
spectrum by a limitation of the domain D. The theory of black body radiation
is a well-known example of this procedure: Conowina Rayleish and Jeans
one supposes the radiation enclosed in a recta.nplar box with perfectly
~waUs. The ampli~de .; of the waves is zero at the boundaries and
therefore may be developed ill a Fourier series and it is not necessary to use
an integral expression.
A similar procedure is used in solid state physics where the Born-von
~

periodic boundary conditions enable us to use a Fourier series also.

lti. the general case in which we rulve a juxtaposition of a continuous spestrum and a discontinuous spectrum of eigenfunctions the last, although
1

See the standard

book~

on

c~~d(:Ultu

(e.g.

WElf,STER,

[1955] p. 153.)

Ch. 1,

S1

OPERATORS

27

consisting of an infinite number of functions, does not form. a complete set.

The combination of the two spectra, however, does form a complote system.
The development of an arbitrary function 1 consists then of a sum of a series
corresponding to the discontinuous spectrum, and of an integral providing

the continuous spectrum.

1/1 -

~ p" I/It +

(J1. '"J. dJ.

(1.33)

a formula that includes as a particular case the series and the integrals of
Fourier.
1

At least the functions that occur in physics.

Comparative table of the properties of function space and If-dimensional vector space
Orthogonal basis vectors

Components of a vector

l
I

----_.._------

00

flJl f!'s . !f'" ...

~ = S fJi'{;

rx",,, i;xle(

(1.1)

i= 1

'f =

ly=:Ey/e,

1,1

(1.21)

2: YtV~

j L.;-ngth

or norm of a vector

023)

{J.3)
(x' y) =

Scalar product

2: >:: y,

(U8)

I
I
,~.,
1-------_._----.----------:.- ------"---'- ---- - .
Express,ion for one (;omponent

: :/;!-I

Linear l)perators (mappings and

coordinate transfonnations)

Bilinear forms

Hermitlan forms

c'-x

)';

17.. ,=

(Uk)

- ;

"'""

.,
2.h (jitX~

.-\,t'k

(x . Ay) -~"

O.22~

~~~ t'~(!t;~.J:1f

A{:::xf,t'.)

I
. .----- - . -----------------~-l------I
I
I

(ej' x)

Xl

P';'= L (,i!P~,

(I A)

=~ 1: e, 0:, '.
L

-'(I'

(1,24)

1p~ "~, Arpl: "= l~ V-',(!ik

(1.25)

(ifh.;}.

(1.28)

i,k

(x ' Ay) = (Ax -

(!.ll) I

(If' Alp}

'.~

(A1p' rp),

(1.29)

I. --------------------------------- -I .---- ---.(~:~.:~:--- "- -'-~~l~~-I-~----------~~:~-=~:~~~ .--.------------~--~--to main axes

I
Ax,
I Reduction
E~genvalues
!
A}f ,= :E !XiX~
(l. 1l a)

L=lge:l~t~~~.

C{.Xi

___,_~______!_________
X'

XI

1 ______ _

(1.12,>

a:,~-.

<I,k.

('P'A'P)'=2;(Xd3~{J,

(l29a\
(1.30)

(1.31)

'0
p:'"
iO')

VI

CHAPTER

'rHE PRINCIPLES OF QUANTUM MECHANICS

1. Waves
1.1. CLASSICAL WAVES

In classical physics wave motion in a continuum is described by the threedimensional wave equation,

(2.1)
A possible elementary solution of this equation is given by

",(x, y, z,

t) == A exp {-2niV (t- IXX+P;+'VZ)}

(2.2)

where A is the amplitude, 11 is the frequency, c the velocity of propagation,

and CI., P and y the direction cosines of the normal to the plane wavefront.
The wavelength A = c/v which we prefer to replace by the wave vector k,
which has the components
k" = 21tfXV/C;

k, == 2ttpv/c;

k z = 2TCYVjC.

(2.3)

The magnitude k is 21t times the number of waves per em; the direction of Ie
is the direction of the wavefront.
Each complete solution of the wave equation (in a box or in infinite ~e)
can be obtained by superposition of elementary solutions. The function
that satisfies (2.1),
00

t/I(x, y, z, t) =

"=--00

Ak exp i{k r-2?tvk t}

(2.2a)

represents an arbitrary disturbance which may be electromagnetic or elastic.

The right--hand-side of equation (2.2a) represents a Fourier series or, in
case of an infinite space, a Fourier integral and either side of the equation
describes at a given moment t a spatial distribution of waves. If l/I(x, y, z, t)
29

30

THE PRINCIPLES OF QUANTUM MECHANICS

[eh. 2,

is substantially different from zero over a localized region the function is

usually called a wave packet.
If we take together all waves of the same frequency, that is of the same
absolute k-value, but of different directions (2.2a) becomes
GO

Y,

= L: a.t(x, y, z) exp {-21riVlt}.

(2.4)

t=O

The coefficients of (2.2a) and (2.4) are generally complex because the
different spectral components usually have different phases. The relative
phases will determine whether we will have constructive or destructive interference, one of the most striking properties of waves.
The product a: Qt(x, y, z) is the intensity of a particular spectral component Vic of the wave at a certain point x, y, z. On the other hand, if we look for
the average value
of l/I* '" over a time T long compared to the
periods 1/'1k, we have:

<"'. '">

<1/1*1/1>

= lim

! L: fTa:ase2'rl

T-+~ T

",J o

(Vr-v.)t

dt ...

<"'. "'>

L: a;a,(x. y, z)

(2.5)

,.

or,the average intensity

at a point (x, y, z) is the sum over all the
intensities of the different spectral components.
The expression (2.5) is obtained by integrating separately the terms in
which ", ~ v" and those in which 1', = V The first set of integrals are oscillatingffunctions of T with a constant amplitude. After multiplicatibn with
T- i the product goes to zero for T -+ 00. The second integral is proportional
toT hence it is unnecessary to take the limit of the product.
1.2. QUANTUM MECHANICAL WAVES

Early in the development of quantum mechanics a postulate was introduced that the energy was proportional to a frequency

or
where II

E = hv

(2.6)

E = flO)

(2.7)

== h/2n,

It was successively rea1ized that the nature of the wave was not always
electromagnetic and an additional postulate

(2.8)

Ch. 2,

11

WAVES

31

which was originally suggested by electromagnetism and relativity, was

proposed for matter waves.
The result is that a particle is associated with a quantum wave. In particular a free particle is represented by:

21ti (1'. ,.-Et)}

J/! = A exp {h

(2.2b)

The complementarity principle, that is the statement that electrons can

have either a wave aspect or a corpuscular aspect but not both, can be
clarified very beautifully by considering wave packets. Consider the following

two cases.
a) Ifwe have a single wave (2.2b), that is one frequency only, the wave is a
6'monochromatic" wave, the momentum is exactly determined by (2.8) and
the position is completely washed out;
b) How'ever, we do have superpositions like (2.2a) and the possibility
exists to choose the spatial distribution of

t/I*t/I

L a:a e

2xi

(Yk-"l)t

k.l

such that "'--1/1 is only different from zero in a certain region of space, say at
t == O. If we now interpret the intensity as proportional to the probability to
locate a particle, we obtain in this case a reasonable determiaation for the
position of the particle. This occurred at the expense of a well-determined
mom~ntum since we must use not one but many differen~ k-values in our
wave ~ packet.
It}s interesting to notice that the question as to which of the two pictures
is adequate is completely determined by the experiment. This is a special
example of a general pattern iII: quantum mechanics viz. that the choice of
the ~ave function is dictated by the experiment. This choice is called the
state
the system.
If one tries to construct wave packets with the smallest possible spread in
velocity and position, one is led to a Gaussian packet which has tho .property
that the product of the root mean squares of the momentum an<i of the
position obeys Heisenberg's uncertainty principle. Hence (2.2b) or a superposition of similar waves gives a complete and consistent description of the
free particle.

of

1.3. THE FREE PARTICLE

Which equation has (2.2b) as a solution? At first one would be tempted to

.. !uote (2.1) but it turns out that by introducing a slightly different form one

32

THE PRINCIPLES OF QUANTUM MECHANICS

[eh. 2, 1, 2

can give the formalism an analogy to classical mechanics such that one can
generalize to a non-free particle, i.e. bound by some potential field, in a
natural way.
.
The function (2.2b) is a solutio~ of an equation of the first order in time.
(but with an imagi~ary coefficient), i.e.:

~ V2 t/!
2m

= h ot/!
i at

(2.9)

and by comparing this with the classical energy momentum relation

2

L=E

2m

we find the following correspondence:

f, 0
H+-+--i at

Ii

p~-.V.
1

(2.10)

(2. lOa)

Hence by introducing (2.10) and (2.10a) and by proclaiming that:

A physical f/UIIIUltY or obseruable is an operator applied to a wave function~
we can drop the postulates (2 7) and (2.8). The reason is that the operators
(2.tt) and (2.10a) can be related to certain eigenvalues, and upon substitution of the eigenfunction (2.2b) in their respective eigenvalue equations

= Et/I;
Popt/! = pt/!

Hop'"

(2.11)

(2.lla)

one obtains (2.7) and (2.8). Since the operators are Hermitian, the eigenvalues are always real (comp. Cbapter 1, 5). There arc of course circurp.stances in which the state is not pure (monochromatic), i.e. that a distribution of eigenvalues is found and thus the physical observable is not sharply
determined. The interpretation of such a case will be discussed in S of this
chapter.
2. De Scbriidinger Equation
In this section we discuss the ScbrOdinger equation, i . e. the general wave
equation for conservative mechanical systems, and also we will introduce the
many-particle equation.

Ch. 2, 2]

THE SCfIRODINGER EQUATION

33

21. SCHR{)DINGER EQUATION

A conservative system can be described by a Hamiltonian. According to

(2alO) we have
-h -OV! + Hl,fI ::; O.
(2.13)
i ot
The Hamiltonian can usually be separated into a kinetic and a potential
energy:
H

= -1 p-+ V,
?

2m

2.14)

hence we have the following equation, named after Schr6dinger

-:Ii1 01/1
-a-t -

112

-2 V t/J + V(x, y, z, t)t/J == O.

m

(2. 13a)

If V does not depend on t we obtain in general an infinite set of functions

l/Ik{X, y, z) such that the operation H is reduced to multiplication with a real
constant, i.e.
(2. IS)
If we are able t~ determine Ek we find by integrating (2.13a) that t/lk is a
periodic function of the time t. Hence we call this the stationary state of the
system:
.11 ( X, y, Z, t) = Y'k
.11 ( X, y, Z )e - iBJet/1
(2.16)
'l'k
In this state the energy H has a well-defined value Ek The spatial function
t/I{x, y, z) is a solution of the time independent Schrodinger equation:

V2y,+ ~~ (E- V)/I = 0,

(2.17)

which can be obtained by separation of variables from (2.13a), provided, of

course, the potential energy does not depend on t.
After introducing the potential energy the eigenfunctions of H are no
longer simultaneous eigenfunctions of Pop. This has a very good physical
reason because if we would ask for the value of the momentum at a fixed E
we would find that it is no longer a. constant. Hence if we apply Pop to the
eigenfunction we do not expect to find a time independent distinct eigenvalue.
2.2~

THE

n-PAl~"rICLE

PROBLEM

It is important to realize that one has to add a new concept in order to

handle two or more particles . V is now of the form V(Xl' Yl' Zl' X2' Y2, Z2,

THE PRINCIPLES OF QUANTUM MECHANICS

34

[Cb. 2, f 2. 3

, t) and V2 is generalized to
2

02

rt+f1:z..

==-2

OX t

82
+ ."."2.' ..

(2. 13b)

DX2

We have instead of (2.13a)

Ii 01/1

7 -- 1

at

#1 2 2
Vi t/I+ V(Xl . t)t/I == 0
12m,

L-

and the time independent equation (2.17) becomes:

L,2m,
~- v;.;.r+(E- V)t/t = o.

(2.17a)

An n-particle system in quantum mechanics is described in a 3n-dimensional

space: the configurlltion space.
Since the confiauration space is infinite and since the potential goes to zero
at large distances 1, the eigenvalue spectrum consists of a continuous and a
discontinuous part 2
The continuous eigenvalues correspond to functions which are everywhere
l,)Qunded, but which are not square SlIm mabIe since they do not So to zero
at iofinitity. In the case of a central Coulomb force they correspond to
hyperbolic orbits. The discrete eigenvalues form. a denumerable set ftnite or
infinite.
In the last case they approach the zero energy value rapidly when they
,
hav~ their largest amplitude at places farther and farther away from the
attr~tive center.
f

3. Angular MOJDelltum

3:J. OPERATOR

The only physical quantities that we have studied so far from a quantum
point of view are the energy and the momentum of a particle. They appeared
as operators acting on the wave function.

It

H == - - -- ,.
i

ot

p~

a
i ax

It
== - - ......

It is easy to define in the same way the components of the angular momentum. Let us consider a particle of coordinates x, y, z and momentum
p~, p.,;-P. Classically the components of its angular momentum -with.
respect to the origin are Lx == YPz:-zp,; L, = zpx-XP.; L. == xP,-YP:._
Provided of course the (pi-..wiso continuous) potential function has resioDl in
which it it su1JlcieDtly negative, otherwise the discrete eipnvalues may not occur at aU..
This is Dot the case for harmonic oscillators (compare Chapter' It 5.3.
t

ANGULAR MOMENTUM

Ch. 2, 3]

35

To translate them into quantum language we replace p, and p. by the corresponding operators and obtain

-z ~-) ,
1 (Y\ ~
ay
vZ

L. = Ii

etc"

(2.18)

3.2. OPERATORS AND GROUPS

Now we- can connect the operators and the physical quantities which they
represent with some simple groups or rather with the infinitesimal operations

or infinitesimal transformations which generate them.

First let us consider a small, so-called virtual, translation 8;rr that displa<=cs
without deformation the spatial distri~ution of the function .;. After this
translation there corresponds at the point x, y, z a value Ort; which belonged
originally to the point x-8x, y, z. Then we have 6", == ---(8t/1,ox)8x. ~The
component Pll of the momentum is the differential operator ofthis infinitesimal transfonnation multiplied by iii. When the system contains f particles
and undergoes a total virtual translation:
6x == 6Xl =- ~X2 == == 6xf the corresponding variation of '" in tM
configuration space can be written as

~'" ==

f o! t5x
'-1 OXI

and the projection of the total momentum P~ on the :x LUi is represented by

the differential operator of this transformation multiplied by iA:
P ==
Je

If
iJ
+-L-==
i
OX,

h b
----Of
i 6x

Lt=t us come back to the case of one particle. If we perform a virtual

rotation a81r of the distribution of the
oy == -z69%,6z == ybOx we obtain:

t5r; - - or; ox- ar; .5y- ~ ~z

ox

oy

fjz

t/I
==

around the Ox axis: 6x

== 0,

(z ~oy - y~)
';68:4"
iJz

If we compare this formula with (2.18) we see that the component L*

of the angular momentum is represented by the differential operator of its
right-hand side multiplied by ill.
We have

36

THE PRINCIPLES OF QUANTUM MECHANICS

[eh. 2, 3

or symbolically
L

n~
=- i bOx

(219)

This definition is general. It can be extended to a system containing any

number of particles provided one calculates first the modifications of t/I in
the configuration space. We obtain immediately

Lx

hL (0
Yi -

="7

1 1

OZi

a\

-Zi - )

(2.18a)

OYI

Hence there exists in quantum theory a correspondence between the momentum and the operator which 8enerates the translation group, as well as
between the ~gular momentum and the operator which generates the
rotation group. The equation (2.10)

h a
H= - - i ot

shows that the energy is the operator whlch generates real translations in
time. Tbis remArk about the connection between a group and certain operators will be generalized and treated in further detail in Chapter 4, 1c2.
3.3. COMMUTATION RELATIONS

A~rding to (2.18a) we have:

(LxL;-L,Lx)t/I

/ L [(y OZa -z oy
(1)'(z oxa -x oza) =._,,2
== _,,2 L (y!. -x~) t/I,
ox oy

(a
a\ (iJ
a)]
z ax -x ilz) Y oz -z oy '"

from which it follows that:

T 1 __, _,,_ 11 r
L x T y __ -"-'y
..... x
v::;)
l. ....

~,_e

-:-

(220)

Ch. 2, 3,. 4r

POSTULATES OF QUANTUM MECHANICS

37

These are the fundamental commutation relations for the operaton which
represent the components of the angular momentum. One can prove by a
similar calculation the commutation relations of Heisenberg.
Pxx-xPx

It

== -:-'

P.xy- yp~ = o.

(2.20a)

The operators discussed above are corresponding to the most CODlIl)OD

invariants in classical mechanics: energy conservation, conservation of
linear momentum and conservation of angular momentum. If a certain
number of these quantities happen to be conserved in a problem, for instance
Lx, L, and L. the three components of angular momen~ then we can also
say that any polynomial in L,%, L)' and Ls:

p == P(L~, L" L,)

is a quantity that is conserved. In the classical case such generalizations are
not very interesting since they can be reduced again to the fact that the
components of angular momentum are separately conserved. In quantum
mechanics they are very helpful as we will see in Chapter 6.

4. The Postalates of Quantum _

We have now reached a point where we can aenera1ize the results which
we have: obtained and thus make a table that lists the fundamental postulates
of quan~um mechanics. We shall use for this a geometricallanJUl.ge, i.e. the
theory Qf function space which we have discussed in Chapter 1, 4.
Let ':JS make two preliminary remarks:
1. The operators A to which a physical meaning has been given, either
differential or multiplicative, are all linear and Hermitian (Chapter 1, S.1
and 5.2). This is obvious for real multiplicative operators. For the case of
momentum and energy which are differential operators we can easily verify
equation (1.29) by partial integration provided that '" and tp vanish at the
boundaries of the domain D 1.
\Ve shall see later why a physical quantity can only be represented by an
Hermitian operator.
2. The SchrOdinger equation, (2.11) or (~.17), and more generally equation (1.30) for the definition of the eigenfun~ons of an operator are linear
and homogeneous. Thus the wave functions are defined except for a multi1

To verifyeq. (1.29) for the Hamiltonian H, one can use (2.14) takinl equation (2.108.)

into ai:Count.

38

THE PRINCIPLES OF QUANTUM MECHANICS

[eb. 2, 4

plicative constant; hence they are not really vectors of the function space
but directions or rays. If we wish til t/ld~ to represent a probability we must
suppose that the functions V1 are normalized, i..e.
".

j I{!*t/ld-c =

1.

With this condition ~'" is completely determined except for a multiplicative

constant which has the absolute value 1 and which can be considered as a
phase factor.
In a certain sense quantum mechanics provides a dictionary of relations
between the classical .representations, which are easily seized by the imagination but do not fit exactly the facts~ and an abstract formalism which allo'Ys
us to predict exactly all that is predictable - at least for our present knowledge. The quantum formalism can be expressed in two different dialects:
one is connected with the wave picture and the other (which is absolutely
equivalent to the first) with function space and matrices. This is the reason
why we shall present the first four postulates of quantum physics in the
form of a tri-lingual dictionary.
I

II

State of a system

Physical or observable
~tjty

Wave function tp

Ray1p of the function space

Hermitian operator A
acting on 1p

Hermitian matrix mapping

the function space onto
itself

---4I----------------~--~-----------------'-------------------III

! Eigenvalues of this matrix,

Eigenvalues ex, of this

operator or characteristic
constants of the equation:
A1p = tt'fjl
II

Observable values of
this quantity

I i.e.
diagonal elemellts
this matrix, which
I

of
has
been put in a diagonal
form by a unitary trans..
formation of the coordi(1.,,,

_________________~f__________________~I_n_at._es______----____
IV

State of the system

where A has a definite
value a.t

I
I

Eigenfunction of A corresponding to t.he eigen..

value Ct,

: . - . ._ _ _ _ _ _ _ _ _ _1 _ :_ _ _ _ _ _ _ _

Ray of the function space

which is multiplied by (It
when the operator A acts
it
..

lon

v.. ~oreover we shall allow, following DIRAC [1958], that every Hermitian
operator has a physical meaning!,
lOne can find in the book by VVEYL [1950] (Chapter IV D, section 14) a remarkable

argument that justifies this postulate or at least connects it to a very general and natural
irreducibility postulate.

Ch. 2, 4]

39

POSTULATES OF QUANTUM MECHANICS

It follows from the postulates III and IV that if a system is in the state t/I"
a measurement of the quantity A gives with certainty the value Cl i
VI. But if t/J is not an eigenfunction of the operator A we do not know
with certainty the value of the physical quantity which this operator represents. We can expect to find the different possible values (Xi' CCt with
different probabilities. In order. to make our postulates more precise we
have only to generalize what has been said in 1 of this chapter: '" can be
expanded according to (1.33) in which we shall neglect the continuous
spectrum for simplicity. Then the state appears as the superposition of the
pure states t/I". Each of them is weighted by a coefficient Pic whose modulus is
~ 1 and each corresponds to a given value (%k of the physical quantity A.
JJTe shall assume that a measurement made on the system in the state'" call
: yield any of those values; the probability that this value is (Xi being
fil"
This last postulate foreseen by Einstein, enunciated precisely by Born,
and developed by Dirac can be considered as the keystone of quantum
mechanics . It is the link between experiments and wave theory, giving to
the latter the character of an essentially statistical theory.
If we repeat-the same experiment a great number of times with identical
systems all in the same state t/lwe shall find for the quantity A either the
value (Xi' or <X", or (Xl and the relative frequency of these different results will
be in the limit equal to the corresponding probabilities. The average value of
A in the! state '" is:

P:

(A)

= L CtjPiPt

Lfa.iP:ifJtPi1/!id1:

IjI*Al/Id1: = (l/I Al/I).

(2.21)

an imvo.r~~t formula due to Dirac.

Let us suppose that the function space is spanned by the eigenfunctions
cP 1 , qJ.2, of another operato~ ~. In this system of axes the quantity A is
represented by a non-diagonal matrix (aUt) defined by:
AlPl

= L !Pi aUt;
i

The state

t/I can be considered as a superposition of states (J'k

VI

L YklPk
k

We have

(1.25)

40

THE PRINCIPLES OF QUANTUM

ME(:H.~NICS

[ell.

4, 5

and according to (2.21)

(A) =

f"'*At/ld-r: = ~ f 'Y~q>~q>laik"lt,d-r: = ~ aikY:YIc'

III

I. ') 1

(2.22)

lk

(A) is then a quadratic form of the Fourier coefficients Yi. This form w uicb
represents the average of a number of experimental results is necessal"ily real:
it must be Hermitian, since the condition that ail "'Ii YIc+aki ')1t is real implies
III
all == ald
Hence we see the reason why a physical quantity must be represented only

Y:

by an Hermitian operator.
Let us suppose that the state we consider is an eigenstate of the operator
H, for example '" = lpn' 1ft = 1, 'i'lc = 0, k =I: n. From (2.22) one obtains
<A)

= ann.

(2.23;

If the operator A is represented by a matrix QfIc in the basis fPl' tfJ",

.j

its expectation value for the state cp" is the element a",. of the main diagonal,
In the foregoing sections the time" t has not been explicitly considere(
since the wave functions which have been used depended only on the coordi
nates~ Now' we shall study the change of the states and the physical obser

vables in time. This will allow us to elaborate the meaning of the equationc
(2.22) and (2.23).
5. TIme Dependence of a State and of a Physical Observable

s._. GBNERAL THEORY

The easiest way to tackle this subject seems to be the following. Le1
l/J(x 1 , Yl' Zt, X2, " ~, t) be a wave function which will be written mort
concisely as t/I{x, t); its variations in time follow from "the propagatioI
.equation" (2.13) or (2.13a).. As in classical non-relativistic mechanics tht
time ~ is considered as a parameter and one deals v,ith the function spaCt
';(x), which are functions ,of space coordinates only_ To span the functioI
space one chooses fixed axes which are a complete set of orthogonal functions
forcxample the eigenfunctions t/I,,(x) of the Hamiltonian operator H, i.e
the solutions of the SchrOdinger equation (2. 17a) which we shall call, follow
ing :Qirae, the Schrodinger axes.
In the expansion of an arbitrary wave function ljJ(X., t)
I

t/J(x, i)

= L y"Jt)"',,(x),

(2.24
'\,

the coefficients y vary with the time. A conveniellt picture is, the foHo\ving

Ch. 2, 5]

TIME DEPENDENCE OF A STATE

41

the vector tfJ{x, t) of the function space whose length (1.19) is equal to unity
rotates around the origin obeying equation (2. 13a). The problem is to
determine the expression i't(t) of its components along the fixed axes
"',,(x). We have

HI/I

= - ~ at/! =
1 at

~ L I/Ik dl'l
1 A

dt

and since the 1" are ordinary numbers and the t/lt the eigenfunctions of H
we have
HifJ = Ly"H.p" = LYkEkt/l".
k

"

Identifying term by term one obtains

Ii di'k

- - = E"l'ki dt
Finally by integration we obtain

(2.2S)
and -

(2.258)
the f1 being constants which may be complex.
'nie 1"'S are always harmonic functions of time with frequency 'Vi ==
27r.4/1J. In a limited sense equation (22Sa) looks like a Fourier expansion.
Particularly if tP is a steady state wave function, this expansion reduces only
to one term (Ct == 1, c, == 0 for 1 #: k) and we find again (2.16). A steady
state is an harmonic oscillation along a Schr6dinger axis, this oscillation
does not give rise to any radiation and therefore cannot be detected by
experiment; since the modulus of the phase factor exp( -27tivJ:t) remains
equal to unity, the system remains on the same ray of the Hilbert space.
If the state is not steady, i.e. if there is more than one non..zero term. in
(2.25a), the motion in Hilbert space will be some complicated rotation.
Let us su~se that the Hilbert space is spumed by another complete set
of ort!to8onal functions tpJ; the transition from the SchrOdinger system tp is
obtaUted by using a unitary transformation U == (uJJ not depending on time.
The inverse transformation corresponds to the operator U-1

42

THE PRINCIPLES OF QUANTUM MECHANICS

reb. 2, 5

By expanding l/I(x, t) in a series of the functions !fJj(x) we have:

t/I(x, t)

= L 11j{t)<PJx) = L CkV!k(X) exp (-2nivkt)

J

from which it follows:

tlJt)

L Ujk 1 Ck exp (-2nivkt).

(2.25b)

In the basis qJ, each of the components tfJ(t) of the wave function appears as a
Fourier expansion.
5.2. HEISENBERG REPRESENTATION

In expression (2.25a) it is often convenient to attach the periodic factor to

the functions 1/11 instead of the coefficients Ct) i.e. to consider that the function space is spanned by variable axes wbich will be called Heisenberg axes.

I/Ik(X, t) = t/lk(X) exp (-2nivt t).

Then the state t/I(x, t) appears as a superposition of steady states t/!1c(X, t)
where coefficients c" are independent of time.
This change of axes has certain consequences with respect to the representation of the physical observables. This transformation corresponds to a
change in our point of view which makes it possible to indicate a more
precisp connection between wave theory and matrix mechanics and even
classidal mechanics.
J . ...
,
Welhave used a Dirac matrix as an abstract representation of a physical
obserYable A. The elements alit. of this matrix are certain constants defined,
for ~xample in the Schrodinger coordinate system by the equations

(1.25)
We.have been led to this representation because the possible values of a
physical observable (eigenvalues of the corresponding operator) are constants independent of the stat~ of the observed system. But physically the
m~urement of the quantity A is done for a real material system, the siate
of which varies according to the equations (2.24), (2,,25) or (2.25a). The
probabilities of the different eisenvalues at of A and its expectation value
(A) are functions of time.
For example (2.22) and (2.25) give for the Schrodinger axes
(2.26)

43

TIME DEPENDENCE OF A STATE

Ch. 2, 5]

Hence we shall not define the matrix which represents an operator A by its
action on the invariable wave functions t/lk(X), but by its action on the com-

plete stationary wave functions. Equation (1.25) becomes

Al/Ik(X, t) = Al/I,,(x) exp ( -i ~k

=

l/IAx, t)aJ" exp

t) = ~ l/IAx)aJk exp ( -i ~J: t)

(i EJ~E'k t)

Thus we get the Heisenberg matrices

(2.27)
which ~differ from the Dirac matrices only by periodic factors of modulus
unity. One can write

(2.26a)
On the other band, (2.27) gives

h dL4Jt
i dt

EJ-E" AJi:

or ufing the matrix notation, EJ being an element of the diagonal matrix E,

AdA
- - == EA-AE.
i dt

(2.27a)

The transition to the arbitrary system of orthogonal functions If'J is made

by using the unitary transformation U. We have according to (1.27a)
E -.. H = (HJ,,)
A~Q

:=

= U-1EU;

(qJt) == U- 1AU ==

(L uji1AJmu,.).
1m

The coefficients qJIt. are complicated functions of the time since the sum. LIM
is a series expansion which involves all the frequencies (Oil; equation (2.27a)

becomes:
It ~ == h U- 1 ~ U _ U-1(EA-AE)U
i dt
i
d,
== u- 1 euU- 1 AU-U- 1 AUU- 1 EU

44

THE PRINCIPLES OF QUANTUM MECHANICS

[eb. 2, I S

or
AdQ
:= HQ-QH
(2.28)
i dt
the well-known equation of Born, Heisenberg, and Jordan 1 We shall
-

rarely use this equation. The foregoing discussion is merely a chango in

notation, or a change of axes in function space. Considered as rays of the
Hilbert space, the Heisenberg axes are not essentially different from the
Schr6dinger axes. The formulae (2.22) and (2.26a) differ only in the way
the time factors are grouped on the right-hand side. For us, the first one is
more directly connected with the principles. The time variations of (A)
, are apparently connected with the chanses in the state of the system and
with the coefficients Yk(t) of the expansion of y,(x, t).
In order to understaud better the meaning of the expressions "expectation
value (A) of a physical observable A "and "probability 1')',1 2 of a possible
value <x,", we must state more precisely what is understood in quantum
theory by "measurement of a physical observable in a given system".
To perform an experiment under well-defined conditions the observer has
first to deteriirine, by the way he sets things up, the state of the system he is
studying. At the initial instant this state is represented by a wave function
1/J(x,O) which changes according to the Schr6dinger equation (2.13) and
,becom~s ",(x, t) at the time t of the measurement. We d,o not suppose that
",(x, t)! is an eigenfunction of the operator A. At this time the observer
chang~ssuddenly the state t/I(x, t) of the system and transforms this state into
an eiFnfunction
of the operator A by the very performance of the measure..
r
meq.!. The observed value is af. He performs again the experiment and starts
always ftom the same initial state, acting always at the same instant t.
This intervention brings according to quantum theory an unavoidable
element of chance despite the fact that it is always the satl\e: it gives
sometimes the state '" t, sometimes "'A:. . .. But after a ,great number of
experiments tho frequency of the different possible results allows us to
attribute a definite probability to each of them, and to calculate their average
(expectation value) at the time t. These probabilities and these expectation
values are what the theory can predict "a priorin using the expansion of
1fJ{x, t)
a series of eigenfunctions t/J i

in

1 This equation which is lO~es called the equation of motion of matrix mechanics
leads to a direct connection between quantum and classical mechanics. Classical equations
such as Newtons law, the virial theorem, and the Lorentz force can be mimicked by UBiD,
matrix notation. (Compare SCHIPP [1949} Chapter VI or DIRAC {1958] Chapter Vl.)

Ch. 2, S, 6]

45

TRANSITION PROBABILITIES

In order that an experimental intervention of the observer can be considered as a sudden intervention it is necessary that the rate of spontaneous change
of the system be small with respect to the rate at which the external perturbation is established.
A very simple case is when ,one considers the eigenstates "',. of the Hamil.
tonian operator H: all the coefficients in the expansion (2.2Sa) are zero
except en the modulus of which is equal to unity. The equation (2.26) shows
that (A) = ana is independent of the time t and can be considered as a time
average of the quantity A for a system in the state "',.~ Let us clarify the
meaning of this statement: If a series of measurements of an arbitrary
quantity A is performed for different syste~s of the same kind and in ,the
same state t/I,. their average does not depend on the instant at ~lhich each
measurement is done. This is the reMon, why the states corresponding to the
level En are said to be stationary states.
In the old quantum theory the meaning of the expression Htime average"
was rather intuitive. It is not.possible to indicate such a thing for a quantity
represented by a non-diaioiiai matrix.
One ~ust note ttiat the stationary states of the atomic systems with which
the qWmtum theory is principally concerned are onlyapproximalely'stationary, provided "the ~tion dampingu is neglected (compare the next
section). As. the frequency increases this . damping becomes more and more
imp~rta.nt, it accelerates more and more the rate of the sponta.t1eous' evolution br the systems, it decreases their u mean life' which can be less than the
duration of an observation.
-"

,~

:-,,",

,_

~~

~. ~

,,<

,i,

e'

6. TraDlitloa Probaldlities and Radia.t1_ Theory

The original basis of the present quan:tUtntheoty is the Bohr postulate, i.e.
an 'atom -raoi8.tes when it changes from an en~rgy level E. to a.lower en.ergy
level En; during this transition it emits a quantum ofelecttomagnetic energy
flO)

= W

II::

E". - E. .

(2.29)

To calculate the relative iatenlity-~of the different spectralJinesoae must

calculate the probabilities p of thedifferertt' :possible 'transitions. A first
attempt based on the correspondence Principle was made by Kramers.
Later,/Schrodinger using a semi~classicat wave piCtUre established the exact
formulae. Finally by utilizing, the theory of Jeans which had ~ improved
by LOrentZ and Debye~,Diraesucceeded in deriVing the tra11sitioll Pto6abiIi,ties from the general principles of quantum meoba1Dics by rigorous reasoning

46

THE PRINCIPLES OP QUANTUM MECHANICS

[Ch. 2, 6

which made the proof of postulate. (2.29) possible. We shall be satisfied

only with a short and rough justification using some considerations of the
corresponden.ce principle.
In classical theory radiation is produced by a periodic change of the
electric distribution. Let J.l.Jl be the electric moment of an atom. Let us suppose
that this moment is parallel to the x axis and harmonic with the frequency v

fA,,(t) ==

:EIL e"x/c = J.l cos 2nvt.

According to the Maxwell-Lorentz theory, the light radiated by this atom

is linearly polarized; the electric vector being parallel to Ox; tho enet'gy
radiated per unit time is: (if Jl. is measured in e.s.u.)
.
dW
2
4 2
4nv
3 2
== - 3 (2nY) J.' = - 3 (2nv) p,
dt
3c
3c

(2.30)

wh"re c is the velocity of light 1

According to the correspondence principle the emission of light is deterznjned by the el~~moment in quantum theory. Thus when the atom is in a
stationa.ry-s~te",., one must decompose this moment into its c.Wrerent
components related to the different possible transitions, i.e. ODe must form
~ matrix (p,..). This is very simpl~: Jl, as any physical observable, is an
_operatPractina on the wave function which represents the state of the atom,
i.e. on! the l/!" and according to (1.25) we have

p."'. == :E t/I"Pb

(2.31)

Ie

Th," .pl 'beingeigerifunctioDS of the energyH;are usually not eigenfunctions

of'the electric moment oPerator and hence the eigenValue of Jl is no~ determined for the level E ,In oth~r words the matrix. JI.,. has off-diagonal eleme,nts. In this way the possibility for the emission of several lines from a
given state is explained in quantum theory. Equation (2.31) is the quantum
analope of the ex~sion of the electric moment in a Fourier series. When
the 'function t/I. alld t4e generalform of the. operator 1'. are known, one uses
(1.26) to obtain the elements'of the matrix (p"".) and one gets:
, I .

--

(I/t", iD/I.) ==

J"': ~

I/t"Pud'C - ",-.

(2.32)

When m and n,arc both large and when their difference is small,.,the formula
I

QuadruPole emission II nePectecL

Ch. 2, 6, 7]

47

PERTURBATION THEORY

(2.30) must be valid. This use of the correspondence principle leads to the
following result, which can be obtained rigorously using time-dependent
perturbation theory (the so-called "Golden-rule"). The energy radiated per
second is equal to the average number of the transitions, i.e. to the probabili. .
ty of transition P mrt multiplied by the value of the emitted quantum. Then
one gets

1 dW)
P1ft,.:: ( - hv dt

4n {2nv)3
quantal1y translated as P mn = 3
3 he

Illmal.2

(2.33)

It was later shown by Dirac, that a similar formula is valid for spontaneous emission, i.e. when the atom is not subject to an external radiation field.
The preceding result is the basis for the selection rules: if
~==p~==p:'=O

the corresponding line does not exist.

If p!" =: Jl~1t == 0 but Jl:" ~ 0, the line is linearly polarized in the Oz
direction.
7. Perturbation 1beory
1.1. FOR.MULATION OF THE PROl3LEM

The numDer of problems in quantum mechanics that can be solved explicitly is rather limited, a situation similar to classical mechanics, or statistical
lI,lechanics. To study more complex questions one must often be satisfied
With successive approximations starting from the cases for which solutions
are known.
This method, is called the perturbation method: the energy levels E 1 , E 2 ,
and the complete set of eigenfunctions
of a given Hamiltonian
are supposed to be known. They correspond to the Schrodinger equation

"'1' "'2' ...

HIjI,

= Ett/l,.

(2.34)

The aim is to calculate the energy levels E' and the eigenfunctions "" of the
system disturbed by a perturbing energy which is added to the Hamiltonian H
and which is generally supposed to be expanded in a power series of the
paraineter A.. Taking into account the terms of only first order, one gets

H' == H+)'W
and (2.34) becomes

(H+AW)1/I' = E'1/I'.

(2.35)

48

THE PRINCIPLES OF QUANTUM MECHANICS

reh.

2.

One can proceed with this approximation to any order. The formal calculations remain simple, but the practical difficulties become quicldy inextricable. Particularly it often happens that the series which are thus
obtained are divergent. We will discuss here neither this kind of difficulty
nor the case in which the spectrum of eigenvalues is partially continuous.
7.2. NON-DEGBNERATE PROBLEMS

Let us suppose that all the levels E, of the unperturbed equation (2.34) are
single. The eigenvalues E{ and eigenfunctions t/I; of the perturbed problem
(2.35) are slightly different from the energy levels E, and. the eigenfunctions
'" ,. Therefore we can put
(2.36)
E; = E,+AWi
l/J~ ==

and we expand

Ui

1/1,+ AU,

(2.37)

as a series of the orthogonal functions

"t == L e""".

t/!,.
(2.38)

Using Jhose three expressions in (2.35), taking into account (2.34), and
neglecting A. 2., we get

W.p,+

L c"E,t/I, == w,t/I,+ L cnE,t/l,

I

(2.35a)

whereLY{ appears as an operator acting on the

exp8.ti~bn (lJ:2S) and write:

1/1,.

Then we can use the

Wt/I, = L tfi, Wit.

(2.39)

[Th~ matrix (Wit) represents the perturbing function in the coordinate

system

~,.]

Finally (2.3Sa) and (2.39) yield

LI \lI,(wu- w,e5,,-Q,,(B,-E,)] == 0
o
{
~n == 1

(2.3Sb)

(i#:l)

(i == I)

and as the basis vectors '" J are orthogonal, the brackets are all zero.
Then, (a) for i ~ I

(2.40)
(b) for i

=1
(2.41)

Ch. 2, 7]

PERTURBATION THEORY

49

and one can suppose Cu = 0 (in such a way that the norm of t/I; is 1).
The problem is completely solved in the first order and we have to calculate
the matrix (Wit) which according to (2.39) is given by the expressions

f "'~W""d-r: == w"

(2.42)
(2.43)

The preceding calculation is valid when AW,J(Es-E,) can be considered

as small as can be seen from (2.4O). When this condition is not fulfilled,
i.e. when the two levels B, and E, are such that Ej-E, is pfthe same order of
magnitude as the perturbation, the series expansion whose first term is given
by (2.37) becomes poorly convergent or even divergent. This is a quasi
degeneracy and the problem can be treated almost as if there was a true
degeneracy (compare 7.4).
7.3. DEGENERACY

Let us suppose that E is an a-times degenerate eigenvalue of (2.34); this

means that the equation
(2. 34a)

allows a different eigenfunctions 1/I,(i == 1, 2, ... (X) which mayor may not
be orthogonal. These eigenfunctions are determined only up to an arbItrary
unitafy transformation. One can build a: independent orthogonal linear
combinations with these eigenfunctions which also obey (2.34a). As previously, we assume that we know tIle complete set of eigenfunctions t/J of the
operator H which contains the ex functions
Generally the perturbation splits the energy level E into (l different levels,
close to the unperturbed level

"'f.

E; = E+AW,.

E;

(2.36a)

there corresponQ.s a wave function t/!; which does not

necessarily become one of the functions t/I f when A. goes to zero, but-instead
might become a linear combination of the I/I,'s.
Since w~ do not know beforehand which linear combination we will need
to make a proper start if A is becoming gradually different from zero, let us
begin with a set of C( arbitrary linear combinations of the ex-wave functions
To each level

tI

"'; =

1-1

11k

t/lk'

I = 1 ... lX.

50

THE PRINCIPLES OF QUANTUM MECHANICS

[Ch. 2, 7J

We will find eventually which combinations we really do want from the

form in which the perturbation takes place. The situation is similar to the
determination of the main axes of a circle, a problem which gets a meaning
,as soon as we .csqueeze" the circle slightly into an ellipse. This can be
formulated mathematically as follows. Suppose we try to solve the degenerate
problem in the same way as before, then ~he coefficient determined by (2.40)
would become infinite for those indiees i and 1 for which E, = B,. The
reason is that in deriving (240) from:'equation (2.3Sb) we would have divided
by zero~ which is not allowed. We can circumvent this difficulty by making
Wu = 0 for I :;:. i; i, I == 1 ... a. This condition means diagonalization of the
so-called secular matrix. The eigenvectors determine the choice of our linear
combinations (2.44), i.e. the coefficients
We obtain iti first order by taking into account (2.38),

1,,,.

ex

ex

"'; = k=l
L 'P,Ie 1/1" + AU, =k=l
L 1Uc I/Ik +1 L, Cil t/t~

(2.44a)

it k = 1, 2, ... (X

== (X+ 1, +2 ...

where the zeroth order coefficients "la, and the first order coefficients e,l

are the unknowns.

The Schrlidinger equation (2.35) becomes (after 2.36a)
!

(H+1W)t/I~ ==

or

ta~g

(E+A,w,)t/li

into account (2.44), (2.34a) and dividing by A

CX>

L
cu(E-E,)J/i, = L 'YUc(WI/I1c- Wit/!t)
1=+1
k=l

(2.45)

As in the first case let us expand W",,, in a series of functions t/I" but
writing separately the Cl first eigenfunctions t/lJ' equation (2.39) becomes

<Xl

Wt/I" == J-l
L '"J Wjk + ,.+
2: 1 .p. V,,,

is a ,square matrix with ex rows and columns and the elements

are theother'elements of the infinite matrix w. Equation (2.45) becomes

where
Va

(2.39a)

(~a)

Q')

E~' c,,(E-E,)t/!, ='-+1

L (1UcWJIt;-'YIJWf)"'j+~EvyUt.Vl1c"".
il;

i,j-1

(2.4Sa)

Ch. 2, il

PERTURBATION THEORY

51

zeros in order to change them into matrices which are defined for all the
values of the indices from 1 to 00. This allows us to put (2.4Sa) into the
following condensed form
co

2: I/Il[cu(E-E,)- 2: 1Uc W'''+YU W l - 2: l'ik V,k] = o.

1=1

(2.4Sb)

The brackets are zero since the t/J I are orthogonal. The equations which are
obtained this way can be classified in two groups:
(i) I ~ cx, El = E, vlk == 0 and (2.45b) yields

2: 11k WIt -

i'll Wi

where the

=0

( i, 1 = 1, 2, . . . <X)

(2.46)

w,,, are known coefficients (cr. (2.42) and later (2.48). This system

of homogeneous linear equations determines both the perturbations w,

of the level ~d the coefficients "Ilk. The 1a are determined except' for a
multipli~tive constant. In this kind of problem, one bas to reduce the matrix
(Wlk) to its diagonal form using "a change ofaXisn (2.44). This problem is
similar to the problem arising from equation (2.35), but it is simpler since it
does notin.volve the whole function space but only the subspace spanned by
the IX first axes '" ,_ In order that the system (2.46) has non-trivial solutions,
its determinant must be zero. We obtain the secular equation,

Wll -WI

W12

W21

W22 -Wi

Wl

I
~~: I == 0

Wt

w....

(2.47)

-wll

whose tX roots: Wl, W2, W4 are real, since the matrix (WII) isHermi~
(as will:.be proved below). To each root wJthere corresp()Jlfis3 system ,of
coefficients "Ij!, - 1jfl. which are determined by ~ homogeneous equations
(2.46) except for a multiplicative constant. One takes this opportunity to
normalize the eigenfunctions .pi- If the secular equation has multiple roots
the degeneracy is not completely removed and some Yilt remain undetermined.
The practical difficulties are purely analytical. One must calculate first
the elements Wu; of the perturbation matrix: according to (2. 39a). These
elements are given by the integrals
Wu.

f",:Wl/lk d

7:,

I S.o%

(2.48)

52

THE PRINCIPLES OF QUANTUM MECHANICS

[eb. 2, 1

the evaluation of which is generally difficult. One then has, to solve the
algebraic equation (2.47).
(il) Now let us suppose 1 > cx, W,,, = 1'u = o. According to (2.45b)
we have
ex
- ~ Yii, V,le
CU - L
(2.49)
== roll ,
i - I E-E,
E-E,
with (cf. (2. 39a))

= "'~W",,,d"C.

Vrt

0)1l

I'ik V,,, =

1 > ex

f "'~W"'fd'C.

(2.48a)

according (2.44a) and neglecting the terms in A. Since the perturbing

function W is real, (2.48) and (2.48a) show that the perturbation matrix W
is Hermitian. The preceding calculation does not involve any restrictive
hypothesis about the nature of the perturbing function. The perturbation
can even depend explicitly on time: this happens in dispersion theory whose
starting__ e~rt~~tions are (2.49) and (2.48a).
7.4. QUASI DEGENERACY

Let us suppose that all the levels are distinct and that two of them El and
El ar~ closet. Let us write
(2.49)
E2 -;- El being fixed by the nature of the unperturbed system, the auxiliary
~eter" is very large when the perturbation ;"W is negligible, and is of
the _o~der of unity when this perturbation becomes of the same order as the
difference E2 -- E 1 In this last case the perturbing action will almost
couple those two levels as if they were degenerate. As in (2.44) one is led
to put
co

l/I~ ==

I'll

t/l1 +112"'2 +A. 2: ClI"";

'-3

"'; =:

00

"'3 + l L

)=1

(2.44a)
e3l"'J

The calculation is similar to the preceding one: we choose as the initial

l'

There can be tItt'e than two.

(~h.

2, 7]

PERTURB.~TION

53

THEORY

level E = E1 and in order that the eigenvalues Wi are 0 and '1 in the absence
of the perturbation, we replace Wi by Wi -11 in the lower right-hand corner of
the secular matrix. The equations (2.46) which determine the coefficients
Yik can be written
J'il(Wl1- W i)+

'Yi2W12

+Yi2(W22 +l1- W

1'il W 21

O}

=
= 0

= 1,2

(2.50)

from which follows the secular equation

=0
the two roots

WI

and

W2

(2.51)

are given by

w = t[W11 +W22+'1+~(W22+11-Wl1)2+4w12lV21].

(2.51a)

We have finally

We find for 1 > 2 the equations (2.49) again, i.e.

If the perturbation is very weak, '1 is large and by expanding (2.51a) we

obU#n again in first approximation the formulae of 7.2,

E;

~
1'22 ~
I'll

E~ = 1 +A.W11

== El + A(I'J + W22)

= E z + AlV22

Y12

1
1

~ 1'21 ~

(2.52)

o.

We will first discuss the special case in which both WI2 and W21 happen to
be zero. '[he equations (2.52) ~emain valid no matter what the order of
magnitude is for'1 and equations (4~51) reduce to a system of two equations
of the first degree. No c0'!Jpling between the two close levels is established,
and when the perturbation becomes stronger the two curves E:' == E{(A),
E~ = E~(A) cross without mutual modification. Their intersection is at the
point w~ere
W1t

(It

i~

necessary that

=:

11+ w 22

WIt - rV22

I.e"

A=

E2-El
------

W 1 1 - W Z2

and E2 - El are both positive.)

54

THE PRINCIPLES OF QtJANTUl\1 MECfIANICS

[eh. 2,. 7

On the contrary if Wl1 :p 0 this intersection cannot occur. The two roots
are always distinct since the radical involves the sum of two squares. The
two energy curves come closer, the energy difference becomes a minimum
r
which is equal to 2-v'W12}V21 when A = (E2-E1)/(-"Vl1--W22)" and then
the two curves diverge.
It is also interesting to examine the behavior of the wave functions
during this process. It is now a simple matter to calculate the coefficients
7. We find:
yil == li2 = t[1+(W 11 - W22-,,)/D]
112

== yil

:=

![1-(Wl1 -W22 -'1)/D]

where D is the (positive) square root in equation (2.51a). For large positive
" : )'11 == 122 == 1 and 121 = 712 == O. If we let El and E2 "cross over", "
changes sign and the result is that for large negative" we have 111 == 121 = 0
and 712 == 121 = 1. Hence if the system was originally in the state E 1 , the
lower state, it will now be in the lower state (E2 ) again. This is only true if
W12 p: oand as we saw above the energy levels do not actually cross over but
do ooly approach each other closely. During'the closing in, the wave functions get mixed and after separation turn out to be interchanged.
The preceding considerations can be applied to the theory of complex
atomS, where the perturbing function represents the mutual interaction

betw@n the electrons.

They can be applied to the theory of molecules which involves also the
nuclei. This picture is also the basis for the existence of a gap in the energy
surfaces at the Brillouin zones in a periodic structure. An important example
j. the following:
Let \l$ con~der two different atoms which are coupled to form a molecule.
When the distance R between the nuclei is large, it is possible that two states
of1;lle~ystemhave nearly the same energy_ This happens when the ionization
potenti&J.of the e1ectro-positive atom is small and partially compensated by
the aftinjty<of the electro-negative atom for the electron, since a slight
increase the energy ;." is then sttfficient to transform the homopolar state
where the two atoms are neutral into the l1eteropolar state where they are
I

of

ionized 1~
A similar case occurs when one of the atoms has an excited level near the
Nevertheless the bond remains homopolar (Fig. 2. 1) but the number 1,,"1" gives, as a
function of R, the percoatage of ionic state contained ill the 4i.Ctual state -:pi and consequently its contribution to the electric mOlnent of tI\e moi~ule in the ground state.
1

Ch. 2, 17]

5S

PERTURBATION THEOR.Y

ground level. If we are describing, for instance, the interaction energy we

~~ choose for the parameter l a suitable negative power of thedistaace R.
The precise theory of molecular systems is more complex. Added to the
quasi degeneracy there is an essential degeneracy (Chapter 4, 2.2) which
comes from the indistinguishability of the electrons; this is called exchanae
degeneracy. It is the following: to any stationary state in which the electro
1 and 2 playa deftDitc role, 1 being bound to the atom a and 2 to the atom b,
there corresponds another state with the same energy obtained by pertnllting
the two electrons, 1 being bound to b, and 2 to a. 1 This degeneracy plays a.
fundamental role in the calculation of the levels by the perturbation method.
In the case of two electrons one gets a secular equation of the second dqree
which is ~ to the fourth desree if a quasi degeneracy occurs. Then the differencebetweenthetwo interacting levels (ef. (2.49) depends on thCdistaBCC
R and one is obfiaod to use some roup approximations, .Neverthelesa one
can .estimate the general features of the four curves which represent the
energy versus distance for the four perturbed states.
E'

__

--------E~

~-----------------~

.. R
R'

Fig. 2.1. Two enClIY curves as a function of the parameter R. OJf-dialonal elements are
zero at the point of intersection.

To illustrate the theory of the quasi dcpDeracy;aa~oarlYaa posai~ we

have drawn in each of the two ~ only two of thoae.curv., those which
correspond to the two lowest, i.e. more:stable states. Fipre2.1 corresponds
to the case where Wl2 - 0, Fig. 2.2 to W12 " o.

-'t

When n electrons caD be ponnuted tho c:lepDeracy is of order II! One mows
the
electrons which are permutecl can be connected to the same Ducleus but in cWrerent .~tes.
1

56

THE PRINCIPLES OF QUANTUM MECHANICS

According to (2.42) we get W12 = 0 whenever

anti-symmetrical (Chapter 4, 3.1).

1/11

[eb. 2, 7

is symmetrical,

"'2

7.S. APPUCATION: DIATOMIC MOLECULE

.H one tries to apply the seneral perturbation theory to cases (analogous to

the preceding one) where the electrons which are permuted are bound todifferent nuclei, 000 finds a difficulty which is worth pomtin, out. Let us
consider for example, as Beitler and London did, two hydrogen atoms which
are brougltt together to form a molecule. The permutatiQ,n of the two -electrons changes neither ~ expression of the energy nor its value, but only
the ptlrtition 01 the Hamiltonian into a pl'incipalterm tmd if perturbing junction.
As an example the interaction potential between the proton a and the electron
1 belongs to the principal term if one considers that these two' charges form
a neutral atom and the same interaction potential belongs to the perturbing
function after the permutation of the electrons 1 and 2 has been performed.
e'
Ea

(tonie)

(Homopotar)

E.

,..R

R'

Fig. 2.2. Illustration of the non-crossing rule: Off-diagonal elements are non-zero. Notice
the labeDing of tho curves.

It follows that the zeroth approximation wave functions are orthogonal

only jfthe'Duelei are infIDiteIy distant from each other. Fora finite but large
distance the onho!onality is only" appfeximative, which is not surprising
'becalJ.St>ttiese wave function'S are not eigenfunctions of the same Hamiltonian. At the same time the operator W which represents the perturbing
functi()nchanges when the electrons are permuted an~thus it depends on the
wave function 1/11; on which it acts. Taking into account these two facts in the

Ch. 2, 7]

PERTURBAtiON THEORY

57

definition (2.39a) of the matrix WIlJ one can easily perform a first order
calculation, following the calculation done in 7,,3 of this section. The
further approximations are more difficult.
Let us make a last remark: we have the study of non-degenerate, degenerate and quasi-degenerate problems divided in three parts, 7.2, 7.3 and 7.4,
but the method which is used to solve these problems is always the same, the
last case is the bridge between the two others.

GROUP THEORY

t.

ne Role of Group Theory in Quantum Medtaalcs

In pnerai, aroup theory is instrumental in solving those problems in

quantu.m mechanics where we have deaeneracy because usually the degeneracy is due only to the ftaCt that the Schr&:tinaer equation allows certain.
groups, i.e. the SchrOdinaer equation remains invariant if the system under
consideration undergoes a set of transformations on the variables that enter
into the wave function and this set of transformations forms a group_
Take for instance an atom with the nucleus at its center: because the field
is sp~caUYS)'Dl1D.etric, the Hamiltonian function H is invariant under an
arbitrary rotation around the ~nter. For this reason the energy levels
depend only 011 the quantum numbers n and /, whereas the wave functions
(leaving out tile spin) depend also, on the magnetic quantum number m.
~,meaJII a depneracy of orientation.
"1--<.m.lder as an example the 11 =- 2 level of a non-relativistic hydrogen
atom. i t ' . III. I values: I == 0 (m == 0) and I == 1 (m =- 0, 1) and is therefore four-fold deaenerate since the energy of such a system depends only on
n!' a part of this degeneracy is accidental, which means that it has nothing
to do with symmetry, since any radial potential different from the Coulomb
.' law wi1I aive rise to a'splitting of the 1 == 0 and 1 := 1 level. The other part is
clue to spherical symmetry, since the three levels m == 0, 1 of I == t stay,
of course, deaenerate as 10Bg as the potential has spherical symmetry.
Application of a linear electric field to this system will split these four levels
illto a pair with m == 1 and two levels which have both m - 0, but different
I values. This is explained in the following way. Take the axis of quantization
(that ~.thedirection in whiQh the angular momentutn operator is diagonal)
alOD, tile field. Under the combination of both the radial and linear fields
thore remains still some symmetry, since reflection with respect to a plane
perpendicu1arto the direction of the linear field leaves the system. unchanged
58

Ch. 3, 1, 21

EXAMPLES

S9

A magnetic ield; however, has the symmetry properties ofa ~'.nd.~heJlce

will remove also the m :a 1 degeneracy.
If we deal.with diatomic mo~es, the symmotry of the lleW{cylindrical
or coDical) permits rotatioJlS~&round the molecular axis aa4 refiectiODS about
the symmetry planes. This leads ~o another type of degeneracy which will be
discussed in Chapter 4, ~ 4. In general, in the case of more complicated
molecules, in order to have degeneracies it is sufficient that the molecule
presents certain symmetries. These symmetries are comparable. with the
groups in crystallography_
Finally and particularly the indistinguishability of partidet of the same
kind, as for eumple, e1ectrODl arollnd thesamo nucleus. nwaDi that ~
. . H m.
. .' . 1....-..
-
uad..~ a ............
hA. bet'ViQi)h .,..,...:.L.e
Hamil
. :taman
~ iavadant.,
~~...~",Il
iI"'R~
of the same kiBd. This is the origin of a new degeumcy:
classiea1 theory: theexcbange degeneracy as was poiated.()utmtht.,papers
of Heisenberg on the helium m.olecule.
To each of these types of degeneracy corresponds a group oftransformations. The study ~each of these groups containing rotations, rei1oxioD8,
and permutations allows one to determine separately all the properties of
the wave functions which are of kinematio origin and pexmits' one to solve
the dynamical problem completely. In general thiS is very<1iftlcu1t.
Fina1.1Y quantum: -theory&diaea physical quantities as operators amfthele
belona p,.se~to c:ertam ~9'PP8. (compare Chapter 2; 3.2).Thae
definitidns, which are simple in" one-electron . , . . . . -became _cn;~ttt
cated iIi the case of complex sysiems particularly when. the spin. lias t(J be
taken ib.to account. It is obvious that group theory is the most,_d' and
essentially the-simplest method to define all the physical quantities in complicated cases.

_onsaeeable.ftom

1. Examples
2.1. GENERAL CONSIDBRATIONS

(i) Let aCOllecUQIl of n objects be placed in a ..tWa o.~ by

the indices 1, 2, . II. If this order is upset, reaultiaa in,for,iaatance. ,3, 7,.6,
... k, we haye performed a permutation. We indicate this operation by the
symbol

1 2 3 .. ")

PI

== ( 3 7 6 .. k

In order to specify the sense of the preceding symbol, we assume that

60

GROUP THEORY

[Ch. 3, 2

the indices are attached to the objects as labels: Pl is an exchanae of the

objects 1 and 3, 2 and 7, etc. 1
To let the collection of objects undergo two successive permutations,
P,2 =: (: l ; :::), amounts obviously to the effect of a single permutation
S

1 2 3 ... )
4 9 . ... ..

= P2 P 1 = ( 5

We say that s is the product of the two permutations PI and P2' using
the convention that the first operation will be written at the right-hand side.
This will permit us (if necessary) to place a symbol representing the collection
of objects on which they operate at the rightl. This rule is general. that is
we will use it for arbitrary operators acting on different mathematical
objects: vettors, functions, etc.
Among the permutations defined in this way, there is one that does not
change the place of the objects. This is the identity, B, which obviously satisfies the relatioDs:
EP

PE

= P;

_(11 22 ....
.... n)

E-

Permutations do not commute: PI P2 ::f: P,2Pl this can easily be verified.

However, to every permutation P = (~ ~::::) corresponds an inverse
,~~. == (~~:::~), which puts the objects back into their place and which
I
coqseq\lel1t1y satisfies , -lp = PP -1 == E.
~ ~t of nJ peq1lptations of n objects form a group, the sy1'1J'lMtrlc group
4Jf.J)J;der n!, which we indicate with the symbol 9',..
/' '(li) If we let n go to infinity and if we distribute our objects on a straight
~~, on a plane, or in a space of 3 or more dimensions and if we go to the
limi~ so that now the redistributions are continuous, then there will correspond
I

..

a point to each object and a continuously variable index m to each point.

In 'three dimensions all these points form the set specified by the three
coordinates x, y and z.
Ifwe';Dow permute our objects, it will give rise to a well-determined correspondence between an arbitrary point P of the space under consideration
ItiIYalso possible to number the boxes in which the objects are placed. This makes a
second type of permutation possible, i.e. the exchange of numbers between the boxes.
The same formulas apply to this case. Quantum theory considers both types of permutations simultaneously (camp. DIRAC, section 6S, first edition only).
lOne can make the opposite convention. The one we adopt is the most convenient with
regard to applications in physics.
.
1

Ch. 3, 2]

61

EXAMPLES

and its image point pi, thus we have effected a transformation A of the
onto itself. We will express this by the notation
P -+ pi

== AP.

The identity is defined by BP = P.

From two successive transformations P -+ pi = AP, pi
a unique transformation,

p --+ P"

s~ce

== cP

-.+,

P"

= BP' results

= BAP,

their product, which can be abbrevi~ted to c = BA.

We will occupy ourselves only with reversible transformations, that is one-toone correapondences. Hence we assume that after a transformation A it is
always possible to bring the space back into its original order by a welldefined transformation A-I. The transformation A -1 is the inverse of A and
we have,
P' .... P = A-1p', where A-1A = E.
This is Qbviously only possible if the correspondence P !; pi established by
the tr8J).sfonnation A is one-to-one.
(iii) Special cases of spatial transformations are: translations and/or
rotations around an axis or a point (these operations are established in the
concept of a crystalline solid), the motions of an arbitrary fluid (see BIRKHOPF 1[1950]), images formed in optics, etc.
Un~ar and reversible transformations of affine or unitary spaces are the
most ;important groups.
T~e permutations themselves are transformations of a discontinuous
space consisting of n points.
'/

2.2. GROUP POSTULATES

From these examples we can deduce by abstraction the general definition

of a group:
Let "there be a collection ~ of operators or elements: A, B, fuiite or
infinite in number; they will form a group if the followingfotlr conditions
are fulfilled:
I. The product BA of two arbitrary elements is an element of the collection f,
"
(3.1)
BA = C, c belongs to <D.
n. The identity E is a part of the collection <6 and satisfies

EA

==

AB

==

A.

62

m.

To each element A corresponds another operation

inverse of A, belonging to tI 8J)d defined' by:
A -lA

IV.

[Ch. 3, 2

GROUP THEORY

==

AA -1

A -1,

called the

B.

(3.2)

The product of these operations satisfies the associative law,

(AB)c == A(BC).

(3.3)

gen~

we have AS rF BA. A group in which the operations are

commutative is ca1fed Abelian. If the number 9 of operatio1l! belonging to t6
is finite, the group is finite and 9 is its order.
But in

2.3. FUR.TQIl EXAMPLES OF GROUPS

1. The collection or set of rational numbers form. a group .under the

operatiOll of multipUcation:
L 'I11e product of two rational numbers is a rational number,
u. the identity is represented by the number 1,
m. the inverse of (A/B) is (B/A) := (A/B)-I.
This group is Abelian and of infinite order.
from t~ PDint 91111ew 01 mJIItiplication the collection of integers (positive
anflf neptive)
not form a group.
~. However, they form a group if we consider the operation II addidon.
Th~ o~ation resultinS from the combination of two elements A and B
is then written A + B instead of AB
.' .L The sum. of two inteprs is an integer: A + B = C,
n. the identity is represented by the symbol zero,

do

A+O == O+A ==

A,

m.

the inverse of A is - A.
This group is Abelian. It is an additive group of infinite order.

3. -Let us 80 back to the~. space of three or n dimensions. A vector "

represents an. operation, a displacement for example.
I. The sum of two vectors is a vector
== c,
n. -. +0 == 41, the operation B is the null vector,
m. the inverse of G is -II.
Hence the set of displacements or vectors form an additive Abelian group
since .+b -= 11+.
.
The displacements are a special ease of the projections of the space on

+"

63

EXAMPLES

Ch. 3, 2]

itself, one of the few in which tPe operations do commute. Affine geometry,
i.e., the me,tric used in vector spaces, is the study of the group just mentioned
't!ith one 'additional property: each operator or vector II may be multiplied
with a real number if we deal with a real space~ or with a complex number jf
we deal with a complex or unitary space. This is called "" additive group
with multiplicators. The mapping of a vector space upon itself results from
the combination of two processes, the process of addition and the process of
multiplication by numbers. These numbers are operators acting on the
vecton. These vectors are themselves elements of an additive group. The
operators could be of a more general form than simple multiplicators;
for instance matrices. Hence a vector-space, on which a system of matrices
acts can be considered as an additive group submitted to a 8Ystem ojoperfltors.
The extension of the theorems for the theory of groups to this particular
case is of tremendous value.
2.4. GROUP TABLE

In summary, a
group consists of a set of symbols satisfying the post1i(ate8
--"'u
I and IV for w1Jich the "multiplICation rules are given a priori. These symbols.
,-

represent operations which are not specified by the abstract .grouP. theory.
An abstract aroup outlines the properties of a certaitl numbe( of concrete
groups~ which are realizatio1J3 of the abstract group, in a kind of Pythagorean ta~Ie (CAYLEY [1854]), usually called group table. For example: Each
otthe t~o Tables 3.1 and 3.2 represents an abstract group whose properties
are those of the group .9'3 of permutations of three objects and those of the
groul>~ of operations which brin, the equilateral triangle into coincidence
witb1tself including the possibility of turning it over on its face. We can give
the following significance to the symbols in the table: A is a cyclic permutation
(~ ~ ~) or a rotation of in ~ound an axis normal to the center of the triangle;
B := ~2 is the inve~~ of the previous permutation or rotation of
around
the same axes (~ : ;); c, D, F thC transposition of two objects or rotatioJ1S
around the medians of the triangle [Compare Fig. 3.3]. '
Table 3.1 looks like an ordiDary multiplication table. In Table 3.2 the
sequence of operations in the first column is not in the order E, A, B, C, D, F,
butintho-orderoftheirinversesE,A- 1 = B,B- 1 == A, c- 1 = C,D- 1 == D,
J

tn

F-1
J

==

p.1

At tho intenection of tho row c and the column

obtained by e1fectinl first c and then

A.

we find the operation

= AC

64

[eb. 3, 2, 3

GR.OUP THEORY
TABLE

3.1

TABLE

EABCD-F

EABCDF
E

A
B

C
D
P

EABCDP
ABBDPC
BBAPCD

CPDBBA
DCPABB
PDCBAB

3.2

EABeDJI

A-I = B
B-1
A

DEAFeD
ABBDF'C
CFDBBA
DCPABB

c-

D-1
F-1

==
=

::=

C
D
F

PDCBAE

3. Subgroups
3.1. DEP1NITION

In the poups of space rotations around a center 0 those operations which

arc rotations around an axis Oz form a subgroup, 'because they satisfy the
postulates I to IV. The identity E is a null-rotation around Oz as well as
around any other axis. One can also form: a subgroup in the group .93
which we just examined. The operations B, A and B form a subgroup .Te,
the subgroup of the rotations around an axis normal to the triangle. This
SUBII'8UP forms a cyclic group because it can be generated by powers of a
single operation A (8 == A 2 , B = A 3 ).
From these considerations we form the following general definition:
I

lfwe eanfind In a group ~ a set of operations .Te,jinJte or infinite, such that:

1. '(he product of two of these operations belongs to :tl',
2. :JfI contains the elements A, B, it also contains their inverses
A-I, Ii-l, ... and as a result o/this the element E; then;F is a subgroup ofrl.
We have defined at the same time the cyclic group generated by iterating a
untq.iIe operation A, i.e., formed by the "powers" of i.. In order that a'cyclic
gr~up be finite, it is necessary that a power p of A reproduces the identity:
Il"= All where p is the order of the group.
In a vector space of n dimensions let us put aside all the vectors of a subspace 'with m < n dimensions. These constitute a subgroup of the main
additive group. In this way a plane 112 going through the origin constitutes
a subgroup of the ordinary space 113 and the vectors lying ona straight line
8 1 in this plane form a subgroup or this subgroup_ The line must go through
the origin since a 'subgroup must contain the unit element, i.e. the null-vector.
We" want to study the influence of operators on the subgroup. Ifwe subject
the space to a system of projection matrices, these operators will, in general,
bring the vectors belonging to 52 out of this plane, i.e., the subgroup 112
does not allow the same operators as the main group
If the operators

It

/I

ut3.

Ch. 3, 3]

SUBGROUPS

65

are multiplica.tors it is always true that the projected elements of the

subgroup belong to the subgroup. An interesting case is the case of the
additive subgroups, which allow the same operators as the main group_
For example the operation could consist of adding constant vector lying
in the plane 8l2 Thus these additive subgroups are subspaces which are
invarlont with respect to the operators considered. Hence it is natural to adapt
the axes to this division of 1l3; if 112 happens to be an invariant plane, We
designate it as the xOy plane and take oz normal to it. We will come back
to this point in 7.
3.2. COSETS OR COMPLEXES ASSOCIATED WITH A SUBGROUP

Let l be a group of order g, ;e one of the subgroups of order h.Botb 9

and h may be infinite, but we will first consider the case of finite groups.
Let E, A, B, c, .... F be the elements of ;e.
10. Multiply.n these elements at the left side by the same element D of .Tt'.
As a reslllt of the definition of the subgroups all products thus obtained:
DE, DA, DB, are again belonging to ;e. On the other hand the,- are all
different. (J)A will not be equal to DB if A " B). Hence this multiplication
amounts to writing all elements of ;e in a different order. If a subgroup I ,
or an arbitrary group, undergoes a transformation from the left by multiplication by one of its elements this means a permutation of its elements.
2. We take now an element 82 of < that does not belong to Jt' and form the
prodUcts S2A, S2B .... S2 P This collection constitutes a complex of elements
or co#et, all different from each other. We will call this a left coset or a
complex associated at the left to the subgroup Jt' and indicate it by 8 2 3f'.
No one of these elements belongs to Jt' because if S2A would belong to
;e, then S2AA-1 = 82 would also belong to ;e according to the definition
of a subgroup.
These elements do not form a subgroup: the product S2AS2D is certainly not
a part of the subgroup since we know that AS2 B does not belong to Jt'.
If S3 is an element of t:6 and is neither a part of ;e nor of S2.;tl', then we can
form with 83 a second coset S3.;tl' and we can show in the same way that it
does not contain an element of ;e nor of 8 2 3f'.
If t is a finite, group we can exhaust this procedure. As a result we have
the equation:
(3.4)
Each of these cosets contains the same number 'h of different elements

66

OROUP THEORY

[Ch. 3. 3" 4

as .1t', hence we have:

9

= Ih.

(3.5)

The order of a subgroup of a finite group is an integral dlrJuor 01111. order

of the group. The integer I is the index of the subgroup.
One defines in the same way right cosets to a subgroup.1f' and these will, in
general, not coincide with the left cosets. We have
l

==

.1t' +.Tt'S2 +.Tt'S3 +

... J't's,.

. (3.4a)

The nature of the cosets does not depend on the choice of the generating
elements ~, S3 They are uniquely determined by the structure of "
and rt>. To sho~ this proposition let us co~der an arbitrary element
82 of 828. We claim. that .S2.1t' :; S2J't'. As a mauer of fact s; can be
written as t2D uit bcJonp to. S21, where D is a ~~ elemeD~ ~r Jr.
Hence sil consists of the collection of elements.f2DA, SJDBt S2DP, i.e.
those of S21 written in a different order (compue 1).
In the examPlo of 2.4 the subgroup .ttI3 (A, B, B) possesses a single coset
~3 =- Dds
P~3 .
The ume eonsideratioDs
. apply for infinite groups. The index I of I could
then be finite or infinite.
Let St, for example, be the set of vectors in a three-dimensional space,
considored as an additive group, I the Bet of vectors _ of the xOyplane
andr~' a, 'lCdornot located in 4f'. Then ,4f' is the set of all the vectors , + "
taU.the points S' are in a plane Z parallel to xOy. Each of these
planescetfllPond to, a coset associated with 4f'. ~ere are a continuous
bJIlaite number of these planes (compare Fig. 3.1). (Sec p. 69.)

-'

-~

fro. ..
i

4. Cf)Djagated EleDlents. Classes

4.1. THE CASE OF LINEAR SUBSTITUTIONS

Let

Us consider in an ,,-dimensional vector space, the grOUP (of infinite

order) of.all the linear veCtor transformations with complex coeflicien+l.

One of -them will be
(1.4)
or in -matrix language

Y:r::AX.

(1.4b)

s,'

: We can consider this from two points of view: either the axes stay' fixed
and we let a transformed vector y = Ax correspond to the vector x (A

Ch. 3, 4]

CONJUGATED ELEMENTS. CLASSES

67

represents a mapping of the space upon itself), or the vectors % and y are
identical in space and the axes are rotated. If (1.4) is considered as a change
in basis vectors we find according to (1.3a) that the basis vectors underwent
the transformation A -1.
Let us take now two transformations A and S and

(3.6)
then by using the first interpretation, i.e. the axes e, are kept fixed, we ~y
that the matrix 8 produces a mapping of the space upon itself; which in
general differs from A: we say that the two transformatio1l3 A and 8 are

conjugated.
With the similarity transformation (1.7), or the second interpretation
used for the Illatrix S, we know that they go over into each other by a change
of coordinates S. Hence two conjugated linear transformations are in general
different but equivalent in the sense that either one can be obtained from the
other by a transformation of axes.
4.2. GENERALIZATIONS. INVARlANT SUBGROUPS

Two elements A and B of a group ~ are conjugated if one can find a third
element s of the same group, such that
(3.6a)

A c~ is a (JOll_on of aU operations conjugate to a given operation A:

we ob~ ~e class from this ~lement by considering s in (3.6a) as represent4lS',:4R1ccessively all the elements of the group tI.
In:' an Abelian group SA

==

AS

or SAS- 1

= A holds for every s. Hencoeach

element forms a class in itself.

In each group the element B forms a class.
Let .If' be a subgroup of ~. Then s;'l's -1 is again a subaroup (s is a futed
element). the conjugated subgroup of I . Indeed, if AD == C and aU three
belong to I, then SAS- 1 SBS- 1. = SABS- 1 == scs- 1
Hence a subgroup .JIr has as many conjugates as thefe are operatiOns s'ia
the group; but not all of these are distinct from each other.
It often happens that 3l' is identical with all its conjugates, that is to say
that SFS- 1 belongs to 3l', whatever the element F of ;ff' is, for every s
belonging to <. In this case I is an invariant subgroup or normal diviSor.
For example the subgroup d 3 of the group $P1 (lee 3.2) is invariant.

68

GROUP THEORY

[Ch. 3, 4

4.3. FAcrOR GROUP

If ~ is an invariant subgroup in f#, let us decompose rI into .Yl' and its

cosets, 1

(3.4)
We claim. that each term of this sum may be considered in its turn as an
element of a new subgroup, called the factor group of 9J and designated by
the symbol t/.Yl'.
Indeed if we take the product Si .Yl'Sj:/l' we obtain a coset containing all of
the elements of t which are of the form 8, ASJD, A and B belonging to:/l'.
We claim. that this product: is one of the cosets associated with :/I' in (3.4),
that it is the one which contains the element St =:II S,81 and which we designate

by s".Yl'.
Indeed
S,ASJB

==

-1

s'SJSJ ASJD,

and as .Tt' is invariant (a neceSsary and sufficient condition) SilAS) =

which is an element of :/t' as well as DB = c. Hence we have
SiASJB

= SiSJC ==

S"C ..

If we take for A and B all other elements in .Yl', keeping s" fixed, c will vary
but remain in :/t' and hence we will have
(3.7)
i

we

'1

'2 =

82 :/1', .... ' , == S,:/I', the symbols '1'

'2;,. :F, may be considered as the elements of a new group, the factor
group t/.;te, using this notation (3.7) takes the form of the usual definition
of a group product fF,fF} == :Fie.
Evidently we have :/1':/1' = :/1', which simply results from the fact that the
product of two elements of ;It' is also an element of this subgroup. Hence
.Yl' is the unit element of ~/:/1'.
The definition of a factor group amounts to blurring out the individuality
of the elements of :/1'. They blend into a single element as is also the case for
all the elements of each coset associated with :/1'.
All preceding propositions can be extended to infinite groups. It could
happen that the factor groups are infinite.

If

put

=:/1',

The left and right cosets are the Same in tho case of an invariant

follows immediately from the definitions.

subgroup~,

as

Ch. 3. 4)

CONJUGATED ELEMENTSCLASSES

69

4.4. ABBUAN GROUPS

then SAS- 1 == A for every s. Hence each subgroup is invariant

and can be used to define a factor group.
Let us take for example a three-dimensional vector space 113 The vectors
in the plane -2 == xOy form an invariant subgroup ;l' (SAS- 1 :a: S+A-S -= A).
As we have seen in 4.2, a coset associated with ;l' consists of the set
of vectors which have their points in a plane Z perpendicular to Oz the
z-axis (see Fig. 3.1); to each of these planes, i.e. to each value of z, corresponds
a coset. They form a continuous series.

If AS ==

SA

~~~~~~~~~~~~--~~~7

,/
,/

/'
/
/'
/

2//

/.
/'

" 'st

/' /

'"

;'

Pig. 3.1. Coset associated with .11'

== xay.

The e~ent F. of the factor group can be obtained by disregarding the

differenCes between the different vectors 11, fl, ..," .... which have their points
in the plane Z~~ that is by projecting them all on the axis Oz. This is expressed
in mathematics by the symbol of congruency. (For this notation see also
problem 3.11):
1J == 11' == ,," (mod \12)
The notation means that v and v' are identical if one disregards the (vector)

difference lying in the plane 1l2 In the same way we treat angles intrigonometry where 45 and 405 are considered identical disregarding the difference
of 360. The latter is irrelevant in certain cases as for instance the calculation
of the trigonometric functions.
The figure accompanying the previous formula is identical with (3.1).
The plane Z is the space Bt2 ; all vectors v, 1", 1''' etc. are vectors between the
origin and a certain- point in. Z.

70

GROUP THEORY

[eh. 3, 4, S

To each value of z corresponds an element Fz of the factor group '*11 =

8131-2. This group is continuous and represented by a one..dimensional
vector space, the axis Oz. More generally (the proof is the same) if < is an
n-dimensional vector space, has m dimensions (m < n); each element of
!F, of f/I consists of a set of vectors Vi' .;, v~' ... that satisfy,

v ==
~

v~,

==

1)~'
(mod
J

Yt')

any <IYt' is a vector space ,vith n - m dimensions obtained by projection on

an axis, a plane or a hyperplane.

s.

Some Properties of the Group of Permutations

(Symmetric group)

or n

Objects :7,.

5.1. NOTATION WITH CYCLES

If p = (~ ; ~ 1~) is a cyclical permutation, we could write behind each

term the one we have to substitute for it . We obtain the following notation:
p

= (1 2 3 4 5).

In a similar way we \\rrite

'1 ., 3 4 ~\
:...
~) = ~ 1 ~ 3X2 4)
( 54123
{:.
,

the permutation can be separated into two cycles, each set of elements forms
an independent cycle that tenninates in itself. An object that does Dot move
fro~ its place will form a cycle on itself. In this notation one encloses it
between brackets or sometimes one omits it altogether. The order of a cycle
is tlie number of objects contained in it .
,Only three things are important in this notation: the number of cycles,
the objects contained in each of them, and the succession or order in which
they appear in the cycle. One can arbitrarily choose the order of cycles:
(1 5 3) (2 4) = (2 4) (1 5 3) or the object that appears first in a cycle (1 5 3) :;::
(5 3 1) = (3 1 5).
A transposition is a permutation of two objects. The following permutation
is equivalent to a single transposition:
f

G~ ~ ~~) = (2 4)(1)(3)(5)

(24).

Each permutation is obviously a product of transpositions:

(1 5 3)(24) = (1 5)(5 3)(2 4)

or

(2 4)(5 3)(1 3)

Ch. 3, 5]

71

GROUP OF PERMUTATIONS

but in the cycle notation we usually arrange the objects such that the diffetem
cycles do not have an object in common. A permutation is even, or odd nit is
the result of an even or odd number of transpositions. This is called the
parity of a given permlltation.
One can show - and it is almost obvious - that the parity of a pennu~..
tion is unambiguous, i.e. independent of the manner in which one decomposes it into transpositions.
The inverse permutations of a given permutation can be obtained, in the
cycle notation, by reversing the order of the terms in each bracket. The
order of the brackets as a whole is irrelevant since they are independent.
(1 5 3)(24)-1 = (3 5 1)(42)

= (1 3 5)(24).

5.2. CONJUGATED PERMUTATIONS

If we have

n)

s = ( 1. .2 ... . ;
J1 12

'II

1 2 ~ .. n )
( k 1 k2 . kll =

(i11 i212 .
ill)
. I,.. ;
1

then,the conjugate of s with respect to Tis s' = TST- 1 , where we have to read
the operations from the right to the l~ft, so we obtain
I
- 1
S=TST

k
z
ka)
.
11 11, " I,.

\' k 1

Hence ,to obtain the conjugate s' of s",ith respect to

we have to carry
out the permutation T on both lines of s. This rule can be transferred ..uy
to the cycle notation: suppose the s requires several cycles~ It is then obwOJlSly sufficient to carry out the permutation" Ton the terms. ofeach,cyclc.:in
order to obtain TST -1, i.e. we neither move the parenthesis nor chan.~the
number of cycles, nor the order in which they appear. Take for example,
s = (1 5 3)(24), T = (123 4 5)," then, formingthe conjugate, wohaveTST-. 1 ==
(2 1 4)(3 5)~ In brief, to go from a permutation to its conjugate mtaDl:to
change the labels of the objects which one permutes without mOttifyiDgtb.e
manner in which they are permuted.
From this we see that a claSs ofpermutations (compare ,4~2lis.complllely
determined by the number of cycles and the order of each of them.
T,

72

GROUP THEORY

[eb. 3, " 6

5.3. ALTERNATING GROUP .sif4 OF n VARlABLES

This is the group formed by all the even :Permutations of " objects, a
subgroup of index 2 of /,.. Its coset is a set of odd permutations;
these permutations do Dot form a subgroup since the product of two odd
permutations is oven. d"js an invariant group of !/,., according to the rule
of 5.2.
Most of the preceding results were obtained by Cauchy.

6.

isomorphism ad

HoJllOlllOl.Pldsm

6.1. DEPINITION

If we have two groups !J and (1/' such that:

1. To each element A of t corresponds one and only one element A'
of <6' and vice versa.
2. If AD == C then AI B' == c' for every element in the group.
The group tables of these two groups are the same and actually they differ
only in the designation of the elements. In this case the two groups are called
isomorphic (or have a holohedral isomorphism). From the point of view of
abstract group theory the groups are identical. However, the elements refer
to different mathematical objects. We have seen an example in 2.4; the group
9'3 and the group of opetations which describe the coincidence of an equi..
lateral triangle with itself are identical.
6.2.; GENERAL THEOREMS

.t

,Let us suppose that the correspondence between tJI and fI' is not one-toone,
means that to an element A' of ti' there will correspond several
dUlerent elements At, A2' A, of 16. Any element A of <6, for example A"
multiplied by one of the elements of (I corresponding to B' in '6' should
lead to an element Ci which corresponds to A'B' --= c'. (see Fig. 3.2).- In
this case, the gI'-Oups are called homomorphic, that is the isomerism is
merohedral
These concepts are used in crystallography and 'one can find a number of
examples there. A simple arithmetic example is the following. Let V,. be
the grOUP of permutations of n objects. With the even permutations we
ass~ociate .the number + 1 and to the odd permutation the number -1.
The multiplicative group tN' consists of two elements + 1 and -1 and is
homomorphic to f/". The element + 1 corresponds to the alternating sub-

group and - 1 to its coset.

73

ISOMORPHISM AND HOMOMORPHISt.{

Ch. 3, 6]

In senera! if a group t6 possesses an invariant subsroup 3t', then the group

~

is homomorphic to the factor group t6/~. All the elements of ~ correspond to the unit element P1 of t6/:JIt' and those of the complex s.~ to F,.
The converse of this obvious statement is also true and forms a fundamental
theorem.

.'.'.c'

At

,.

"at .

l'

--_..... ........
..,

At

/1\\

belo", to

//I~

;/,~
All

_, 8j .........'"

C:,

.. ........ --....
A/ .) -

c;.

e..... e,l

---

betongto1

Fig. 3.2. Diagrammatical representation of the homomorphism between two

sro~pa.

TheOr.m: lft is homomorphic to ~' and letting ,H' be the unit element o/tl

l
,

then:

1. The seto! elements of t that correspond to E' forms an invariant subgroup

of i;

2. t6' is isomorphic to the lactor group (I/3t'.

The proof is as follows. 1. If the elements 11 and 12 of t correspond to H'
thenl, = J,l:z,correspondstos',B' == B/. We conclude thal the set of elements
J112 1,.. 9f '6 which corre~pond to B' forms a subgroup lof 'I. This
subgroup is invariant because using an arbitrary element x of f6, then xnc- 1
corresponds in <6' to
X'E'X,-l

==

X'X,-l

= E'

and we see that XIX -1 belongs ,to ~ for an arbitrary x.

2. If 81 and 82 are two elements of <, not belonging to rt' and corresponding to the same element s' in <6', then s i 1 S2 corresponds to S' -1 s' = E' and

74

GROUP THEORY

[Ch. 3, 6

belongs to ~: s1182 = 1",. Hence 82 = 811m lies in the coset Sl ~ associated

with ~.
Inversely two elements S11", and S11,. of this coset correspond to the same
element s' in f'. Hence the element s' corresponds to the complex 8 1 ;/1','
just as E' corresponds to ~. We find that C' is isomorphic to t/~.
6.3. REPRESENTATIONS OF A GROUP

A special example of isomorphisms are the so-called representations of an

abstract group. Let us suppose that CD' is a group of linear transformations.
To each operation A of t corresponds a matrix A created by the linear substitutions of a set of variables called the basis. We say that ~' is a representation of~. If the connection is isomorphic the representation is faithful. 1
For example, ,if every group element is represented by the number 1, the
multiplication rules are preserved but the relation is not isomorphic, hence
not faithful. One is unable to recognize the structure of the group table from
this representation. We will designate representations of groups by b01d face
capitals 'I or also by $l(G) or !1J(o). In particular 9) stands for "Darstellung" . 2
The order of the representation matrices is equal to the dimensionality
of the representation space 9t, i.e. number of variables employed in the substitution.
For !instance each abstract group has an obvious representation, the
identity' representation in which we let correspond to each element of the
group the substitution Y = X i.e. a one-dimensional matrix. To the product
AB = "c correspond~ in this representation the product of the matrices
1 1/ 1.
6.4. EQUIVALENT REPRESENTATIONS

If ~ is a group with elements E, A, B, C , and if t 1 and t2 are two of its

representations consisting of matrices E 1 , AI' B 1 , C 1 , and E 2J A 2 ) B'2'
C'2 .. t~en-we will call '61 and ~ 2 equivalent if we can convert one in~<? the
other by a change of coordinates, i.e. if we can find a matrix S such that
A2 = S-lA 1 S; B2 = 5- 1 B 1 S; ... shortly l2 = S-l~lS.
Obviously~.the two representation spaces Btl and 8t2 are identical.
Also used: Utrue", which is an incorrect translation from the German "treu" meaning
faithfuL
I The reader will have noticed that we used mainly lower case bold face letters for
vectors in order to avoid confusion.
1

REDUCIBILITY OF

Ch. 3, 7]

REPRESEJ:"-rI'!\T!O\I~S

75

7 Reducibility of RepresentatiG113
7~1.

INVARIANT SUBSPACE

If we have a representation 9J in an n-dimensional vector space, it often

happens that there exists a vectorial subspace n11 \-vhich has only m < n
dimensions and which is invariant with respect to the transfortnations of (I,
i.e. such that all mappings .Ii of t transform the vectors of m1 into vectors
of 1 It is natural to adapt the coordinate system to this invariance" We span
the space ftll by the first set of m unit vectors and we leave the remaining
n-m outside this space. A vector oX of 91 1 will have in this case the COfllpOnents X m +1 = X m +2 = ... = Xn = o. Since after the transformatiol1
x' == A x remains in Ul 1 , the preceding equations lead to: X~+l = X;"+2 =
... = x~ = o. The mapping x' = Ax can be v"rritten

X m +l =

X'II

Hence the matrices A of fI have the form

A= (Poa SaQa)

(3.8)

where Fa and Sa are square matrices l of the order m and (n-m), Qa is a

rectangular 11latrix, and 0 a null matrix. Hence, the matrices Po, Pb
generate a representation of f in the space Btl with m dimensions.
In case we can, find such a space Btl and its corresponding axes, the
representation '6 is reducible. If we cannot find such a space the representation
is irreducible. These definitions are independent of the concept of groups
and of representations because they apply to an arbitrary set of matrices.
The matrices Sa, Sb .... also form a representation of the group t9 in a
space which one obtains by ignoring the components Xl~ X2 X m ,
x~, x~ . _x;" of the vectors in Dl. This space is geometrically obtained by
The word matrix is used here in the loose sense of "array of numbers" not in the
usual strict definition of "array of numbers that transform like a direct product of vector
components'" since the latter obviously does not hold for subsections of a matrix.
1

76

GROUP THEORY

[eb. 3, 7

projecting the vectors of tJl parallel to il 1 , i.e. by projecting all those that
differ only by their nl first components into one single vector. We will designate these by the symbol 0l/Di 1
.
Indeed, if we consider the vector space at as an additive group with operators defined by the matrices of the system '#, Uti is an invariant subgroup
\vith the same operators (invariant in the double sense of the word with
respect to tlle transformations of t and also because the group is Abelian).
miDlt is the factor space as defined in 4~4 of this Chapter. I-Iencef the
matrices Sa form a representation of (I itt the factor subspace 8l/Dl 1 7.2. COMPLETE REDUCTION OR DECOMPOSITION

The most interesting cases are those in whlch one succeeds, by a. conven. .
ient choice of axes, to decompose the space
into two independent in~
variant subspaces Ul1 and tIl 2 All the rectangular matrices Q. become zero
il~nd the matrices of the systenl 9J take on the fonn of So step""~JI~fe matrLy~

Po
(o

0\
~'SI1) ~

'Inus the representation, or more generally the system of matrices II decomposes into two representations or into two separate matrix systems, one with
m, ihe qther wit.h It - m dimensions.. Tras is called a complete reduction or
decom1'1d,Sition . This is indicated by a S},mbolic plus sign:
i

'11+~2'

(3.9)

9l = fRl + tllz

(3 ..9a)

!f

e.s

hence Ut1 is isomorphic with Ut/Dt1 Thtoreffi! Every systllm of reducible Ul'iitar,v matrices is decompoaabky, i.e.
for a unitary system we can always go from (3.8) to (3.8a) by a suitable
ehanae of coordinates..
It is sutB.oient to take in the unitary space It - in which the tranaformations of the system tI are performed - a set of axes such that the first m
span the space 1 and the remaining n - m span a space 812 that is orthogonal
to til' i~Oo--~a space that contains all the vectors "perpendicular" to at t Since -the transformations !I are unitary they conserve the orthogonality
relations among vectors. They transform. III into Jtself and will do the same
to ~ as this space is orthogonal to Ill- The result is that the second one is

in,variant too.

Ch. 3, 71

7'7f

REDUCIBILITY OF REPRESBNT..J\TIONS

7. 3. REDUCTION OP THB UNITARY MATRICES OF A GROUP INTO THEIR

IRREDUCIBLE PARTS

An isolated unitary matrix can always be brought into the diagonal form.
For a system of unitary matrices this is only possible if these matrices
commute with each other (compare 3.3). rHe assume of course that all tllC
matrices of the system are transformed with the help of the stIrne transformation matrix.) If the matrices are not commuting, we can only hope for
the simultaneous reduction of aU the matrices of the system to Hboxes n
of the size ml, m2, .... m,,(ml +ml + .... mil == n) along the diagonaL If this
reduction is carried out as far as possible we say that the system of matrices
is reduced and the parts, each corresponding to one set of boxes (one for
each matrix element of tf), are called irreducible .
""hen we qeai with the representation ~ of an abstract group ~~ it
important to decompose this representation into irre,/u[:ible representatioru;
since, as we will see below, only these have a fundarnental ll1eaning in physies
and mathematics. Once we have found these ].1 irreducible; constituents ~ 1 ;
t#2 . ~ . t6 p we can again use the notation introduced ir: equation (3~9)"
~ = tfl+~2+.

4.

J)~

None of these preceding definitions implies that the group Cd is

7.4~

tlnite~

EXAMPLE

We will again use the group [/3 since this is the simplest example of a
non-=Abelian group. The elements are the six permutations (1), (1 2), (2 3),
(3 1), (1 23) and (1 3 2). In order to make use of the group table in 2~4,
Chapter 3 we indicate the isomorphism: (1 23) == A; (3 21) == B; (I 2) = n;
(2 3) 5: C and (1 3) := :F.
It is easy to find a representation of this group., Let us take for tbeobjects
which permuted the three variables Xl, Xl and X3, i.e. the three projections
of a vector % on three rectangular axes -1, 62 and ',. The different elements
of this group are represented by the system of mapping equations:

(1 2)

-+

Yt ==
Y2 =

13

X2

Xl

e: X3

(1 2 3)

-+

Yl

==

Xl

Y2

=:

x]

Y3 ==

Xl

etc~

We obtain in this way one 3-dimensional representation of V 3 from which

the matrices can be written as:

78

tCh. 3, 7

GROUP THEORY

010\
(1 2) ~ { 100)
\001/

010\
(1 2 3)

--+

( 0 0 1)
\100,

etc.

The following simple geometrical remarks permit us to reduce these matrices.

The permutations of [/';:. leave the sum Xl + X2 + X3 lllvariant. Hence the
planeXt+x2+X3 = o (plane9i2)andthenormaltothisplaneXt = Xl =X3
are two invariant orthogonal subspaces of the representation space Bt3 Let
us take an axis along the normal and two axes
and in the plane \ll2 .

e;

e;

e;

Fig. 3.3. An object that allows tho six operations of the group .9',- Indicated are the threefold axis (perpendicular to the plane) and the three two-fold axes (in the plane).

It iseonvenient, in the sense qf obtaining more symmetrical fonnuias, to take

these last two axes neither Perpendicular to each other nor of unit length.
"~le will take el in the plane of e 1 and e2 and e; in the plane of e2 and e3.
't."'.\1 ~
V7

t.f .t~ke.
.<~t
"

(e; is directed along the bisectrix of el and -'2' etc.)

lne corresponding change in variables is found by taking the transposed

Ch. 3,

71

REDIJCIBILITY OF REPRESENTATIONS

79

inverse of the transformation of the basis vectors (compare equations

(1.2) and (1.3):

1\

s=!(~ -~ ~)~s-l=I\;_~ -1) .

o 1 -1
1 1 -2
Hence the new variables are

Fig. 3.4. l'he choice of unit vectors that establishes a decomposition of the representation
into two irreducible representations.
..

From which we find the matrices

The reduction cannot be performed any further . There is no invariant

axis in the plane m2

GROUP THEORY

80

3~

[Ch.

Hence we have found a three-dimensional representation of the group 9'3

and we reduced it to two irreducible representations. One is the identity
representation, consisting of the matrices 1, and the other a two..dimensional
representation in the plane fR 2
The group f/ 3 has a third irreducible representation. It is one-dimensional
and can be found very easily. The alternating group .RIa is an invariant subgroup of ~3. It consists of the even permutations: (1), (123) and (1 32)
and we may write (compare 3.2)

.93 == d,+(12)J7I'3.
(1 2)J:l3 is the coset associated with d, and consists of the odd permutations (1 2), (2 3) and (3 1). If we let the number -1 correspond to the
elements of the coset and ifwc let the number + 1 correspond to the elements
of the subgroup, we obtain the antisymmetric representation of 9'3 The
identical representation could be called the symmetric representation.
As a matter of fact any symmetry group f/ ft must have these two representa...
tions as is obvious from the explanation in 6.2. It is possible to prove that
for !/3 only these three irreducible representations exist. As a last remark
only the two-dimensional representation is faithful (c.ompare problem 391) .
It

7~S.

FINITE GROUPS

Theqrem: The representations of a finite group can aU be considered to be

unitarY and therefore completely reducible .. Let ~ be an arbitrary representatio~; of order n of a finite group ~. Let us take an arbitrary Hermitian
form: for instance, the unit form F == x~ Xl + x! %2 + ... XJt constructed
with the variables of the representation 16. We submit ~s form to all the
substitutions of ~ and add all the results. This way we obtain an Hermitian
form that stays invariant under all these substitutions, since they only perlnute the terms in the sum. By a convenient choice of coordinates we can
bring this form on its main axis and by a "choice of units", into the unit form
itself (compare problem 3.12) .
The transformation just described is then used to transform aU the matrices
of the representation ( into '" == S~s -1. This new representation, equivalent with ~the first one, is unitary since it has the property that it leaves
"the length of a vector" (x~) x~ + (xl) xi + ... (~). ~ invariant.
It is often convenient not to limit oneself to unitary representations,
as in the example of 7.4.

x:

Ch. 3,

81

UNIQUENESS THEOREM

81

8. Uniqueness Ibeorem: The decomposition of a given representation !I from

a group ~ into irreducible constituents is only possible in one lvay
More precisely, if one finds two decompositions
t;I == t1+"2 +

;t

WII and

= W1+ ~2 + ... ~~,

we must have p :: pi and the two sequences are formed by irreducible representations which are one by one equivalent after changing the order in a
proper way.
In modem algebra which deals only indirectly with the notion of representations this proposition is connected to a more abstract theorem due to Jordan,
Hoelder, and Noether.l
In order to give an idea of the proof of the theorem without goinS into
details we will show it for c)nly three dimensions. "fhis gives the possibility of
specifying the precise meaning of the process of decomposition with the
help of a simple geometric example. In the case of a three-dimensional space
two hypotheses are possible, either the irreducible invariant subspaces are a
plane 81 2 == .%Oy and a line Btl == Oz" (If the representation is unitary the
line is perpendicular to the plane) or they are Li.e three axes til := Oz,
9l~ :; Ox and tI~! == 0)1"
1. Suppose the first assumption is true:

81 3 = al1 +-2
i.e. e~h vector of Ita that iOC$ through the origin can be decomposed unambigUously into a. component lying in III and a component in 8f.l and eB.ch
of t~se will stay in its subspace under the transformations of the representati0ns 'I of the "group '6. It is impossible to find an invariant ,plane Iii
that does not coincide with fIl 2 Indeed if'it existe~ it would cut 812 along a
line L. This line would be a.n invaria.nt subspace of 812 since it is the intersection betWeen two invariant subspa.ces. But III is irreducible, tho tine L
cannot exist and each invariant plane Iti has to coincide with 8ll o In the
same way it is impossible to find an invariant line Iti outside the axes
Oz == tll1 because if it existed it would determine with this axis an invariant

plane different from Uta .

2.. In t1ie second case we have

tllJ == \1 1 +91; +8l1';

1

Compare for instance

SPBlSD.

[1937] section 11.

82

GROUP THEORY

[Ch. 3jO

8~

9'

the irreducible representations are one-dimensional, i.e. the matrices of "

are diagonal. In this case there is no irreducible invariant plane since the
intersection of such a plane with the pIaile illl m~ (which is itself an invariant
plane) is invariant. Finally there could be a line L different from the three
coordinate axes that forms an invariant subspace on itself. In order that the
vectors 11 on this line stay on it after the transformations of the representation
!I, it would be necessary that all components of the vector v%, v, and VI: be
multiplied with the same aumber. In this case the representation would have
one dimension instead of three. Hence the only invariant subspaces are the
three coordinate planes which are each reducible to their axes.
We see from this example the uniqueness of the decomposition ofa threedimensional representation and a simjlar proof can be given in the general
case of n dimensions.
9. Schur's Lemma and Related Theorems
The theorems discussed below are crucial in the theory of representations.
Consider two vector spaces \ll and e, one with m, the other with n dimensions . Let:
1. tIt be a system of mappings or bilinear transformations of \ll onto
itself, consisting of the matrices A R' BR

2. t,~ be a system of mappings of onto itself, consisting of the matrices

As, Bs. .
Th~' matrices of the two systems correspond one by one to each other.
~R and lis could be two representations of the same group, but it is not
necessary to introduce the concept of a group at all.
3. Finally let there be a mapping T of 01 into e. T is a rectangular matrix
such that to every vector x of there corresponds a vector y of ;

y = Tx.

(3.10)

1'he converse in general is not true. To a null. .vector of CS corresponds a

subspace ~' in 8l (this is an invariant subspace of the additive group f!l,
comp. 4.4) and to a vector y which is different from zero corresponds a
coset associated with Bt' (compare 6.2).
Hence T established a homomorphism of 81 into e, or at least between a
part of the space and the space tJl, since it may be possible that there exits

Ch~

3, 9]

83

SCHUR'S LEMMt\.

vectors in e which are not used at all in our homomorphism, that is they do
not" correspond to any vector of R. 1
Having supposed all this we assume:
1. that the system r#R is irreducible;
2. that the matrix T establishes between the vectors AR% and AsY,
BRx and BsY, etc.... the same correspondence as between x and 1, i.e.
AsY

= TAR.%';

BsY

= TBRx;

(3.10a)

or, by taking into aCCGunt (3.10) and leaving out the vector symbol x
As T =

TAR;

BsT = TBl{;

(3.11)

. . .

From these two hypotheses we will show the fonowing theorems (compare
Fig. 3.5).

Fig. 3.5. Sylbbolic representation of the assumptions on Schur. Lemma. Spaces It and e
are represented by point sets.. The mapping T by connecting lines. e' is that part of S
actually used in the mapping T. Since TXt) is equal to zero, then Tx a , Txv, etc. are also
equal to zero. We prove that either . ' = 0 or Bl' = 9t

Theorem ~I. The relation T existing between 81 and ~ is either an isomorphism and Det T =F 0, or T is identically zero. Indeed if we consider the
subspace Dl' of Ul that corresponds to the null-vector of and we let %0
be an arbitrary vector of this space then by hypothesis we have y = T Xo = 0 .

For instance the equations

tllXl+tlJXt+tlaxa;
)" = tUXl +tasXs+/aaXa,
establish a relation sut;h that to ~ch vector It of a three-dimensional space 1t8 there corresponds a vector, y~t the plane y10y, in the space e and if tbis space has more than two
dimensions, its vectors lying outside this plane do not correspond to any vector ina"
The vectors in It that correspond to the null-vector in e are those which lie on a line
whose equations are determined by,puttina the left-hand sides of the preceding equations
, equal to zero. To a given vector y =1= 0 there corresponds a set of ve;ctors in 91. !1aving their
origins at 0 and their end points on a line parallel to the line lnentioned above4
Yl

(Ch. 3, 9

GROUP THEORY

84

The matrix T makes the vectors AR%o, B.. %o, correspond to AsIa,
B s 10' , according to (3. lOa). Because Yo == 0 the latter are all zero.
Hence A.. %o; Ba.o; all belong to the subspace at' which appears to be
invariant with respect to the transformation of the, system ' ... But we have
supposed that this system is irreducible and we are left with the following
alternative: either It' = It, and T%0 == 0 for every %0' ie. T sa 0; or al' == 0
and the null-vector of
uniquely corresponds to the null-vector of fR.
In the latter case we know as a result of the fundamental theorem of 6.2
that the relation between Rand e established by T is a one-to-one correspondence or isomorphism. The collection of vectors 1 of e which correspond
to vecton .J: in It do not have to fill up,the whole space e, but only a subspace
e' which is isomorphic to It and invariant under the transformations of!f,This isomorphism has as a consequence the reversibility of T, i.e. the exist
ence of T-l or the mapping of e' upon Il.
17Jeorem 11. If '6s is also irreducible, e' is identical to e and 8l is is0morphic to
They have the same number of dimensions and (3.11) can be
written as

e.

As == TA.R T- 1 ;

Bs == TBR T- 1

....

ffs

= TWa T- 1

The two systems {#s and ~R are equivalent. They transform into each other

through a change of coordinates.

Theorem 111. Suppose this identification is made

tIt

== '6s ==

(f,

then

(3.11) can be written

AT

== TA; BT=TB;

7)ze matrix T commutes with all the matrices of the Irreducible system 'if
,and we will show below that this matrix is necessarily a multiple of the unU
matrix in " dimensions, or T = A.I, A. being a number.
Let us consider the equation Det(T -AI) == O. This equation will have at'
least one root which is unequal to zero. Let us use this value for A.
The matrix T -1 I will commute with all the matrices of the system !f;
for every value of ,t since T commutes with all of them. The preceding the<)-l
rems confront us with the following alternative, either T - AI establishet!
also a 9ne-to-one projection of It on and the Det (T -lI) 0, whic~
is inlPossible, or T - A, I == O. This establishes the theorem.
SummariziD, we have: A matrix T, which commutes with all the matr~
of the trr.ducible rep,uelltatlon of a group t, is necessarily the multiple of ~
unit matrix. If the matrix relates two non-equivaJelJI irreducible l'epresenta(iold

Ch. 3, 9, 10]

CHARACTERS OF A REPRESENTATION

85

like the relations (3.11), it is identically zero, a statement of major importance in the theory of groups and quantum mechanics.
If the representation is reducible we can easily construct a matrix which
will commute with all matrices of the representation and which is not a
constant times the unit matrix.
Suppose the reducible representation is transformed by a similarity transformation S to a set of step-wise matrices. We construct a diagonal matrix T
that has elements 1 1 , at the places which correspond to the diagonal positions of the first box of the step-wise matrices, elements A.2 on places corresponding to the diagonal elements of the second box, etc. This matrix will,
according to the previous theorem, commute with all matrices in the
representation presumed above. If we bring the reducible representation
back to its original form by the transformation 5- 1 and if we transform T
simultaneously, the commutation relation will be maintained and the matrix
T will not be a constant times the unit matrix, provided of course we take
11 #: A2 :p
Hence we find that if matrices exist which commute with all the matrices
of a certain representation and if they are not proportional to the unit matrix
then the representation is reducible.. If they are proportional to the unit
matrix, the representation is irreducible. This forms a simple criterion
about irreducibility and it will finally lead to a prescription for finding the
irreducible parts of a reducible matrix system ( 12.3).

10. Characten of a Representation

10.1. DEFINITION

Let' A, B, C ... be the matrices of a representation f# of a group ~. The

characters of the representations are the traces of the matrices.
Equivalent representations have the same system of characters, i.e. (compare
3.5) if A' = 5A5- 1 ; B' = SBS- 1 ; ... then

Tr A'

= Tr A;

Tr B'

= Tr B; ....

The trace of a product of two matrices is independent of the order in which the
matrices appear. A similar. statement for an arbitrary number of matrices
holds only if" the order is changed cyclically, which gives

TrSAS- 1 = TrS- l SA

= TrA.

This proof is only valid for finite matrices; in the case of infinite matrices we
have to consider the convergence of the sums.

86

GROUP THEORY

[Ch. 3, f

to

We will designate the character of the matrix A by X(A). ~Its value depend.
on the particular matrix we have selected, just as a function depends otithe
chosen value of the variable xs
The characters of the matrices representing operations belonging to the
same class (in the sense of 4.2) are identical as they can all be represented
in the form SAS- 1 (S runs through all the representation matrices of the
group) and Tt S.AS- 1 = Tr A. For this reason the character is said to be
a function' of a class instead of a function of an element as was suggested
above.
If the representation is irreducible the character is called primititie. 1
Let t: be a representation which is decomposedinto its irreducible elements
(3.12)
where the integers ml indicate how many times a particular representation is
contained in ill. The representations 'lI 0, 'lJ 1 , are not equivalent.
Obviously the characters have a similar relation

(3.128)
The symbols X o , X t , ..... designate the character systems of the irreduci
ble'" representations, i.e. the set of characters X o('-')' Xo(B)~.". Xl (A),
X 1 (B)'i ... etc. They are different only if A"and B are elements belonging to
differe~t classes. Therefore the number of relations (3. 12a), which are distinctJ
is the same as the number ofelasses. We will see below ( 11) that these
character sets satisfy a number of relations between themselves.
10.2. THE NUMBER. OF IRREDUCIBLE REPRESENTATIONS OF A FINITE
GROUP

We have seen the importance of the notion of irreducibility in the preceding paragraphs. (A crucial theorem which we will prove in 11 brings
this out very clearly . )
WeJmow that an arbitrary representati~n of a given group f can 00.
posed accordina ~o (3.12) into its irreduci~le parts. These irreducibl~ ~~s
may differ from one arbitrary representation to another and there seems to
be no limit to the number of possible ineducible representations resulting
from these decomposltions. Actually this is not the case.

decQ,,-

1 Some authors, comp. e.g. VAN DBIt W ABltDEN [1949] reserve the word character Cor the
irreducible representation 'Qnly and use trae. or spur otherWiset

Ch. 3, 10)

87

CHARACfERS OF A REPRESENTATION

The llUTllber of non-equivalent irreducible representatlo1l8 of a finite group

_ I.f equal to the number of classes into which we can divide its elements.
We establish this theorem by studying a special representation of the
group, the regular representation, which is one of the most natural ways of
representing a given abstract sroup.
JO.3. REGULAR REPRESENTATION OF A GROlJP

Before we introduce the regular representation in a more formal way, we

want to point out that multiplication of all group-elements with a certain
fixed group element induces a permutation among the group elements. As
we can see from the group table, all these permutations are di~tinct from each
other.
An arbitrary permutation of elements can be represented by a matrix in
the following way:

o 100
1234)
(
P - 2341 -+

0 0 1 0
0 0 0 1
1 000

This matrix contains only one unit element per row and per column. The
permutation induced by the multiplication by a given element and repte$enteel by this type of matrix is called the regular representation.
, In order to establish this idea in a more formal way we let a variable x"
and a basis vector, correspond to eacb, operation s of a group. The collection
of v~ors s spans a new space Pt the so-called group space, which has 9
dim~sion8 in the case of a finite group of order g. A vector .: of this space
can be written as follows:
(3.13)

the sum being extended over the g symbols .

We define the product of two vectors ~ and" ==
~"
If

;: r,,., xllY,n

L }',t by the rule:

(3.14)

represents the basis vector that corresponds with the operation ST of the

&roup. The expression (3.14) may be read either as a double sum or after
rop1aci.ng the products by the proper group elements as a single sum in
which each element is repeated 9 times.

88

[eb. 3, 10

GROUP THEORY

It is often practical to refrain from the vector description altogether.

The , and " are then considered as hypercomplex numbers! and defined as
linear combinations of group clements with (real or complex) coefficients
X3 and Yt- Hence there is no longer any need to use bold face letters and we
will refrain from doing so from now on. The symbols s, T, .... are the hasU
of the hypercomplex number system and all the quantities obtained this way
form. the group algebra. 2 The structure of the algebra is determined by the
rules that define th.e products u = ST, i.e. by the multiplication table of the
group ( 2.4).
The two expressions "group space" and "group algebra" as \vell as the
notions of "vector in the group space" and "hypercomplex number of the
algebra" are equivalent. We will use the latter expressions jf we want to
stress the multiplication rules (3.14).
From the last equation. we find that an arbitrary operation A of the group
generates a projection from the space p upon itself,
~

(t

(IS

--+-

e'

Ae

= Lxsas
s

corresponds with the product

L,xa-ttt.
t

of the operations

and s and we

have s = A -1 T) ..
The collection ofprojections or mappings A/orms the regular representation
of the group.
Let us: designate the components of ~' by x~ then the preceding equation
is equiv~ent with the set of substitutions
t

= a, b, c, " ....

Hence .the matrix A can be written

= (a ts)

= (<>$tO-1t)

= (Ja

t .'f-

1)

l3.1S)

in which the rows and columns are labeled with the help of the elements
of the group themselves: E, A, B, .... S . T . and h",ts-1 is equal to 1 if
A = TS-! and zero otherwise. The matrix contains mainly zeros and only
a single 1 per row and per column.
If we return to the example of the group 9' 3 of permutation of three
objects, referring to its group table displayed in 2,,4 (second form), we
see that the~D:iatrjx A that represents the abstract element A is the following
1
2

Compare problem no. 3.

'"fIle ~Nord algebra j:-, ~Pg;ested because both

SUIt) 3.nd.

product are defined...

ORTHOGONALITY RETALIONS

Ch. 3, 10, 111

89

arrangcl11ent of 0 and 1

A=

001000
/ 100000
010000
000010
000001
000100

The importance of the regular representation is due to the following theorenl. All irreducible representations of a group r9 can be obtained by reducing
its regular representation ~,. . The number of times an irreducible representation appears in the regular representation is equal to the dimensionality of the
matrices of that jJarticu/ar representation.
The proof of this theorem can be established directly from the result in
11.2.. [Compare problem 3.8 where it is explicitly indicated that one
can obtain this proof from (3.23).]
1'\aklng up again our example of Y3; if one would reduce tile regular
representation of this group, one vlould find tIle unit representation once,
the alternating~ one-dimensional representation once, and the irreducible
two-dimensional representation mentioned in 7.4, twice.

11. Orthogonality Relations (Finite Groups)

11.1.. GENk.RAL FORMULAS

Let r! ~nd t' be two irreducible representations of the finite group t and
(alk), (aik) their representation Dlatrices; nand n' the order of these matrices.
An important case is the one in which t' is unitary, but first we will deal with
the g~nera1 case. Consider a rectangular matrix S = (S,p.) with n rows and n'
colurnns and let us form the sum

(i)
extended over all operations A of i .. (The number is of course equal to the
order of the group g.) T wiU be, like S, a rectangular matrix with n rows and
n' columns. Let us write this ex:pression explicitly:
tile

L (all S'p a~il + bll SIJt b~;

j.

1 ).

I'

Let {--: a;'ld C' be t\VO matrices of t. and

~f

cOJLresponding to an arbitrary

90

GROUP THEORY

[eh. 3, 11

element c of t6. We claim that for an arbitrary c the following expression

holds
CTC,-l == T.
(il)
Indeed

eTC; -1

:;;::

L C~4SA'-IC'-l,
.....

,
if we call CA == D, then A - 1 C,-1 == (C' A')-l == D,-l and D and DI
correspond .in _ and !l'to the same element D == CA of rI.
If A runs thro,uah the group (I, i.e. represetlts successively all the operations
of f#, D - CA. goes through the same group, but in a different orda- (the
same idea was used in 3.2).
Hence
(,"'TC,-l ==
DSD,-l ==
ASA.,-l = T,
which proves (ii).
For arbitrary c the result

.0

:s
CT

= Te'.

(ill)

If we apply now Schurs' lemma we find that either C and C', i.e. !!I and {I'
are non-equivalent and hence T :5 0, or t'6 and "lJ' are equivalent and T is
a constant times the unit matrix.
Flrs_ Case: t'6 and t' are non-equivalent; (i) gives

L ASA'-l

==

o.

The ~trix 5 is arbitrary. If we take a matrix that is everywhere zero except

for the component

S"1

we obtain

(3.16)
and in the unitary case this means

r aUca~! = o~

(3.16a)

In the relations (3.16) and (3.16a) the indices i, k, v and l are arbitrary,
i.e. we ~."an arbitrarily located element in the collection of matrices, and
keep the position fixed in going from one matrix to the other. These relations
are important since they characterize the inequivalent representations.
SeCiJnd Cafe: 1M and t' are equivalent. We can choose a new coordinate

ORTHOGONALITY RETATIONS

Ch. 3, 111

system for the representation t'/J' such that ~'

can be written

where

= 'li.

91

In that case our theorem

is a constant depending only on S.. Writing this out, we have

L r. ausJt ali

JIt.

= tXbn ..

A.

Since the matrix 5 is arbitrary, let us suppose that all the elements are zero
except one of them sJt == 1; the preceding equation then becomes
(iv)
where the constant tX, determined by the choice of S, i.e. by the indices j a..s~d
k, is independent of i and I. In order to determine C( we have by definition,
(v)
(The reader should not confuse this summation, which is extended over then
dimensions of the representation space, with the summation
extended
over the 9 operations of the group ~~)
In eq1.lation (iv) let i == I, and if we perform successively the two summations with respect to the indices A and i we find, taking into account (v),

L.

~
L

i m. 1

.J.a:~
,.,.~ 1 :::r:

) ' (I"

.;.J

rlrx = .al5LJ
",~
A

}~

or finally,

In the case when i :F I or j =F k,

L llikaki

~ gJll.

In case the representation matrIces are unitary the equations take the form

L a,t a;, == 0
A

and

L al1,a: == gIn.

(3.19)

The equations (3.16), (3.17), (3.18), and (3.19) are the basic orthogonality
relations bet}veen irreducible representations. The use of the term orthogonality
will be explained in the remaining part of this section.

92

GROUP TIIEORY

reb. 3, 11

11.2. APPLICATION TO THE CHARACTERS OF IRREDUCIBLE REPRESENTA..

TIONS

Restricting ourselves to unitary rrlatrices, we have for (3.19) putting i = k,

L aiia~

= gin.

Hence using the right side of (3.19), we have the following relation bet'~leen
the characters of a unitary irreducible representatioll

(3.20)
Similarly if we deal with
find from (3.16):

t\VO

non-equivalent irreducible representations we

L X'(A)X*(A) =

o~

(3.21)

X(A), X'(A) . are functions of the variable A representing an element of

the group (compare 10.1). These functions are defined only at 9 discrete
points. (Moreover the "shape" of thls function is further simplified by the
fact that-its-functional value is ~he same for all elelnents belonging to the
same class. Hence the characters are actually functions of a class instead of
functions of the elements.) The function space of this function is identical
with the group space mentioned in 10.3. Or to formulate it differently
we! may construct a hypercomplex number by using the character of the
eleinent as the coefficients of every element. Using the geometric interpreta
tion we use the vector symbol

X =

L X(A)

It.

In order to write the preceding equations in the form of an orthogonality

relation it is sufficient to normalize theta with the help of a factor 1/

.Jg:

L
A

.!Jg.Jg
1 X<P)(A) .

rt1)*(A)

ODg

(3.22)

where the superscript labels the irreducible representation.

The normalized primitive characters of the inequivalent irreducible represen. .
tatiofii oj a group <; form in the group space a system of orthogonal jilnctions
or a system 0/ orthogonal vectors.
One may express the equations (3.19) in a.similar way. The result is that
according to well-known theor~ms X(A), X'(A), . ., are linearly independent

Ch. 3, 11]

ORTlfOGONALITY RELA1"IONS

93

of each other. The same holds for the matrix elements ail" ail Of,
according to (3.16) for aik, a~.A. ..... 1
Similar statements can be made for continuous i.e. (infinite) groups. For
example if we have the group of rotations around an axis, equation (3.22)
becomes the well-known relation

~f2ltexp (im'lp) exp (-imlp)dlp =

2n

0mm"

Finally let us consider a reducible representation. Equation (3.21) applied

to (3 . I2a) gives

(3.23)
and

1 .t...J
'"'
-g

X ("
A)X*()
A

==

2
111.0fm 21

+ .... ,

(3.23a)

where mil is the nUID.ber of times the representation Jl is contained in the

reducible repres~ntation. From this we conclude that:
1. (3.20) is the necessary condition for the irreducibility of a representa-

tion {IJ and (3.23) sho\vs that tills condition is sufficient.

2. The necessary and sufficient condition for the equivalence of tyVO irreducible representations 'lI and W' ~f the group C!f is the identity of their character
system~

Indeed this condition is necessary because a change of axes will not

change the traces of the matriceso It is sufficient because, if it were fulfilled
we would have

! L X'(A)X"'(A) =
9

1,

contradicting (3.21) the necessary condition for non..equivalence. Hence the

characters completely determine the irreducible representations and the
irreducible representations completely determine the characters.
11.3. CLASS-SPACE

We have mentioned in 10.1 and 10.2 that the characters are functions of
a class. WewiU indicate the p classes into which we can subdivide the group
t: by C 1 ,. C 2 , C 3 , C i . Cp Each class contains Ill' hz, hi, .. " ., hp

elements of the group.

1

Compare problem

s.

94

GROUP THEOR.Y

If we introduce X,
written

[elL 3, 11, 12

== ~hJgX(C,) the equations (3.20) and (3.21) can be

1' ....

,, _ _

,=L1 XiX: == 1

and

L X;X: == o.

(3.24)

fa 1

We are. led in this way to consider a class space of p cJ.&nensions which is

derived from the group space by ignoring the differences between the elements
of the same class. The successive irreducible representations fI, " . .'.
correspond to the vectors X, X' . . of this space with components X, ... ,
X;
According to (3.24) these components are ortho_onaLBecause there
can only be p orthogonal vectOR in a p-dimensioDal space, the number of
irredudble re.presentatio1l3 0/ the group ~ mutt be leas than or equal to the
number of cwaea. The theorem in 10.2 states that they would be equal, ic.
the vectors X form in the class space a complete system of orthogonal axes.
This theorem :is also true of the vectors X with components X(A) in the
group space .
-

t\

..

...

111#

e.

12. Sam of a Ous; Projection Operaton

12.1. DEFINITION OF THE SUM OF A CLASS; STRUcrURE COEFFICIBNTS

Suppose we take the sum of all elements belonging to a given class i:

Ci

=::=I

AU)
1

+ A(2i) +

A( I}
II,

(3.25)

where".'~

the number of elements in the class. If these symbols represent

abstract group elements, c, is a hypercomplex number (see 10.3). Most of
the time we will apply this defJ11ition to a representation of the group.. In
this case we write:
~(Q+
A(r)
(3.25')
C, -- n~(f)+
1
.n. .2
~
lat
and the matrices C t are obtained by adding the corresponding elements of
the matrices A 1 , A 2' etc.
If we take a product of two of these quantities, the group elements at the
right... hand side can be grouped .into classes, that is, the result can be written
as a linear oombination of sums of classes, the coefficients citedbeina either
zero or integers:

(3.26)
To show this we take the conjugate of the left-hand side with an arbitrary

SUM OF A CLASS

Ch. 3, 121

95

element x which will only permute the elements in each sum of a clasa. As a
result the right-hand side will never contain a set of elements which do Dot
fill a whole class, because all possible conjuptions would generate the miasing members of that class, and because the left-hand side must be invariant
under any conjugation .
The coefficients Cal' which can be determined from the group table, reflect
the structure of the group_ With the help of these structure coeffjcitmt8 we can
determine the character of all possible irreducible representations of a given
group.
The hypercompl~x numbers C, formed by the sum of a class commute with
all elements of the group, since
(3.27)
where x is an arbitrary element of the group; X-I c,x = c, because conjugation only permutes the terms in the sum accordina to the definition of a class.
If we now take the representation by matrices of (3.26) and apply one of
the theorems of 9, we conclude that these matrices are constants times the
unit matrix i( the representation is irreducible. (It is not di:fftcult to obtain
the explicit value of this constant but its value is not needed for the following
argument.) If two matrices are multiples of the unit matrix the product of
their:traces is equal to the trace of their product times the dimensionality of
the matlrices.
Applyi~ thls to the matrix representation of (3.26) we find that

,.

h,Xth1Xi; == n 12 cUt1h,X,

(3.28)

1-1

becaUse the character of the sum..of-a-class matrix is equal to the character of

that class X, times the number of elements in the class hi. 11 is the dimensionality of the matrices of the particular irreducible representation we are referring to. (This number is of course equal to the character of the unit element
Xl') Note that the relation (3.28) only holds for primitive characters.
12.2. CHARACTER TABLES

In this subsection we will give the complete proof of a statement made in

10.2, namely: The number of irreducible representations is equal to the
number of classes. This was partially verified in 11.3. However, the follow...
ing proof is independent of the arguments used in t~t section .
In equation (3.28) the representation was not specified. If we sum over all

96

[eb. 3, 12

GROUP THEORY

irreducible representations we have:

r'

r'

L hlX1")hkX~) = L n<p) L Ciklh,xf

p=l

p=l

lt

),

(3.29)

1=1

where r' is the number of irreducible representations. The right-hand side

contains a factor, which we will evaluate separately:
r'

r'

L n{p) X1/l) = L
,,=-1

JY(Il)(E)X<p)(C ).
1

1'=1

This is simply a change in notation since n is the dimension of the representation and X(E) the character of the unit representation. However, according
to problem 3.8 the characters thus obtained are the characters of the
regular representation, and these are all zero, except of course if c, = E.
We find
r'

L X<P)(E)X(P)(C,) = Xreg(cz) = g6'1 ,

Il

since the dimension of the regular representation is the order of the group.
Substituting this into (3.29) we obtain:
r'

L X<1t)(Ci)X(P)(C;l) = g/h,

(3.30)

,.::01

Cr

1
The symbol
indicates the sum of a class of elements inverse to the elements in the class C,. This may be either the same class or a different one.
In bot1), cases it will contain the same number of elements.
If th~ representation is unitary, we have
I

,,'

L X*(p)(Ci)X(Jl)(C

j)

nlht .

p=l

This orthogonality relation looks similar to (3.20), which can be written as

follows
r

L h,X*<P)(Ci)X(Jl)(C

j)

= g,

(3.20a)

i=l

but is entirely different in nature. Equation (3 . 20a) contains a sum over all
classes i = 1, ... r while (3.30') contains a sum over all irreducible representations p. == 1 . . . r'.
The theorem is now easily proved if we consider the double sum
r

r'

L L h,X(")(Ci)X<I')(Ci )
i=ll'=l

depending on which _sum. was taken first.

== gr =- or'

SIJI\;f OF A CLASS

Ch. 3, 12}

97

As an exalnple we give the character table of the group mentioned in 2.4.

It has three classes C z: E the unit element, C 2 the two rotation elements
around the axis normal to the center, and C 3 the three turnover elements.
Using the results of 7.4 the following character table is obtained:
I

-------I-~------i--------~---

p=l

.~t=2
i~i. ~-:: 3

1
1
2:

I.
I

i l l
II
1
-1
I

--- 1

()

F'or other cha.racter tables see for instance

KOSTER

[1957].

12.3. PROJECTION OPERATORS (IDEMPOTENT ELEMENTS)

A matrix that generates a given irreducible representation is called a

projection matrix. Consider a set of matrices already in box form and multiply
by a diagonal nlatrix as indicated below"

I
x;
.
xx
X

f~~r'

--'

......... ..,

...

\\

l xxx

I '

!.

lxxx/
,

I xx x

_. . . . .~~ __ ~~ .. ~ .. _4~

_ _. _ .

i xx t

--""'

""

/ 00;

! ...xx;
-..... ...........................

00 i

x .......
x -~,I.....--.........
...-....-.f~ .---

i
.~

; 000
.~ 000

000

This DJ8trix has zero elements except at those places on the diagonal which
corre~pond with the box we want to "project". The irreducible representations can be considered as "components" of the reducible representation.
The projection matrices 8 are also referred to as idempotents because they
have the p~operty elt = B for any power 11.
The result at the right..hand side is a set of matrices that are again isomorphic
(or homomorphic) with the group_ They form an irreducible representation
if we omit all superfluous rows and columns (those that contain nothing
but zeros)~
If the representation was not in box form we could bring it into this form
by a. transformation A ~ S -1 AS ~-= AD" The representation in box form is
designated by a superscril,t h. Conversely if v{e consider the inverse transfor':'
mation applied to the equation
.
(3.31')

[Ch. 3,

GROUP THEORY

12

representing the matrices depicted above, \ve have

S8b(~)S-lSA~-1

= SA

b (I-l)S-l

or

e(PJ,A

==

..4(11).

(3.31)

It is possible to construct a priori the matrices s(p). If we form all products

8(J1) A the result will be a set of matrices ti("> which can be reduced in size.
This is done in this case not only by omitting rows and columns that are
zero (as mentioned above) but also by noticing that certain rows or columns
are linear combinations of other rows or columns and keeping only those
that are independent.
The procedure can be s1lD1med up as follows:
Projecting a certain irreducible representation out of a reducible one consists of three steps. (i) Convert the matrices to box form; (il) Multiply by
the diagonal matrix indicated above; (ill) Omit the unnecessary rows and
columns in the result. It turns out. however, that steps two and three can be
taken first, making the first step" which is in practical cases of course the
most elaborate one, unnecessary.
After this introduction we display the projection operators or idempotent
elements. Th~y are
e(l')

== (nlJ/g) L X<I')(A -l)A

A

=:

(n,t/g) L X:(i)C,

(3.32)

where 9 Ul the number of elements of the (finite) group, nil the dimensionality
of the itreducible representation jJ., .U the characters of this representation.
The s~ is taken either over all the matrices that form the reducible representation of the group~ or as the right. .hand side of (3.32) indicates the sum
over all C i Since the characters of all elements are equal as long as they
beloJig to the same class it is obvious that the second sum is identical with
the. }first~ provided the representation is unitary.
'The matrices 8(") can be constructed in a straightforward way from a given
representation if the characters of the irreducible representations are known ..
These can be calculated from the structure coefficients as mentioned in
12.1 and these in turn follow directly from the abstract group table. U"uq.Uy,
however, the necessary character tables can be found in the literature..
If a calculation is perfonned t it is of course advisable to investigate first
whether ~p.reSentation Jl is really contained in the irreducible representation..
This cati'be done with equation (3.23)~
<Finally we hav~ to demonstrate that the products 8(1') A really form for
all A an irreducible representation p.. The necessary and sufficient condition
is that the chat1h,.~ers of 8(1') A are equal to the characters of the irreducible
I

Su~J

Ch.. 3. 112]

99

OF A CLASS

representation Jl
(3.33)

In order to show this we first point out that

.de(l')

A-I

==

8(1'),

s(jJ)

commutes with all A:

(for every A belonging to !f)

since a(}4) if made up of class-sums which are invariant under conjugation as

mentioned in 12.1.
Let A be, for the moment, an irreducible representation, for example v.
Then from theorem m, 9 we conclude that
8(1')

C I.

We calculate the constant c by taking the trace of this equation

X(s(p == n" L X(I')(A -l)X(.d Y)

9

=:

nfJ.611., ..

(3.34)

(We must assume that A contains the representation v only once.) From
this we conclude that
c :: 6". ~
(3.35)
If if is reducible and in box form, the matrix a(}l) will be a diagonal matrix
with the diagonal elements zero except at the places corresponding to tho
box p., according to (3.35). If the representation is in box form equation
(3.33) holds. But then it should hold for any representation, because tho
characters are invariants.
The p~escription for performing a reduction is the Conowin&. We have a
set of basis functions and operate with t.he prescribed operations of the
group in: order to create matrices. Now, depending on how complicated the
problem is, it might be worthwhile to see which representations are contaiiled
by taking the trace of each matrix (one of each class is CllOugb, if the classes
were not known this is the moment to find out) and determining the coetTicients in eq. (3.12a)
If a certain irreducible representation is contained one constructs the
projection operator (3.32) and inspection of these matrices usually tells us
which rows or columns are supet;ftuous.
If not obvious we mUltiply the projection matrix with the basisl! The
resulting linear combinations are either zero (in case a certain row contains
nothing but" zeros) or partially dependent on each other. We take from each
set of dependent linear combinations one and drop the othen.
As an epilogue we should like to point out that the idea of projection
operators is solely based on the theory of characters. First, the only reason
~

100

GROUP Tf-IEORY

[ell. 3, 12, 13

that the projections (!/p).4 are the ll1atrices ~Ne want theln to be is that the~
have the proper characters, and that fact is sufficient for irreducibility.
Second, the way the operators G(p) are constructed is to make use of the
orthonormality of character systenls, and this was used (3.34)"
If a certain irreducible representation is contained several times. in a
reducible representation!t one has to proceed with more caution.. In this case
the projection will result ill a linear combination of the multiple occurring
representations. It is necessary to perform an additional orthogonalization
in order to obtain the box form. "There is an arbitrariness in this procedure
similar to the arbitrariness in tIle choice of orthogonal wave functions
in a degenerate eigenvalue problem~
In the important case that the reducible representation is the product of
two irreducible representations of the rotation group, the method does not
apply, because these are continuous groups. This case can be treated in an
entirely different way (Chapter 5 : 5~3).
13. Representations of the Permutation Group
13.1. YOUNGTABLEAUX

In the special case in which we are dealing with the permutation group it is
possible to obtain irreducible representations in a clear and concise way.
The ideas explained below are originally due to A. Young, WilO published a
nu~ber of papers on group theory around 1900. 1
If! we consider an element of the pernlutation group and if we employ
the hotation with cycles for instance,

(1. 5 3)(24)(7 8)(9)

(3.36)

then we can write these numbers in a so-called Young tableau as follows:

(3.37)

A. Young was a country clergymanQ

Ch. 3. 131

REPRESENTATIONS OF THE PERMUTATION GROUP

101

Every row corresponds to a cycle and hence the numbers in a row can always
be cyclically permuted . It is the custom to draw the different cycles in decUma,
order, that is to say the first row is the longest cycle, the second is either t4e
same length or shorter, etc. If there are two rows of the same length the order
in which they are written is irrelevant. The shape or contour of this scheme
determines a class of equivalent permutations. (Compare the end of 5.2.)
This is a one-ta-one correspondence. If two schemes are of the same shape,
they belong to the same class of permutations and if they are not the same
shape they belong to a different class of permutations. By the shape we mean,
of course, the'empty box, i.e., the tableau without the numbers. Instead of
describing the shape or the class of permutations by indicating the length of
each row or the length of each cycle, one can do just as well by telling how
many cycles there are of order one, how many cycles there are of order two.
how many cycles there are of order three, etc. The number of cycles of a given
order, that is, the number of rows of a given length, is sometimes called the
rank number. If n is the total number of elements which we are permuting,
we have the following equality:

(3.38)
where the rank number at indicates the number of cycles of order k.
The number of permutations that belong to a class C{.,} characterized by
the set of rank numbers Ql,. Q2' 03 t ~ a" == {a i } is equal to:

(3.39)
The derivation of thls formula is easily demonstrated with the help of example (3~16). If we take the corresponding Young shape we have h, i.e. in this
8 open places and hence can fill in the numbers in 8f (h!) different ways.
The nWIiberof rows of length k is ai and they can be permuted in any order.
Hence we have to divide the total number of ways hI by al! in order to get
the total number of classes. Now we can cyclically permute all the numbers in
each one of these rows which gives a factor kale since each row allows k
cyclical permutations. That means JcD"a,,! in our example (3.37) 3 for the
first row and 22 21 == 8 for the next two rows. The number expressed by
(3.39) is found in different places in mathematics and physics. In the theory
ofnwnbers it is called the partitio numerorum, i.e., the number of ways one
'tan assort a certain number n, of objects into piles with al piles of one object,
a2 piles of two objects each, etc . To physicists this factor is vleIl known from

cUe

GROUP THEORY

102

[eh. 3. 13

the work of Mayer in the development of the partition function in statistical

meChanics. (MAYER and MAYER [1940] p. 437.)
13.2. THE p. Q-OPERATIONS; IDEMPOTENTS

We will introduce a special set of elements which will eventually lead to the
construction of an idempotent.
Conjugation oean arbitrary element T with respect to p (i.e. TJ = p-1 TP),
will have the effect that the Young tableau of T will maintain its shape. The
numbers in the shape will be permuted in the way described by P. We consider
only the elements p which will create a horizontal permutation in T. That
is to say, the numbers in one row will be permuted but will never leave that
particular row. Such elements form a subgroup_ We will call this subgroup ?I
and its order is equal to nta,,! (Compare Fig. 3.6 for the general Youn,
scheme.) Not all elements represent different permutations. In a similar way
one can introduce those permutations, which, if used. to conjugate given
permutations, create permutations only within a given column. We call this
subgroup !2. Any rearrangement of numbers in a given tableau, i.e . keeping
the shape the same, corresponds to a similarity transformation
_-1

Pop T = P

(3.40)

TP.

This holds for both p and Q.

We introduce now P Q, which indicates that we are dealing with an operatio4 which consists of a vertical permutation followed by a horizontal
.......~I.kl1.,tation. Such elements, which do not form a subgroup, will be used
~} . " . ~~ a hypercomplex number, by the following convention: Sup:
pdse ..... econjugation of the Young scheme cannot be obtained by the opera:.
tion P- Q, 'th~~ the coefficientC will be equal to zero. If, however, it can be o~
tained by the operation p Q then Cwill be either plus 1 if Q is an even permutation, or minus 1 if Q is an odd permutation. In formula we have

(3.41)
is an element of i?,. either a reducible or irred ucible representation of this
element. or, in the true sense of group algebra, an abstract element. The

R.

number$/{(lt) are:
OifR:;'PQ

'(R)

=(

1 if

==

PQ and Q is an
- J if It PQ and Q is an
R.

even permutation
odd permutation.

(3.42)

Ch. 3, 13]

REI'RESENTATIONS OF THE PERMUTATION GROUP

103

We want to study the set of elements P . Q more carefuUy. First a word of

warning. It is. not true that pc. Q corresponds to a similarity transformation
with Q followed by one with P. The second similarity transformation is
Q-1 PQ instead. 'This is easy to see" F'rom (3~40) ,ve have: JP )0" T 5 (fJF)- l
TQP = (p-1 QP )oppop T. Second) it is interesting and useful to have a criterioIl
that tells which elements of R belong to the set F Q and 'Vt-Jcl1 do not. The
elements R belong to the set p .. Q if and on.ly if, two numbers of a. given col...
umn never arrive in the salne rOtH. TIle condition is necessary, as one ccin
easily see from an example.

Fig. 3.6. Tableau.

To show that the. condition is sufficient we separate the Y ()1Ulg tableau

into two parts. The leftpart P, contaiD~ the first i columns, tIle rignt part p,
the columns labelled by i+ 1, ... , at. 'The operation R never brings
t'VQ
..
nUD1bers of the same column into the row", hence the lEi numbers of rl
come from different columns. The numbers in the posi.tions ,+l~ " ., ~1.
of the row r 1 originated. from the columns in p", hence these numbers call
never come from the rows '.+1' 'S+2, . etc.
The second criterion is; if B. does not belong to P- Q. the ,elements R ....: 1 PR
1

'

"?

~-

and Q will have at least one transposition in conunOt;1~

The hypercomplex number' defined in this way has theprqpertythat'its
square is proportional to the number itself, as will be shown below. Such a
number is called an essential idempotent because it is possible to make jt
idempotent by, multiplication with' an ordinary num~rll Hence if.we take
the square of (3~41) ,ve find by using the nn.lltiplication l'1des f<]f 11~11er-

104

GROUP THEORY

rCb. 3, 13

complex numbers of 10.3

,2 ==

L C(R)C(R -IT)T.

(3.43)

With the help of the result (3A.6) of the Appendix to this Chapter we find
{2 == pC and with
p-t C == 8
(3.44)
we have 8 2 := 8. The number p is a real integer as can be seen from the derivation in the Appendix.
The fact that the , is aD essential idempotent allows us to use this operator
as a projection operator. Any idempotent a CulfiDs the equation ';-8 =: 0,
hence the idempotent must have eigenvalues 1 and O. If we multiply the
matrix 8 with a set of basis functions, all components that correspond to a
zero in the diagonal of the a-matrix will be removed. Hence the number of
independent basis vectors is reduced. If the 8 were Dot in dia,onal form
the last arpments would still hold as we have shown in 12.3. An obvious
aspect of projection, viz. projecting twice gives the same result as projecting
once, is automatically fulfilled since 8 2 == 8.
13.3. IRllEDUCIBLB REPRESENTATIONS

We return to the regular representation described in 10. The basis of the

regular representation can be considered in two ways. Either we consider
the basis B, A, B, s, . . . as abstract elements, or we consider the basis
re~Dted by matrices B, A., B, ... S .... These' matrices are operanD, in
turn qn a "subbasis" which is usually not written in the formulas. (This is
rem.inlsceut of second quantization or field theory where the operator wave
ftmcdonoperates ott nothing "visible" either.) The distinction 'between these
two! cases is irrelevant but helpful in order to make the following pOint.
f
,.'
,Wewant to use the projection operator or idempotent B to induce a new
~pfCSentation in the regular representation. Wedo this by operating With 8
on the usubbasis" and hence the elements B, At B, are replaced by 8,
AS, Be, ., sa Again one can either consider this as a set of abstract
4

elements as ,above, or as a set of matrices EI, Aa, Be, ... , Sa, ... where 8 is a
matrix 'which bas the property that the eigenvalues are either zero or one.
Apin, this cJ.iatinctjon is ~levant but wo wanted to make it clear that the
projection.-..operators are written on. tM right. 1
1

In the laDlU... of aroup alaebra all hypen:omplex. numbers obtainOd by muitiplyina s

at the left by the hypetCOmplex numbers or the _bra form a subset or the allebracidtoCl
tho left ideaL In this language the minimal (simple) left ideal corresponds to our irreducible
representation.

REPRESENTATIONS f)F THE PERMUTA1"ION GROUP

Ch.3 , 13]

105

We want to calculate the characters of the representation induced in this

way, since we finally ,,'ant to show that the special choice of idempotents
resulting from the Young shapes leads to irreducible representations and
irreducibility is easily demonstrated if the chara,cters are known.
The representation matrices induced by the idempotent are

s -+ ASs =

A-

La C(R)R
(3.45)

That is, according to 10.3 the element

is represented by a matrix:

Hence the diagonal elements of this matrix. are obtained by taking s

= T and

the trace is

We indicate with {a,l a certain partitio, hence a certain class of elements

and henqe a certain shape. The hypercomplex numbers ,{QJ or t{tt} derived
from thi~, shape will induce a representation of which the element A has a
characte~iX{tI}(A). In order to see whether this representation is irreducible
f

we calcate the sum (3.23a) and show that the right-hand side is equal to
one~ USing (3A.7) we find indeed
g-1

L X{1I}(A- 1 )X{4}(A) == (gp2)-1 L L L c{a}(s-l A-l s)(fal(T- t AT)

A

= P-2 r L ,{tI}(S-l A- 1S),{G}(A) = 1.

A

(3.47)

The result obtained implies that all irreducible representations of the

symmetric gr~up can be obtained by a straight forward method. (Compare
problem 3.13.) The tableaux are not only useful for this purpose but arc also of
great -help in the decomposition of representations that were irreducible
with respect to the symmetry group of n elements but are redu,~ible \vith
respect to a svm.metrv )!TOUt')
I'" of 11. - 1 elenlents. if: for instan.ce. \ve .ha~le the
.;

'"

~i-

~'

106

GROUP THEORY

[eh. 3. 13, AI

permutation group of three elements, the Young tableaux are:

r-~-'-"

l_J I

I
I

-~

We can decompose very simply each one of these irreducible representations

into tllose which are the irreducible representations of the permutation
group of two elements. Ac.cording to the following scheme:

11J ._-

r1
I

L~

r---l

~--lI

.,.

~,--l

IL
1

____:l

The plus sign indidates a direct sum of t\VO permutations. This ide;a is
extens~ely .'used' in electronic and nuclear spectroscopy.. If one wants to
find tlte;,. ,.>';
~.Jtevels. of a n-electron system from the knowledge of the
wave
and energy levels of a n - 1 electron system. In such a proce~
dure it is necessary to form products of linear combinations of wave functions but not aU of these products obey the Pauli principle:. Hence one is
interested in only certain representations of the permutation group. In this
case the pennutation is with resfJCct to tIle electrons .
~

" ~:;.:!+.l'

"?:

In order to prove the equation mentioned in 13.,3, Vie study the properties
of th.e hypercomplex numbers C (compare MOLENAAR [1930]) . First of aU
\ve notice" ~that
((E)

1;

C(p) = 1;

'(Q):;-~

Ch. 3, AI]

107

APPEN11IX

which follows sip.1ply from the definition of

t(PR)
since either R

::; PI Ql

(PR)

==

It i"! also easy to see that

(3A.3)

'(R)

:=:

and according to

~.

(3A~1)

we

$"

ha,;I,;~:

(PP t Ql) = ((Ql) and (R) = ({PIQ1) := '{Ql)'t

'*

or R
Pi Q1" In this case the right-hand side is zero . 1~h(~ Jeft. band side is also
zero because PR =F .P P l Ql is also not a product of a .P and a (I . Hence the
second possibility satisfies the equation as vFelL
Following the saIne line of reasoning one can ShO~1 the relation

(3A..4)
since either

R ::..~ PQ,

then on the one ha.nd \ve have

'(RQ)

= (Pl Ql Q)

= {(Ql)C(Q)

llsing (3A$3) and (3A.2). On the other hand

(a)C(Q) = '(P1 Ql)C(Q) = ~(Ql)'(O)

according ~o (3A.l). Or if R. #: PQ then both sides art.;' zero since also

RQ -::f. P j Q1 (:) == P1 Q2'
T~le have no".v enougll mate-rjaI to show the first. su.tnnlation fonnula:

L ((APB) = L '(AP)'(~)4
P

Either: B

= P 1. ()t

(3A.5)

and we have, according to (3./\.4)

L'CAPB) = L'(APP 1)(Ql) = I((AP){(Q-c),

since p and PP 1 are both summ.ed over all elements . l~he right-hand side, using
(3A.l) can bt:' \vritten

L C(AP)'(P 1Ql)
p

:=

L C(AP)'(Ql)
p

Or if B ::/.:. P:t Q i the right..,h.and side is equal to zero and since according to
the second criterion B- 1 PB and Q h.ave a transposition in (~ommon: Qz =
l
B- P2 B Hence replacing p by PP2 in the sum 'we have

2: C(APB) L C(APP2 B) = L {(APBQ2) = L C(APB)t:{Q2)

=:

f!

.,

according to (3A.4)p Since Q2 is a transposition, we have {(Ql)

the sum is zero because

2: '(APB) = - L (APB).
p

=:

--1 and

[eh. 3, AI

GROUP THEORY

108

The next formula we want to prove is,

L C(R)'(R -IT) == pC(T),

(3A.6)

\vhere p is an integer~ Call the left. .hand side L and replace

If ",'e now sum over all elements p then we have:

using (3A.3) and

It

by

p-1 lt

(3A~5).

The number p represents the total number of elements that have the property
that they only create horizontal permutations in t]le tableau. If we replace
It by PR then we have

PL == LLC(PR)C(R- 1p-l p )'(T) =

P

PLC(R)C(R-t){(T)

according to (3A.3). ,Hence we find (3A.6) where the integer p is defined by

p

== L_C(R),(a- 1 ).
The

la.~t

formula \ve want to prove is the following double sum:

12 L ({S'-lR -, lS){(R) = p2.

It

If we again call the left-hand side

L and replace R by PR and take the sum

over P, we obtain:

PL == LLLC(S-lR

- l p -1

S)'(PR)

L L L ,(s-l a - lp -l)C(s)C(a),
F

acc;Ordins to (3A.S) and (3A.3).

Ifnow we replace R by p-l. then we have:
P

L == L r L C(S-1:R- 1pp- 1X(s){(p-1 R)

P

..

== p}2 L C(S-llt-l){(S){(R)
R

apin using (3A.3). The result ist that by replacing s by ps and summing over
P, we will have:

APPENDIX

Ch. 3, AI]

109

using (3A.5) and (3A.3). The sum over P can be performed by replacing
s by PS and \ve obtain, using again (3A.3):
P

L = L L L (s-lpp-l)C(a- 1)C(p"'l s),(R.)

PitS

== P L L '(S-lX(R.- 1){(S){(R) = pp2.

It

This completes the proof.

CHAPTER 4
GENERAl.; APPLIC.t\TIONS TO QUANTUM 1\IECHANICS;
~
"TIGNER'S THEOREM
1. Invariant Properties of the SchrOdinger Equation
1.1. THE T\V() GROUPS OF THE SCHRODINGER EQUA'rION

'There are two important groups that leave the Schrodinger equation,
Lt~", the Ifamilton operator II, invariant.
1, rfhe p~rmutation group - the exchange of positions in space - between
identical particles, either electrons or nuclei. This group will always leave H
invariant . The quantum theory makes the fullest use of the indistinguisha[,ility of identical objects.
2e 'The group or rotations and reflections. This group plays a role only if the
potential energy has certain symmetries.
l'hese groups are related to the space coordinates of the constituting
particles only, i.e., their operators are orthogonal linear substitutions in the
conJiguration space (the axes in the ordinar.,v space stay rectangular)~ They
occ"hr in ordinary as well as in relativistic mechanics.
The Lorentz group which acts in space-tinle, does not leave the Schrodi.Jiger equation invariant, but only the Dirac equation .
1.2. TRANSFORMATIONS INDUCED IN FUNCTION SPACE BY THE. TRANS..
FORMATIONS IN CONFIGURATION SPACE

If Xl ~ X2, " " Xn are the coordinates of the particles, Le~, the coordinates
in the configuration space r and if s represents an operation of one of the
groups we mentioned (for instance, a rotation of the entire helium atom
aro\lnd its nucleus, or a permutation among its electrons, both of which
can be expressed in six-dimensional configuration space), then s can be
expressed by' a system of 1'1 equations
n

x; = L UikXk,

(4.1)

k==1

wher~

the matrices cr are always orthogonal in the real sense of the word.
110

Ch. 4, 1]

INVARIANT PROPERTIES

III

We will use the following shorthand

x -+

sx;

X' ::

S -1 X'

(4.1a; Ib)

where % represents the set of coordinates Xl .... x,., i.e., a point in the configuration space.
Wh..t are the repercussions of the operator s in the function space or
state-vect<;r space _1 Or according to the language of H.. Weyl: "Which
transformation induces s into the space Ul?"
Generalizing the statements of Chapter II 3.2: The operation s substitutes
in r-space the point x' = sx for the point x and at the same time carries
the value the function ..p had at the point x to the .point xl'. We may use the
follo"ring picture. The operation creates, so to speak, a redistriblltion of the
wave function distribution in r . . space as one redistributes the masses in
ordinary space if one describes the motion of a fluid. The coordinate system
being fixed, we obtain in this way a ne\v wave function ""(x) and we have
by definition
(4.2)
tfJ(x) -+ ifJ' (x) = sfjJ(x),

but, according to the statement in italics, we have for every value of x,

1/t'(x')

=:

s.jt(sx) = tjJ(x)

(4.2a)

or
(4.2b)
In the equations (4Q2) and (4.2b) the elements s of the group ?J can be
considered/as operators acting on the vectors or rays of the function space,
as mappi~gs of the space 91 on itself, or as mappings induced in the space m
by the group ~.
We can easily see that these are linear. They are unitary since we have for
two functions '" 1 and t/12'

(1/11

' " '})

==

(s!/ll Sl/J 2)

because the transformation (4.1) is equivalent with

mation of coordinates iri r ..space.
1.3. EXPRESSION OF THE INVARIANCE OF H

We suppose that the potential energy

V(x) = V(Xl' x.2' . Xn)

all

orthogonal transfor-

GENERAL APPLICA'fIONS

112

stays invariant under all operations s of

v'{x~

t.

[eh. 4, 1

Hence Vie have

... x~) == V(s;t) == V{x)

and since s is an arbitrary element from the group ! we may write accordin&
to (4.2b)
(4.3)
i.e., the operation s does not change the function V. We \. -a,all say that the
functions which satisfy the condition (4.3) are symmetrical with regard Itl
the group '6 or invariant under the operations of the group ~~
If we consider the product of two functions, for example, V and 1/1.,
we have according to (4.2b)

s[V(x) y,(%)]

= V(S-l X)I/I(S-1.x)
== sV(x) . s./I(x),

and if

.,P is invariant
s[V(x) "'(x)] = V(x)s.p(x).
0

More generally the invariance of an operator such as the Hamiltonian H,

i.e. , its invariance under the operations s of the group '{I, is expressed by
the equation

sHtJ!

:=

HstJ!

usually done in the theory of operators, we omit the object ./1 on which
theyi operate and write
sH :r= Hs.
(4.4)
Of, 8rS

Our hypothesis on the SchrCSdinger equation can be written

s(H-E)'" == (H-E)st/l.

(4.5),

1.4. CONSTANTS OF THE MonON

The equation (4.4) shows that the operator s commutes with the Hamiltonian. If we remember the quantum mechanical meanina of H(H =- ilia/at)
we see'that the operator S dOes not chanae in time. Considered as a physical
quantity (for this it has to be Hermitian) this opefa.tor is a constant of the
motion~ The same conclusion follows immediately from the equation (2.27)
of Born, Heisenberg and Jordan.
This remark is the basis for the quantum mechanical derivation of the
classical theorems about momentum. (Chapter S, 3.)

Ch. 4, 2]

113

WJGNER'S THEORE?vI

2. Wigner's Theorem
2.1. THEOREM

Equation' (4.5) shows that:

If t/J is an eigenfunction of the operator H and corresponds to the eigenvalue
E, then st{! is also an eigenjwtctiol1 of H corresponding to the same eigenvalue E .
\VIG~'Ell [1 927]..theoremQ
From this fundamental remark follow some important consequences.
First let us suppose that the Schr<sdinger problem is solved: the sequence of
eigenvalues E and eigenfunctions J.jJ is known. They form a complete
system of orthogonal functions. For the sake of simplicity we suppose that
the energy level spectrum is discrete.
(i) If E is a non-degenerate eigenvalue: then s1jJ is equal to Vi except for a
multiplicative constant J.l of modulus one. In the special case that the group (I
has only two elements, the ide;D.tity E and the elements s, the constant p,
satisfies the equation p,2 = 1 and hence must have the two values 1.
l'\his occurs in the pemlutation group of the two electrons of tile helium
atom. If jJ == + 1 the function t/J is symmetrical. If Jl :=: -1 the function ~ is
antisymmetrical. For more precision we refer to section 3.1.
ii) If Eis a degenerate eigenvalue of order ex; tJil' 1/12
1/1. are the orthogonal eigenfunctions describing the states of the energy level E . Let y" be one
of them, then st/l i is an eigenfunction corresponding to the same value E for
everyrelement s of ~ . Hence stjlt is a linear combination of the functions,
c

z: "'/eSt;
IX

SI/Ji

(4.6)

k==1

In the same way each element s of the group gives rise to a matrix S == (Slti)
ofotder the elem~nts of which are generally complex numbers. The whole

set o/the matrices or the whole seto/the traM!ormations (4.6)form a representation of the group
To prove this let

9)"

be another element of rJ

STifJ, == s(nJt,) == s 2: IPI Ttl

f= 1

=- Ie.LI ';"SlJ Tj, == z: l/I",STJkJ4O

(4.7)

It

Hence to the element ST corresponds a matrix ST.

If the functions '" f. are orthogonal and the elements s are unitary operators,
this representation is uoitary because we have

(s'c/lf. :sipk) == (y". f/I.) == 8Ii

114

reb. 4, 2

GENERAJ.J APPLICATIONS

or according to (4.6),

L (if!

j'

t/ll)S1t s!k =

j, I

Hence

L bjl S:7t S'k

=:

O;k

)1

"s*s = ()~ ik
4-i

Ii

LJ~

s*s =

I ..

the ~ fundamental functions which describe the state

corresponding to the eigenvalue E are fixed except for an arbitrary unitary ,
transformation A.
If we replace the functions '" i by linear combinations of these functions,
It is known

tha~

IX

i/I; = k=l
L "'leakl'
the matrices S become
S' == A-1SA

and the representation we obtained by a change of the axes in the "eigen

space" of the level E is equivalent to the first oneil
2.2 . GENERAL SOLUTION BY aUCCESSI\'E REDUCrJONS,

Wigner's theorem is usually presented as forrr.tulated above. It seems

suggestive to study the question in a more general way and in a certain way
from the reverse point of vtew.
Consider an arbitrary complete set of orthogonal functions X,'" Theyspan
the rdnction space and we use them to represent any operator s or H as an
infinite matrix, according to (1.25)

SX, =

L X"Sk

(4.8)

li

Th.e elements St,t , Hid are complex nUDlbers and not dependent on
time.
Let us suppose that s, T ... are the elemellts of a. group c. The matrices
S' == (S".), T' = (Tii)t . (primed in order to distinguish these matrices
from the finite matrices S ~d Tused in the first part of this section) form a
representation t' of t of infinite dimension. This representation is unitary if
the group f is unitary. The proof is the same as above (compare (4.7~
Let us reduce 16 in its irreducible elements
"

(4.9)

In this series we will find sev~ral times the same irreducible representation,
often even an infinite number of times. The reduction is obtained by a unital:y

Ch. 4, 2]

WIGN!3R~S

THEOREM

lIS

change of coordinates, Le., by a clloice of new axes lfJl, lfJ2~ which are
suited to the problem. Thus each matrix takes the box form corresponding
to (4e9)
o ....
o . ., . "
0
S1 0
0 S2 0 .. . .
T=
0 T2 o ...... ,
(4.9a)
0 S3 .... *
0
0
0 T3 . ~ .

(Tl

s=

where S 1, Tl

1#

..

"

..

..

\~

..

..

..

..

are matrices of the representation t'!J 1; S 2, T2 ...... matrices

of the representation

etc. ~ " ....

Cfi 2 ,

In other words the orthogonal functions <P, are arranged in partial sets each
of which span an invariant subspace of the function space 91.
The Hamiltonian H is, however, not an operator of the group rI. The
matrix which corresponds to this Hamiltonian in the systelD of axes can be
,,'Iitten as follows: all the matrix elements connecting the functions of one
subset to another are gathered in a unique symbol

H=

HIt

HI2

Ht3 .....

HZ1

H22

H 23

H3t

H32

H33 ..

..

..

.. ..

(4.9b)

4>

where the H fk are submatrices . Most of them are rectangular, for the different
boxes S, T. ... do not necessarily have the same dimensions.
Let
form the products HS and SH (when multiplying two matrices
the submatrices are dealt with as matrix elements).

us

H i1 .S1 ~
HnS z \.
TJ. 3Z S: ..

/II 1i S t

HS .....
-

~H11S1
8
."

31 8 1

..

"

.;r

SH

/S1 11 11
(S2 H 21

S3 R 31

\.."

S!HI2
S2 H 22

SJII 32

SlH 13 ~

... )

S2.1!23.....
S"!,H:13 I

to

""...1

!.e.
Schurs' Lemma can now be applied and it leads to the fol1ow"ing conclusion
l When Si and S" are unequivalent Bik :::::: 0
2 \Vhen ,S i is equivalent to Sit; ~ f1!k i;-; a multiple of the unit mattlX and
~te havt!

116

feh. 4, 12

GENEllAL APPUCATIONS

where a;. is a number.

It follows that H also has a box form but in a less restricted manner than

the S

matri~"

Let us replace (4.9) by:

'6 == "0!JO+nl-1 + ... n,t#,

(4.12)
where only the non-equivalent irreducible representations occur: <!J0, no
times,

.1 n,
t

times, etc. . .. The matrices (4.9a) and (4.9b) take the form
80

'0

0
<It

(4.13)

'0

"
$,

$2

lIa1 (0)

<If

"

.,. (0)

lIa. (0)

"

"

\
0

...

H-

(4.

N11

cn

1121 (1)

"21

U)

"22 (1)

.....

..

.. "

if

,.

."

Ch. 4, 2]

117

WIGNER'S THEOREM

Hence H is similarly decomposed as the matrix S of t6, i.e., in big square

matrices sit~~ed along the principal diagonal and each corresponding to an
irreducible representation ~, of f. All the matrices B Uc(1)
multipleaof
the unit matrix and their dimension is the same as the dimension of the
Q18.trices ~ I

are

HUc(l) = H;k(/)1

(4.11a)

H:'(l) being a number.

Very often (we shall see this in examples below) the big square matrices in
Sand H are infinite. The difficulties which come from this are easily solved
tcf. WEYL [1950}), since (4.13a) can be written by analogy with (4.12)
H = Ho+H t + .. H(/) .

\Jl

= 910 + Btl + tllz .

The function space 91 is decomposed in subspaces It,. These subspaces are

invariant with respect to the group C!J and the operator H at the same time.
Each of them corresponds to one of the non-equivalent irreducible representations of 'lJ.
This decomposition comes solely from the symmetry of H with respect to
the group <fl. It can be said that this decomposition is of kinematic nature.
The axes or orthogonal functions which lead to this decomposition are
represent~d by symbols with three indices fPlnm,
I labelting the big square matrices
n the smail square matrices
m the ,-rows and columns of the small square matrices.
These 'functions are not yet the Schrodinger wave functions for which H
is diagonal. This is obvious considering the steps from (4.9b) to (4. 13a). But
these functions are not yet con1pletely determined: we can still perform any
unitary transfonnation in each of the subspaccs 91 1 a.ad we shall use this
transformation to make each of the big squares matrices of (4.13a) diagonal,
i.e., the matrices obtained by replacing the submatrices H,,,(/) by the corresp.onding numbers H!k(l). The indices i, k are the m-values, (/) is the angular
momentum which characterizes a representation. Inside a box the element.a
are labelled b~ n., a label necessary because the representations I are multiple
occurring irreducible representations.
The result so far obtained can be formulated without any reference to a
particular quantum mechanical problem. We have a set of reducible matrices
s. The matrices S can be '.l/ritten as a sum of big boxes (labelled by I) each

118

GENERAL APPLICL\TIONS

[eb. 4, 2

consisting of n, small boxes of dimension m, . If there is a matrix B that

dOmmutes with all the elements of the set or group, then H can be written .....
a sum of bi'g boxes (again labelled by I) each consisting of It, small boxes.
These latter boxes each contain a different diagonal matrix of dimension
By rearranging rows and columns v/e can also say that the big boxes (labelled
by /) consist of m, small boxes of dimension nl:

m,.

It,
-------,

i
----~"'-------~

\
,

\\

!f ~'

12

lill

... f
'. i
---.----~--.-.-.~-H;.
0
H4~
(I
I
H
LI'

i
~

'. . , "

1
j

a1 '.

)1

"12"

',.

' . I

,I

-"---4-------':"+~--

.Jt l&-sufficient to choose in each subspace 1t" !:;orresponding to the rep~niiltion f I

a set of axes "'.,. which are eigenvectors of HI' These are the :.
ei&enfanctions of the Selttadinger ~quation.. The matrix. (H!t{l)) of dimension;'
", ~mes diagonal, To each of its diagonal elerrlents Enl which are the.
enersy 'evels of the system) there corresponds a subrrtatrix in H, which is a
multiple of the unit matrix E",I, of the same dimension as raJ. Thus we
obtain, after bringing roY's and columns in their original arrangement,
I

'.

,
\

'\

\\

119

Cb. 4, 2]

This matrix has clearly the same arrangement as the matrices S (4.13):
each of the big square matrices corresponding to an irreducible representation!, of ~~ntains as many small square matrices as t' contains 9i 1 i.e. nl._
The set of the elements included in a big square matrix form a term system
in the atomic case, i.e., a system with fixed I-value and different n. .values.
The combinations between the elements belonging to different tenn"systems
are subject "to selection rules which depend on the group (I (cf. chapter 5, 7).
It can be proved that they are not allowed if (9 is a symmetrical group.
It is easily seen that one can reduce H to the diagonal form (4.14) without
modifying the fonn (4.13) of the "matrices S - (see Appendix 1).
The equality of the values E", of E in each small square matrix comes from
the symmetry properties of H "ith respect to the group'll and forms an
essential degeneracy: it is impossible to separate the corresponding levels by
a perturbation W unless this one alters the symmetry of H. If W is invariant
only for a subgroup ,;tt' of 9J, the representations which were irreducible with
respect to f cease to be so with respect to .Yl'. Their reduction leads to a
separation of the levels coalescing in E". The eigenvalue spectnlm becomes
finer. 1 The energy levels are labelled with 3 indices or more instead of only
two. For example, this happens when a hydrogen atom is placed in an
external magnetic or electric field. The spherical symmetry is replaced by a
cylindrical or conical symmetry.
The preceding theory gives us valuable information on how to perform the
pertur~tion calculation. Let tX be the dimension of the representation 9J,
of 'I, ", (I == 1,2 ... ex) the eigenfunctions which form a basis for this
representation, i.e.. , which describe all the states belonging to the ex-fold
degenerate level
in the table (4.14). In order to calculate the first order
perturbation W, of this level it is generally sufficient (see Chapter 2, 7.3) to
solve a secular equation (2.47) of order a . In this equation only the matrix
elements Wit (I, k = 1,2 ... <%) resulting from the mutual interactions of the

E.,

states '" I occur.

But if we know the subgroup ,;tt' of 11 \\~hich leaves the perturbation function W invariant and the number Pof the representations resulting from the
reductions of t6, with respect to the subgroup JIf', then we know the level Elfl
is divided into only P < ex distinct levels and the order of the secular equation
is lowered from ci to p. Besides we know a priori the basis of the new repres~n...
1 The group tit' ja formed with certain elements of ., . Then the invariant subspac~ or
tho group !I remain a fortiori invariant under the elements of f but those sutotl'ac~ can be

further subdivided..

120

GENERAL APPLICATIONS

reh. 4, 2

tation i.e., the corresponding eigenfunctions. The theory of the Zeeman

effect Chapter 5 ( 4 and 9) is a simple example of this method.
For a n-e~ctron system, atom or molecule, one group of the Schrodinger
equation is the group 9',. of the permutations ofn electrons. The function H
is always invariant with respect to f/" because the electrons are physically
identical. It is impossible that the degeneracy corresponding to the group t.~"
can ever be lowered by any perturbation W. This is called exchlllJge degeneracy.
That is one of the important results in the theory.
Sometimes it happens that the levels located in several different squares
are the same. This is an accidental degeneracy which can be removed by any
perturbation, even one having the same symmetry as the system.
NOTE .. When H is invariant with respect to several groups (rotations,
permutations ... ) one can consider each group isolated or combine them in
a single group. The first method is more convenient if the elements of two
different groups commute as this makes them absolutely independent.
In every important case and particularly in the case of the permutation
groups the methods of the theory of groups give the possibiljty of building
a priori the irreducible representations. Thus the structure of the matrix H
is determined. Its decomposition into systems of elements, each system
occupying a big square in the matrix, can be determined before we calculate
the elements and the wave functiOns.
Histprically the symmetry properties of H with respect to the group of
rotatiops and retlections were used implicitly without reference to the theory
of gro'ps. But introduction of group theory allowed physicists to clear up
the p~utation problem and brought clarity and unity in the special
symmetric problem.
2.3. EQUIVALENT DESCRIPTION

Let us make a last remark, one which is almost trivial but which may
prevent misunderstandings. If we consider the equations (4.8) or (1.25)
literally we can think about the matrices S J of the irreducible representations
corresponding
,of ~ as matrices of rotation of the axes in the subspaces
to the small square matrices of (4.13). We su.ppose that the basis fUllctions
if/",. are known: i.e., we have:

-Ill

.'/',

l'

.- SljJ' 171m -- "L .tpI..'

SO)
111m' m'm

'f'lnm -+ !fJ Imn -

m'

with

(4.88)

Ch. 4, 2, 3]

ABELIAN GROtJPS

We can also suppose the axes are fixed and consider these matrices as
mappings of the/unction space on itself Let'" be a wave function expanded in
a series of orthogonal basis functions,
JjI =
I,

L
II,

Plnm ~/lnm

We have

'" = 1/1' = sljl = L !l'nm L tJllnm' ~!".

lnm

11t'

_. '\- p. t . j.
lm~' tp lnm' :;

-- ' lnm'

with
-- '"
L..J SO)
m'm f3 b,m
P'Inm' -m

(4.8b)

an equation that determines the linear transformation which the operation s

exerts on the Fourier coefficients or components of '" along the fixed axes.
The functions Y,'JJm span the invariant subspaces \}llli. The two points of view
are equivalent. We shall use either according to the case we are dealing with.
3. Abelian Groups
All the elements and all the matrices of the representations commute.
So th~ matrices can be simultaneously diagonalized (Chapter 1, 3.3).
The irreducible representations are all one-dimensional.
Following below are two simple examples.
3.1. PERMUTATIONS OF TWO OBlECfS

This group !?'* contains only two elements E == (1) and s == (1 2) with
the only rule of multiplication S2 = E. In order to represent this group let
the element s be represented by a one-dimensional matrix such that the '
number S satisfies the equation S2 = 1.. Thus one obtains two irreducible
re~ntations of 9'2: E, S == + 1 and E = 1, S = -1.
It follows that in the space of the states for an atom with two .eleotrC)Ds
-(helium) tho matrix S of the equation (4.13) is decomposed in only hVo big
square mafnces: In the first one the number 1 occurs along the diagonal
(obviously they are infinite in number). In the second square matrix the
number -1 occurs in the same way. Once H has been put in the form (4.14)
we obtain

122

GENERAL APPLICATIONS

/100

010

II

001

fCb. 4. 3

II

.I

s=

----

(4.15)

....

0
0
o .., 1

-1

0
o -1

1
i

II

"./

I
!

/E1O
0
o E,.o

H=

I
!

t
I
!
I

. I
I

.~

Ell

Ell

(4. 15a)

Hence, there exist two systems of eigenvalues corresponding to the two sys...
teDl.$' of eigenfunctions. One of them satisfies, after (4.1 S), the equation
sy,lt =
these are the symmetric/unctions. The second satisfies St/lk := - "',,'
these are the antisymmetricfunctions. (We delete: with respect to the group
of permutations.)

"'i'

3.2. PLANE ROTATIONS (AROUND A FIXED AXIS) (GROUP gl)

The operations commute, since any two arbitrary rotations of angle q/

and cp satisfy cp' +q> == qJ+<p'~l Hence the irreducible representations are
one-dimensional. To each element sf' that is a rotation through an angle

or

1 When cp is incommensurable with 1t we can consider that all the elements

the group
are built by iteration of the rotation ffJ to any dearee of approximation (two angles whicb
differ by a multiple of 2n are conside~d identical). The group ~t can be consid.cred from
this point of view as a cy~li{; grollp ,of infinite order (closed group with one parameterj.

Ch~

3, 4]

123

ABELIAN GROUPS

cp, there corresponds a one-dimensional matrix, i.e . , a number X( <p) and

because
we have

X(rp)X(cp') = X(,,+cp').

Since X(rp) is a continuous function of cp and since So is the identity X(O) == 1.

We shall suppose (this is not the only possible case) that the representation
is faithful and unambiguous, Le*,
X(2n) = X(O)

Let us put: ..(<p) =

We have

=:

I.

eta(cp).

),(0) = 0,

A(cp+cp')

= l(cp)+.A.(cpf).

A functional equation with the solution A = mq> with e 2nml = 1, \vhere m is

a positive or negative integer. Finally:

+ 1, :-r 2 .

m = 0,

It

..

An atom in an external magnetic field has the symmetry of the group ~241
Every matrix S(cp), representing in function space a rotation q> around the
field, is decomposed acoording to the scheme (4'113) \\ith

St(cp) == e ifJ ;

So({J) = 1;
S~(<p)

= e-i'P;

S;(~)

,S2,(q

= e2i ., .... ;

= e- 2f. ;

1\

..

The elements can be ordered according to the value of m, the magnetic

quantum number but this classification is generally insufficient. The study of

the complete equation can only tell us how each value of rtl occurs in the
final representation.
Let us, however, consider the case of a single electron with spheri~
coordin~tes f, qJ,

8 . The theory gives us immediately information about the

wave function t/I'III(r, cp, 8) corresponding to the invariant subspace labelled
by the number m. These functions satisfy the equation

s.l/Im(r, lfJo, 8) = t/lrn(r; q>o +tp, 8)

In particular

(/'0

:=

e'''''y,,,.(r, tpo, [J).

== 0

l/lm(r., rp, 8)

= e1mVI!rlr, 0, 9)

:=

e1ffUPt/lm(r, 8}.

124

reb. 4, 4

GENERAL APPLICATIONS

4. Non-AbeUan Groups. Rotations and Reflections in a Plaae

This group ~; is obtained by adding to the rotations around an axis the

reflections through planes containing this axis. This is the collection of
diatomic molecules ,\\"hich have the symmetry of the cone or the cylinder.
Let us call the reflection through an arbitrary plane containing the axis T,
and the rotations s. The rotations commute with each other but not with T.
We can see this immediately from Fig. 4.1.

~2

2'

I ----- ,/~

I .I
.
I

II

!,

I
I
I

I
I

If

'

I!!
I!I
i

I:

i
f

Fjg. 4.,1. Illustration of the non-commutativity of rotation and retlection operation in a

system of cylindrical symmetry.
>

One call go from the position 1 to the position 3 in two different ways.
A rotation sip (1 to 2) and a reflection T (2 to 3), or a retlection T (1 to 2') and
then a rotation
(2' to 3). Thus we obtain

s_.

where the succession of the operation must be read from right to left as
usual. First~let us consider the subgroup P)2 of ~;. It induces in the function
space a representation which we have studied in the preceding section:
the invariant subspaces are one-dimensional and are spanned by a set of
basis functions t/J m' such that s(,p 1/1m = e iPJrrp t/I m. Let us suppose m > O.

Ch. 4, 4]

NON-ABELIAN GROUPS

125

Likewise 1/1_ m is characterized by:

~

stpV'-m
.1,

_
-

-1""1.1,
'i'-m

But the equation (4.16) gives

TSfPt/!m

TeimfPif.tm

= ehnfPTt/lm = S_tp nfrm

or taking into account the last of the preceding equations

Tt/I". = t/I-m.

(4.17)

As could be expected the operation T leads to a coarser reduction of our

representation.
It is necessary to combine every function t/I". with every function t/J-m
in ord~ to build a two-dimensional subspace of the space of the states
invariant with respect to every element of the group .@~. In this subspace
the element Scp is represented by the diagonal matrix
(4.18)

and the element

by the non-diagonal matrix which results from (4.17)

(4.18a)
Then the representations and the levels are classified according to the
absolut~ values Iml, +m and -In being associated in the same representation. The eigenvalues for which Imf = 0, 1, 2~ ..... are labelled by the symbols
E, 1t, L1:t . . .. l~he state m = 0 is a particular case:
s"t/lo =
so(q = 1, as (4.16) gives us no more information To is
determined by the condition T2 = E, i.e., Ti = 1. Then we can take at will

"'0'

To =

+ 1.

There remain two one-dimensional representations which are distinguished

by their property with respect to reflection 1 and -1 (cf. Chapter 5, 8).
Finally we obtain the follo\ving irreducible representations and systems of
eigenvalues

126

GENERAL APPLICATIONS

[Ch. 4, 4

Appendix 4.1

We will si:ve some examples: Consider the case of a two-dimensional

irreducible representation ?J.2 which occurs m-times in f. This representation is part of a 2m-dimensional subspace 81 2m which is subdivided into m
irreducible 2-dimensional spaces. The large boxes of the matrices H and S
which correspond to
can be represented according to (4.13) and (4. 13a)
in the form:

"2m

S11

S21

Sui

8 22 1
I

S2

==

,..

S11

S121

S21

S22t!

f
1

H2

H21
0

=:t

H~2

H~l

0
0 . H~m
H;".
H~2 ~ 0

0
H;l

H;"l

H~l

. . .

that is t~ say that, if we designate the 2m basis functions by fPt, CPt, cpz, "2
lfJ""qJ:', the matrices S2 transform as:

= Sll fPl +S12 cp~ .

Sf/I; = S21 <P1 + S22 <pi '
S<Pl

H2 q>1 ~ H~l fIJI +H~2tp2 + . H~m<P".

H,alfJl := H~lqJl +H;2(fJZ+ ..... H~mCPm
. . ... . . . . . . . " . .
H 2 q> 1 == H'11"'1 + H'12 <P2 + .. H'1m tp".
H 2 ({J2' = H"21 fi'1 + . . H'2m({Jm

(4.A2)

"

. . . .

..

. . .

..

..

We try to find the eigenfunctions of the Hamiltonian H2 by putting

1/1 == ex 1 <1'1 + Cl2lp2

ifJl

z=

0Cm 'Pm

131 'Pi + P2 cP; P",({J:n.

127

NON-ABELIAN GROUPS

Ch. 4. 4]

As a result of the equations (4.Al) and (4.A2), we have

st/l == S 11 '" + S 12l/1' ;

StjI' =- S 21 '" + S 21. !/I'.
H"t/! = (cx I H;t+2 H21+. (t.H;"lWl
+(11 H~l +a:2H;2 + ... am H:" 2 )lP2 + ...
I

H 1 "" = (Pt H; 1 +

Pm H :"1)q> ~ +

+(P1H~2+

. fJ.H~2)<P;+ ....

We will reduce H2 to its diagonal form. and we will have H2Y, - E';;
H) t/!' == Et/I' if we satisfy the equations
cx1H t1 +a2H~1.. ".H:'1 == Ea 1
(XIH~2+(X2H22 tX".H~2 == Ea2

. . . . . . .

.. .

(4.A3)

and a similar set of equations in /J, with the same coefficleil~ H; 1 t

We have to reduce in both cases the same secular problem in m dimensions
(and not in 2m dimensions). The equation
H;l

..

H:"1

H 12 -E

H:" 2

. . .

..

. . . . .

==0

has in seneral m distinct roots. As a result we have a set of ", 90nstants

(%1 .,. given by (4.A3) and m constants PI ... Pm which are eqUal
the
first set.

to

CHAPTER. 5

ROTATIONS IN 3-DIMENSIONAL SPACE: GROUP

1?}3

1. Spherieal Hanaoales aad Representation of the Rotation Group

We will follow the historical order. Let us consider an atom with only
one electron in a central field of spherical symmetry (Bohr-Sommerfeld).
We suppose that the origin of the polar coordinates Tt 8, tp is at the nucleus.
The potential energy is att arbitrary function of the distance r. The Schrodinger equation and the Hamiltonian H are invariant with respect to the
group Pl3 of the 3-dimensional rotations about arbitrary axes passing
through the nucleus.
According to the Wiper theorem the eigenfunctions of this Hamiltonian
can be classified in a number of different systems. Each of them acts as a
coordinate system in function space and constitutes the basis of an irreducible
representation of the group ~3. In other words, each of these eigenfunctions
bcpomes after rotation a linear combination of functions of the same system
~ the matrices of these transformations form an irreducible representation
or; the group.
,/We shall show that the solutions of the hydrogen atom problem as
.c8Iculated by Schrodinger have this property_ These solutions can be written
as follows:
(5.1)
!

with
y,(m)

= eimfP(sin

8)-m p~m)(cos 8),

(S.1a)

where Y,(III) is the usual notation for the Laplace spherical harmonics.
The associated Legendre polynomial is defined by
d(l-m)

Pfm)(z)

= di -

(1- Z2).

(5.1h)

'
The Laplace spherical harmonic is homogeneous in cos 8 and sin 8. Its
degree I, the azimutluzl quantum number or the quantum number of angular
128

<"'h. 5, 1 ]

SPHERICAL HARMONICS

momentum is a positive integer: I = 0, 1, 2 ... Given a certain value l.

. after (5.1b), (~-m) can only have integer values between 0 and 21. Hence the
magnetic quantum number m can assume only the (2/+ 1) values m ==
-I, -1+ 1 . 1-1, 1; to a given value I there correspond (21+ 1) independent spherical harmonics Y,(JR) all of degree I in cos 8 and
8.
The third quantum number n is always a positive integer which can indicate either the radial quantum number of the Bohr-Sommerfeld quantum.
theory (and then assumes the values nr = 0, 1,2 ... ) or else the total quaatum number n == 1+ 1, 1+2 . 1+ n, + 1 ... , which is the usual choice.
The energy levels E(n, /) are independent of m. They are {21+ I)-fold
degenerate and one can choose as corresponding eigenfunctiODl (21+ 1)
independent functions of the form

sm

!",(r) Y,( lp, 8)

Y,

being a linear combination of the

Y,<"').

Thus the eigenfunctions t/I are products of two factors, the first of which
I.,(r) only depends 'on the particular interaction between the electron and the
remaining part of the atom; the secon9 one exhibits the symmetry properties
of the- operator H with respect to the group ~ 3.
Since /",(r) is left invariant under an arbitrary rotation s of the system
about the nucleus we have
"'~~) -+ s"'~~) = !II'( r)s y/"')( qJ, 8),

but st/l~~ is also an eigenfunction of the energy level Ell" hence sYf-) is a
linear combination of the (2/+1) functions Y,(lII)(m = -1, ..., +/):
p=+l

S y,(m)( qJ,

8) =

2:

y,(P) S~!

(5.2)

p=-I

where 5, = (s!2) is a matrix of dimension (2/+ 1).

The Yf m ), in the space of functions f( qJ, 8), span a (21+ 1)-dimensional
subspace which is invariant with respect to the operations of the group ~3.
The equation (5.2) is a known property of spherical harmonics. It follows
immediately from the invariance of the degree 1 of an homogeneous polynomial in (x, y, z) 1 by a rotation of these coordinate axes. Moreover we
1

As x + iy

= T sin I) eiql, Z = r cos I) every homogeneous polynomial of degree

I in

x, y, z has the form: r' Yz(<p, if). Tbe YI("') are basis polynomials, homogeneous in x, y, z
and defined on the unit sphere by' the condition that the difference between the exponents

of (x+iy) and (x-iy) be equal to m (cf. (S.la. The expansion or an arbitrary polynomial
n
Y, in a linear combination of (21+ 1) polynomials
) follows from this definition.

yt

130

R,-)1"A,!'fONS iN 3DIMENSIONAL SPACE

[Ch~

S, 1

conclude from tb.e wen-known orthogonality properties of spherical

harmonics that ttlese matrices S I are unitary.
The function: space 81 is decomposed in subspaces Ut, invariant with respect
to the group ~ 3. Each of t.hese invariant subspaces corresponds to a representation !'J, of the group ~3 consisting of matrices Sf of dimension
(21 + 1). It can> be shown that these representations are irreducible. This is
almost obvious as a result of the theory of spherical harmonics. This representation occurs an infinite number of times because the radial or total quantum number n can increase tv + 00, 1 remaining the same. The spaces Bt,
filling the big square matrices of the formulae (4.13) and (4.14) are of
infinite dimension. They can be decomposed in an infinity of subspaces
It(n, 1) with dimension (21 + 1) each corresponding to an. energy level
E(n, I) and with one "copy" of the representation fj 1 filling each one of the
small square matrices .
In particular if the totation is around the axis Oz through an angle
OJ~(rp' == scp
tp+w:rJ, it follows from (4.2b), (5.2) and (5.1a) that the
matrix S(cor:) takes the diagonal fonn

exp [ -ilwg]

o
.

"

..

0
exp [ -i(l-l)coz ]

&

(5.3)

...

All the~ representations are of odd degree (U + 1). One can wonder why
this is so land if no other exists . According to Chapter 4, 3.2, concerning
the plane' rotations PAz, the exponents m of the exponentials exp (imt:z)
have to be integers in order to obtain an unambiguous representation for all
values .of the parameter 00%. The matrices S l are the only ones which give a
faithful representation. But if we give up the condition of faithfulness for
arbitrary values of the parameters and maintain this restriction only in the
vicinity of the itkmlty w. == 0, it is possible to find other matrices.
~ommerfeld has sho~ in the old quantum theory that internal
quantum nu.!!'Nrsjwhich can be hal/integers" must be introduced in order to
deacribe Certain spectra. It is known that these half integer numbers COLlle
from the spin (cf. 6.2J~ By analogy with (5.3), 'We find for the rotations fi)z

~xp.[~ij~J. e~p.[-.i~.-l~~]:.:: . . ~ . . )
o

exp (ija>,] I

(S.3a)

It{)1"ATION GROUP

131

withj = 0, -i, 1, t ..... All the exponents of the matrix are either integers or
b~lf integers. We shall add to the matrices of odd dimension in whichj is
integer, matrices of even dimension withj = t(2p + 1). But in theJatter, ~ere
is not a one-ta-one correspondence between the representation of the rotation
group and the abstract elements of this group . When ro. is increased by 21t
the matrix elements in (S.3a.) will reverse their signs because the factor
j = t(2p+ 1) :;-.: t+integer
gives rise to a -1 in the matri~ element. To each angle OJ% there are two
corresponding matrices SJ(wz ) and -Sj(w;:) in these representations of even
degree. They are two-valued representations. Between the representation
and the group ~3 there exists only a homomorphism. Both matrices 1 and
-1 correspond to an angle zero or 21t.
2. Rotation Group and Ifwo-Dimeosional Unitary Group
2.L RELATION BETWEEN THE ROTATION OROLrp AND THE UNITARY
GROUP.

E" Cartan then H. Weyl h.ave shown how to build "a priori" all of these
representations 1 The group ~3 of the rotations around a center is a 3
parameter group.. T\vo of the parameters fix. the direction of the axis, the
thlrd one the rotation angle. Hence we have the problem of constructing a
correspondence between rotations in three dimensions and a unitary three
parameter group_
The Simplest unitary group is the unitary unimodular group '&f z with two
complet variables. The transformations (j of this group can be written

f e' = + prJ
U -+ t rtf = -P*I;+a.*rt,.

(5.4)2

Ct". B. CAltTAN, thesis (1894]: and his papers; 'The projective groups which leave no
plane manifold invariant (irreducibles)"', [1913], and (1914}., See also CAJlTAN [1938].
Cartan& work bas been resumed and completed by another method by H. WEn
[1925} on certain points. These studies are very general and they aim at building all the
possible irreducible linear groups "a priori" and at finding their structure (compare the
footnote on p. J 7 and Appendix IIIj
t A mapping of a two ...dim~nsional complex space on ItMlf can be written as:
1

e' == r:tq-i-PYi,

r( = y~+~.
F or simplicity 'Are i.t.1I'lpcse the condition that this will be a unimodular tt'(lnSj:?1'matian, i.e.
11

-:jI.

ti.o- fl,.-

L
rhe matrices of the adjoint and inverse trao..cormations a'fe:
..,
:(1;* '1J$\)
(l ~
~./)

\rilli

(Footnote corJilfURd on

tiff!

next pa:?e~)

=:;

132

[eh. 5, 1

ROTATIONS IN 3-DIMENSIONAL SPACE

The determinant of

(f

is 1; this condition

aa.* + pp. = 1

(S.4a)

which coUPles the real and imaginary parts of a. and fJ and lowers the number
of independent parameters from four to three.
By a stereop,phic projection we establish a correspondence between
every rotation s and a transformation (J of the type (5.4), with certain complex
coefficients ex and p.
Let x, y, z be the coordinates of a point P on the unit sphere:
r+r+z2 == 1. Let x == 0, y = 0, Z = -1 be the coordinates of the
south pole S which we use as center of projection. The plane of projection
will be the equatoriaf ~e and the coordinates of pi, the projection ofP,
wiD be called x' and y (compare Fig. S.l).

Fi,. 5.1. Stereographic projection of the point Pinto P'.

I Q order that

(J

be unitary

a == 0-1
a=

from which follows that~

+cx*,

?' =

-p.

i.e.
(1

(.~P*

!.).

A unitary but not unimodular transformation would only satisfy:

Da.*, f' = - DP*
mod D == 1, (lcx* +PP'* =: 1
where D is the determinant of the coefficients. "fhis would be a 4 parameter transformation.

f5

:=

Ch. S, 2]

ROTATION GROUP

Let US put X +iy' =We have

SP'
p ....

SP

13~

e.
1

y'

X'

y == 1+z -

== x -

1+,,

and

p=--

X-IY

,.

+,,*'
.

1/1+,,*

Y2(1+z)

x+iy == x-iy ...

x+iy ==

'

2,*

1+,,2{

1-,,

Z=--

1+"*
1
or :finally using complex. homogeneous coordinates ~ and ", such that
t = tile with the supplementary condition:
~,. +"". = 1

we obtain

x+iy == 2tt~.,
x == ,,~. + ~,,* t

x-iy ==

(5.5)

2,,,

Y == - i(,,~ - ~,,*)

z = ,e-",,~
To each pair of numben , and " there corresponds a point on the 1lDit
sphere, since as a result of (5.6), (5.5) is equivalent to x"+r+r - 1.
Eve~ unitary transformation leaves (5.5) invariant and coD8eC)uently
transforms a point on the sphere into an other point on the sphere. It ileMY
to see ~t such a tran~formation leaves the anates between two 1iDeI from 0
to two different points C invariant (because the transformation is 1IDitary
and unimodular). Hence the transformation is a rotation.
To each transformation (/ of the type (5.4) there corresponds a rotation
s of ~3' The converse statement is not quite the same for if we change the
sign of and /1: i.e, ~ = -ex, P' == - /1, the signs of ~' and ,,' chango, but
after (5.6) x', J1~, %' remain invariant. Consequently to each rotation s of ~e
group f!}3 there correspond two transformations of the unitary lJnimodular
group ~2: +0' and -tl. Since the transformations are linear there corresponds to the product S2S1 of two successive rotations a product (12(11' The

group "'2 is a twO-dImensional representation 0/ the rotation group ~3.

A rotation about Oz leaves the difference "*-'1'1*
Z and the sum
::II

When the point P is on a sphere of radius r :;C I it is sufficient to multiply these formulae by the Teal constant ,." The conclusions do not change.
1

134

ROTATIONS IN 3-DIMENSIONAL SPACE

,~. +11" invariant. Hence we have

e' ,'. = ee*, 11' ,,'. =

account (5.4) and (S.4a) we find that

c;' = eiCP ,
where

q>

act* =

(Ch. S, 2

"". i.e. taking into

Of,

and 11' = e -itp f/,

is an arbitrary quantity. We see, however, from

x' +iy' = 2t(e'* = e- 2itp2 tT<!* = e- 2itp(x+iy)

that 2cp

= - OJz ,

the angle of rotation, i.e"

00(0 ) =
%

(exp (-tiro;;)
0

0.

exp (tICO,)

).

(5.7)

A similar calculation (compare problem 5.2) gives the matrix which

represents a rotation ro, about the axis Oy

o{wy )

_ (cos troy i sin tro,)

i sin 1<oJ' cos 1-wYI

(5.7a)

and a matrix of the same type a(wx }. An arbitrary rotation s (compare

Fig. 5.2) defined by the Euler angles qJ, 9 and 1/1 is the product of 3 rotations:

~-----------------~7---Y
I

-Fig. 5.2. En'erian angles.

a rotationo.s(qJz) through an angle qJ around the axis Oz, then a rotation ;

s(8,,) around the new axis Oy, finally s{rJ!z) around the new axis Oz. Then we
have s(rp, 8, Vi) = s(t/!z) s(8,) s{lP z ) and the corresponding matrix can be
written

135

I{OTATION GROUP

Ch. 5, 2]

(1( qJ, 8, fjI) = a( t/I z) 0-(8y ) u( qJz)

t8)

!i(cp+tJ!)) cos -!8 +i exp (+ ti(lP-\f!) sin

\ i exp (- -li(<p-ifJ)) sin -1-9
exp (+ -li(tp+.p)) cos 18

= {exp (-

(5.7b)

it has the form required by (5 . 4).

An alternate way to introduce the spinor variables and their transformation
properties is to use the so-called "null-vector method" used by Kramers.
Q)ee KR.A1t{ERS [1937] and BRINKJvlAN [1956].)
2.2. THE REPRESENTATIONS OF THE GROUP 'Wt AS REPRESENTATIONS OF
THE GROUP ~8

The preceding method gives us at once an infinite number of representations of the rotation group, because all the representations of the group fl2
are obviously representations of the group ~3 also. They are easy to construct.
Let us form tensors in the two-dimensional unitary space 11. A symmetric
tensor of rank v has v+ 1 components

e,

(5.8)

(compare Chapter 1, 1.3).

By a transfornlation (f (5.4) on

and 11 we obtain
k=o

en '-il1 li,

(cte-f-{Jrr)v-i( -JB*e-r4Ct*11)i

= L SJ),,,-lcttk.

(S.8a)

k=O

The components of a tensor of order v undergo a linear transformation.

The'transformation matrix is S<fJ) = (S[:. In the space of all the functions
of x, y, Z, to which they are related by (5.6), they span a subspace of dirrlension (f) + 1). This subspace remains invariant under the rotations s. The
transformation (5.8a.) is not yet unitary. It becomes unitary if one uses the
variables (5.8b) instead of (5~8)
qk

efJ-k"k

-,

= 0, 1, 2 .. - v

(S.Sb)

.J(v-k)!k!
because if (5.5) is used one finds

~
*
~ qkqk =
k=O

"L...J (ee*),,-k(fJ'1*)k
.. : - - - - - - - - =
k

(v-k)!k!

1 (l!;;*
*)V = mvarlant.
.
,
,S~ +tl'1
vI
--

(5.9)

136

ROTATIONS IN 3-DIMENSIONAL SPACE

[Ch. 5,12

The matrices S(u> can be multiplied among themselves, like the m~trices (1.
Then we obtain an infinite number of representations of the rotation group,
each arisin,, from a tensor of order v = 0 1 , 1, 2 ....
The customary notations in quantum mechanics lead us to put v = 2j
and to label these representations by the symbols ~J' j =: 0, -iJ 1, i ..
Consequently, their dimension is 2j+ 1. It is convenient to call j-k := m.
The variables q of the equation (5.8b) take then the symmetrical form
II

.
q~)

,i+m"J-m

v'(j.+- m)!(j -

m =j,j-l, ... -j.

(S.8c)

m)!

The group t1IJ 2, from which we started, has as a basis the vectors of the
two-dimensional space " 'I, with v = 1 or j = t. These vectors are usually
called spinors. It is the same as the representation D t which is double valued
as we have seen in 2.1. This space is usually called spinor-space.
By taking the real and imaginary parts of the variables " " and the
coefficients a, fJ explicitly into account in order to have 4 real equations
equivalent to (5.4), it is easily proved that if If; ~', ,,', satisfy this system,
they are satisfied also by

e,

X =

A~

X' = A"

+ fl'1*,
+ JUI'. ,

-p,~.

+l'l
Y' == - Jl~/. + AI7'

where .A. and Jl are arbitrary coefficients. In ordel to have XX*+ YY. = 1,
we mU$t put: AAo + J.LJl* = 1.
The above transforma~on, which is very peculiar since it connects X and Y
simulta,neously with and '1 and their conjugates, shows the extent to which t
the baSes of the unitary unimodular group are undetermined.
Intparticular if we put A = 0, p, = 1 we obtain a' pair of variables (,,*, - ~.)
which undergo the same transformation as (" ,,). This can be immediatelyc,
veTified. Consequently after choosing x+iy ==
x-iy =
Z =~
as the basis of the rotation group ~3 in the 3-dimensionalconfiguration space, these expressions transform respectively as _,.,2" 1 and
i.e. as the three components of the second order tensor in " " or
according to (S.Se) as qf!l, -q1 1 ) and (1/ J'i)qb1 ). Thus the group ~3 can:
be identified with the representation D 1
In the same way, the representation D j where j == I is an integer is nothing}
but the representation D, (obtained in section 1) the basis of which are the:

ee*-'1"
- e",

2,,,,

2rr'.'

To the value v = 0 corresponds the identical representation; to

fJ

1, the group

~t,

Ch. 5, 2, 3]

137

INFINITESIMAL TRANSFORMATIONS

(2/+ 1) spherical harmonics Yf m ). For a more rigorous proof of this state-

ment see Appendix I. These functions are homogeneous polynomials of

degree 1 in (~+iy), (x-i.,v) and z on the unit sphere i.e. they are sums of
terms like:
A6t(l1~*Y'{ <!'1*)T(,~* _ttr/,")l-a-T or A<r1{

,'-ctrt 17 e*l--r: - el1*l-D'l1'-'te*d).

tIC'f

(J

Let us put m = 1:-0'. The above remark shows that these terms are transformed like
or like q~). Thus the functions Yl m ) are independent
linear combinations of (21+ 1) functions of ~*) '1, fl which, though .not
identical with the q~), transform in the same way. They are the basis of a
linear group which is equivalent to D J' j = I.
These representations are faithful. The others for which j = t(21 + 1) are
double valued.
They are all irreducible and no others exist (this will be proved in note II
which should be read after the following section) .
One final remark: in the space IlJ of the representation gjJ the rotations
O)~ around Oz induce the following transformations, which can immediately
be deduced from (5.7) and (S.8e).

el+m,,'--

e,

q:" = exp ( -imcoz)q"..

The corresponding matrices 5<})(oo%) have the form (S.3a) as was postulated
before.
In the following sections we shall assume that the representations fA)
where j: is half integer can playa role in physics. This will be justified in 6 .

3. infinitesimal Transformations and Angular Momentum

Th~

rotation group is a continuous group. It contains operations which

differ as little as wanted from the identity. These infinitely small differences
were termed by Sophus Lie Infinitesimal transformations. The diJferential
operators which give rise to these transformations and which for the rotations are the components of the angular momentum, are the quantum
analogues of classical quantities. This is true for all continuous groups .
The quantum theory gives a physical meaning to discontinuous operators,
such as reflections and permutations, which are considered as pure abstractions by classical theory.
3.1. INFINITESIMAL TRANSFORMATIONS OF A CONTINUOUS GROUP

Let us consider for example the rotations. An zlljinltesimal rotation, an

element of a 3 parameter group, is defined by the three components of an

138

reb. 5,13

ROTA'II0NS IN 3-DIMENSIONAL SPACE

axial vector d{J. ;::: (.t) dt lying along the rotation axis. The length of thiJ
vector is eqqAl tQ tJl~ rptation angle dB, dt is an auxiliary parameter (the time
in kinewatics), (J) similar to an angular velocity with components OJ,x, 0)"
co% (ro=-:'" = d8 x /dt . .. ). The transformation of the coordinates x, y, z by
tbU {QtatiQn is linear and homogeneous in x, y, z according to the kno'wn

tQrmulae
~

:;:;:

(.(}1~-lO%y)dt,

etc.... , x'

= x-w::ydt+w,zdt, etc. . ...

(5.10)

A. finite rotation is the integral of a continuous succession of infinitesimal

trAnsformations (5.10).
:Mor~ generally, let us consider a continuous group with r parameters
Sl, s~, .. . SF) which are all zero for the identical transformation. In the
vieJ.uity of the identity every element of the group is defined if the values of
these p8.(a.meters are given and conversely. Let n be the dimension of the
space in which the group operates, either configuration space or representa;.
tion space. A transformation s is defined by the values 91, 82, s,. of the
parameters and by the transformation formulae

( 5.11)
the lp i are supposed to be continuous and differentiable with F:'~fACt to the s;..
The transformations (5.11) form a group when two succe'\sive transformations
,I,

s :x

---+

x;

and

T ~ x; ~~ x~ I = lfJ i( x~ . . ~ x~;

t1

t,.)

give rise to a unique transformation

U

= TS: X, ~ x;'

f.{Jt(X 1 , .... XII; Ul' .....

U,.),

where the u). are, in the vicinity of zero, continuous and differentiable

functions of the variables sand t:

(5.12)
The. nature of the continuous group, that is its "multiplication rules n , are

contained in these functions.

An infinitesimal transformation is defined by the following equations
dx,
if we put dsl =

(0 A. ds,

(~lfJi)

.~=1

AS;.,

<!sA
0

ds being an auxiliary infinitesimal small coefficient,

Ch. 5, 3 ]

INFINITESIMAL TRANSFORM.A.TIONS

139

as was dt in (5.10),

= '"
'-' (OlfJi)
-

dx,

).

(5.13)

roAds.

AS). 0

The symbol ( )0 means the s)., must be taken as zero in the derivatives.
It often happens, and this is the case for the rotations (5.10), that the
group is not defined by its finite equations (5.11) and (5.12) but by its infinitesimal transformations. The infinitesimal transformations are then defined
by the following equations.

(S.l3a)
'where the (4).,) depend only on the variables x. These functions have of course
to fulfil certain integrability conditions.
Two successive infinitesimal transformations are equivalent to a unique
transformation described by equations which are obtairied in the first order
by replacing a>A. in (S.13a) by the sum (COl + co,). An arbitrary infinitesimal
transformation of a given group is then a linear combination of r basis
transformations with arbitrary coefficients Q).t; each of them is defined by n

functions
(J.)(

(Xi

Xl , Xn ,

= 1, 2, ...' n;

A = 1, 2, ... r.

These r transformations i.e. the rectangular matrix which has as elements

the fu'nctions <x~l) determine the nature of the group completely.
The integrability conditions mentioned above are that the dxi must be
exact differentials and hence they contain the second derivatives of the <Pi
i.e. infinitesimal elements of the second order.
This can be stated in the following form:
Two transformations, sand T, infinitely near the identity do not generally
commute; the difference between ST and TS is then of the second order.
There must exist a third transformation in the same group, r(s, T), which
connects these two products in such a \vay that:
(5.14)

r is called the commutator of sand T; this operation differs only from the
identity by an infinitesimal transformation. It must IJe a linear combination

140

ROTATIONS IN 3-DIMBNSIONAL SPACE

[Ch. Sf 3

0/

the same basis transformations as the infinitesimal transformation oj

s and T.l
3.2. LINEAR SUBSTITUTIONS

This is the only type of substitution we are interested in. The equations
(5.10) of the infinitesimal rotations are linear and the same thing is by
definition true for any representation of an arbitrary group. Hence we
suppose in (S.13a) that the functions ~A) are linear and we write

dX i =

L
L
(x~)xkw.tds == L ailxk ds
k-t 4=1
t

(5.15)

with the abridged notation

We will put: A == (aile)' A(J.) == X~: and call x the vector with components
Xi and use the conventions of Chapter 1.

dx = A%da;

x' == x+dx

= (1 + Ada)x.

(5.15a)

Each infinitesimal transformation s is defined by a matrix A, a linear

combination with the coefficients tnl of r basis matrices ACA.). Let sand T be
two successive transformations. We have
T :

x' ..... x"

:=

(1 + Bdt):t'

add for the resulting transformation

TS :

-+

x" = (1 + Bdt)(l + Ads)x

= (1 + Ads+ Bdt+ BAdsdt)x.

In the same way

ST:

(5.14) shows that the

= (1+Ads+8dt+ABdsdt)x
commutator r has the form

-+

x'"

r(S,T) = l+Cdsdt,

C :: AS-SA,

(5.16)

As a classical example which illustrates the theory of non..holonomic systems in

mec~~ics we consider a spbere rolling without sliding on an horizontal plane. Its infinitesimal displacements have three degrees of freedom: the two anales which ftx the
rotation axis and the rotation angle. But these displacements are notintograble. Its finite
displacements form a five parameter group: these parameters are the coordinates x and y
of its center and the three rotation parameters.
1

Ch. 5, 31

INFINITESIMAL TRANSFORMATIONS

141

C must be a linellr combinatton of the same basis mtIIrica as A and 8 are. 1

Let us verify these results for the rotation group PJ3'. Let us consider the
three infinitesia:sal basis operations, the elementary fDtations P. P7J P. around
the three Cartesian axes. We obtain them by assuming that we have in (S.10)
successively:

1)
2)
3)
According to (S.lSa)
amount dl1:

d9~

(1)7

==

d8% == .w.dt,
they displace a vector

de == R# "d8~
d., == R, dB,
with the transformation matrices (cr. 5.10)

0)
0
(

0
R;:= 0 0 -1
o 1 0

R, ==

(J)~

== 0
CDs = COx == 0
COx == ro, == 0 .

== co%dt,
dB, == co,dt,

in ordinary space by an
(S.lSb)

0
-1
0)
(o

( 0 0 1)
0 0 0
-1 0 0

Rz==lOO.

(S.17)

0 0

The commutators of these three transformations considered in pairs

are obtained from (5.16), and the multiplication rules for matrices. If we

write
one finds:
C;q

==

R~R,-R,R~

:II:

Rs;;

R,Rs-R.R, ==

R;x

RzRx-RxR:e == Ry

(5.17a)

L:c == ihRx' etc.

(5.18)

If we put.
we find again the commutation rules (2.20) for angular momenta. The
above calculation is, as a matter of fact only a particular case of the calcula. .
tion in Chapter IT, 3, because (S.lSb) is the expression of (2.19) when the
1/1'. are linear functions of x, y, z.
If one conaiders the , infinitesimal basis transformations A(A), A(p.) ... " (as will be
done later for rotations) one must havo:
1

,.

eClp) . . A(A.) A(p)-ACp.) A().)

==

I:

cJ.p., y A(v)

y-l

tho cJ..I'." being constants. S. Lie has proved that this condition is suiflcienI in order that
the infinitesimal transformatiotlS considered give rise to a Stoup. The structure of this
group is determined by tho canstants cA,.u, y. These theorems are the basis of Cattan's

work.

142

ROTATIONS IN 3..DIMENSIONAL SPACE

[eb. 5, 3

3.3. REPRESENTATIONS OF THE ROTATION GROUP. MATRICES FOR THE

ANGULAR MOMENTA

Let us ~pnsider the irreducible representations D J of the gtoup 9}3'

Each infinitesimal rotation
p : Vi

= (1 + Rd8)v

induces in the space of representations 9lJ , in which the cODlponents of the

unitary vectors

qJ

q;j), a transformation

are the

p.;' : tp'

= (1 + R()d8)cp

(5.19)

or going over to algebraic language (cf. (5.15) and (5.15a))

.'
m, m f=},)
- 1,

. . . , -}.

(5. 19a)

Following the definition of the representations we know that the rules of

multiplication of the matrices R(i) = (r~~t) are similar to those o/matrices R.
In panicular: to the matrices R x , R" Rz of equations (5.1Sb) correspond
nlat.rices Rlj), RV>, R~J), which likewise satisfy the commutation conditions
(5.17a). This remark is the basis of the proof in Appendix II.

Let us put as in (5.18)

(S.18a)
We shall assume these three quantities are the three components of the
angulqr momentum in the quantum state j corresponding to the representation
DJ
Wei justify this definition by the following more precise statement:
A P?ysical system must have a spherical synlmetry in order to admit the
group 9J 3 Hence we are dealing lvith an atom lvith a radial field, i.e . a
mononuclear system. As a result of the Wigner theorem, there corresponds
to each irreducible representation DJ of the rotation. group for such a
system a big square matrix inside tIle matrix H (4.14) i.e. a "system of eigen. .
values" labelled by the indices j an integer or a halfinteger. We will call this
index according to the terminology introduced by Sonunerfeld in 1920, the
internal quanfuln number or better, the angular quanfuln number . and the
states or the atom with such an energy, the qualltulll states j. Each state
Enj is descJ"ibed by the whole set of the (2j + 1) wave functions corresponding
to the representation DJ ..
As for tIle quantities L~, L~), JJI), we are allo\ved to consider them as the
cOlnponents of the total angular momentum of the ,,-1.tom: first because they

Ch. 5, 31

143

INFINITESIMAL TRANSFORMATIONS

satisfy the commutation relations (2.20) and the definition (2.19), as is

shown by (5.18a) and (S.17a); secondly because they are constants of the
motion. R<;ld8 is a differential operator of the group ~3 which leaves H
invariant by hypothesis. The Hamiltonian commutes with R<j> dB and also
with L~) , so these matrices are independent of time (cf. Chapter 4, 1.4).
For simplicity we will use units Ii:
M<j)

= i R~j),

M~J) =

R~j),

M~J) =

R~j)

L!f:> = hM!j) , etc.

(5.20)

(5.20a)

Hence.forth we shall call these quantities Mfj) the components of the angular
momentum in the state j. The square of these momenta can be written:

(5.21)
3.4 PAULI MATRICES

It remains now to obtain the explicit expression for these matrices belonging to the different representations DJe Let us begin withj = !.
The variables of D i are the complex numbers , and " connected with
x, y, and z by

x+iy = 2'1~*

(5.6)

x-iy = 2~,,*
their transfonnation formulas are those of the group c1/2

{1(e' ==

a, +e+

fJ'1

- j3*

arx*+pp

(5.4)
,

a.* "

= 1.

(S.4a)

For infinitesimal transformations near the identity we will use a set of

small real numbers K, "t, j.l, v with which ,ve express C( and p as fo11o\\ls:
C!

..)
= 1 + K -r- 11\.;

{]
j

= Jl-f.. IV ..

Equation (5.4a) can now be \vritten:

1 + 2K + ,,2 ~t- A2 + Jl2 + ,,2

Hence

1(;

= 1.

is of the second order and negligible. There remains:

de = iAe + (.u + i v)'1,

d'1 = - (jl- i v)e - iAll.

144

ROTATIONS IN 3-DIMBNSIONAL SPACB

[eb. 5, 13

The three basis transformations are then:

l O A, == 0, Jl == 0
d~

dIJ

== i WI,

== iv,

(S.22)

d(x+iy) == -d(x-iy) == 2iv(,,*-'1'1*) == 2ivz

<Ix == 0, dy = 2vz, dz == 2iv('*'1-17*') == -2vy.
It is a rotation around the x axis over an angle d8x = - 2y.
2 A. == 0, v == 0
(5.22a)

A similar calculation sives

dx

:=

2pz, dy == 0, dz == - 2p.x,

i.e. a rotation around the )' axis over an angle dB;, = 2p..
3 Jl = 0, v == 0 .
(S.22b)

dx == Uy,

dy == -Ux, dz == 0,

i.e. a rotation around the z axis over aD angle d.9c == - lA..

The matrices R~i), R~i), i ) are defined by (5.19a) withj := !. The variables q". are then equal to and 11. They are the matrices of the equations
(S.22) to (S.22b) where v, p, A, must be replaced by their values -1d8~,
!<lS" fUld -!d8. Finally, by suppressing the factors d8%, dB" d3:r and
multiplying by i according to (5.20), one obtains:

.r.
e

~*) = t (~ ~), ~*) = t

~i). M~*) = t (~ _~) . (5.23)

here multiplied by the factor t are

The Hermitian matrices which are

the matrices S.", S" S. wed/or the first time by Pauli in 1927 in hia theory
of spin. It is convenient to introduce
M<j) ==

~+)+~t)

::.

(~ ~)

w!) == ~*)-~*) == (~ ~)

(5.24)

in which case we deal only with real matrices.

These operators no longer represent infinitesimal rotations because
multiplying the equations (S.210) by i completely alteR their features.

Cb. S,

I 3]

INFINITESIMAL TRANSFORMATIONS

145

However, these operators satisfy the following commutation relations

== 2Mz,

M+M_-M_M+

M_M:I-MsM_

= M_

(5.25)

McM+-M+M. = M+,

where we have suppressed the superscript -!-, since they hold in general1ike
(S.17a) and (2.20).
Let us introduce three infinitesimal parameters, ds, dt, du; then the
transformations arising from M<!), M<!> and ~i) are
MCi)'

~+)

== { de = "ds
d"

=:

= { d, == 0

d'f == edt

M~i) == { d~

d"

(S.24a)

= tedtl
=

-i"du.

Before investigating the consequences of these formulae, let us verify

that the ~atrices ~n == i~J) , which represent the components of the
angular momentum are Hermitian. This is necessary in order that these
quantities have physical meaning. The matrices R<j), gjj), R~J) give rise to
unita.ry infinitesimal transformations in the state space and this should
be the case in order to keep the system of fundamental bases functions orthogonal. This remark suggests the following general theorem, the proof of
which is, very easy.
The n1atrices of the infinitesimal transformations of a unitary lineQr group
become' Hermitian when they are multiplied by the factor i(i 2 :: -1).
Let us consider a, finite dimensional space in which the coordinates x,
undergo a linear infinitesimal transformation:

(5.15)
In order that this transformation be unitary (L'x~ x.
have the following condition fulfilled:
't"'"

dXI

L x, t

i.e.

ds

~ dx~
+ .t.J
Xl ._- == 0
!

ds

== const.) ,we must

146

for every value of

and

Xl

x,,_

Hence we have
a1k

Of,

fCh. 5, 3

ROTATIONS IN 3DIMENSIONA.L SPACE

if we put 'A'

-aki

= iA
,

al.k

,*

aki

3.5. ANGULAR MOMENTUM IN THE STATE j.

,,,ill

We
calculate the expressions MCj.>, Mf.!}, Jjl~J) and (M(J)2. The coordinates of the space BtJ form the basis for the unitary representation D J
They are the (2j+ 1) monomials q~) of degree 2j in and '1 given by the
formulas (S.8e). We obtain by differentiation

dq~)

(j + m)eJ+m-l11J-mde +(j - m),i+ ml1J-m- 1dl1

=

J(i+ m)!(j -

m)!

The infinitesimal transformations induced in the space BtJ

M~), M~") operating in the space Dli are, according to (5.24a)
)J:J+m-1.-i- m+ ld

S+ :dq!:'

== J+m,-

s_ : dq!!)

'f

J(j+m)~(j-m)!

(j

by Af<j>,

s = J(j+m)(j-m+l)q~~lds,

m)eJ+m+l,y-m-ldt

-.'

= J(j-m)(j+m+l)q~J)dt,

J(j+m)!(j-m)1
:. dq(}}
S z
m --

..... --

mq(J)du
m

Hence/the elements of the corresponding matrices

f <m- M<.!>11n+l) = <m'M~j)-iM~j)lm+l>

l

M(j)

are (cf. (5.15

<mIM~lm-l> = <mIM~)+i~j)lm-l> =

JU+mXi- m+ 1)

J(j-m)(i+1n+1)

f
<nlfM~J)lm> = m.

(5.26)

These are the fundamental formulas for the components of the angular
momentum. Each of these matrices has in each row and in each column only
one element different from zero. Only the last one is diagonal. III the state j,
the component M~J) can assume the values m, L~j) the values mh with
m = j,j-l,c ... , -j. In order to separate the different states corresponding
to these different values it is necessary to destroy the spherical symmetry
(Stem and Gerlach experiment).
V'e have according to (5.21), (5.24) and (5.25)

Ch. 5, 3, 4]

"fRANSITION FROM THE GROUP

147

(S.27a)
If ~ne uses this formula for the state j using (5.26) and the known multiplication rules one finds that the matrix (M<j)2 is diagonal and has the
value:
(M(j)!,m = (i+m)(j-m+l)-m+m 2 = j(j+l).
(5.27)

This matrix is a multiple of the unit matrix. It is invariant under rotations

and commutes with M<j>, M~j) and M~j). Hence in a given quantum state j,
the square of the length of the angular momentum has,an a~\J.lately . known
value. Its component along the z axis can assume (in units h) the' discre~
values m = j,j-l, . .. , - j and these values can be measure4in a lllapetic
field H which fixes, in laboratory space, the particuJ~diJ:ecti9n
O~.
If we know the value m of ~J) it is not possible to
anything accuqlte
about M!1) and M~J), which are represented by non-diagon~ matrices
thus are undete~ned. We are very far, from the classical picture.
However, (5.27) and the last equation (5.26) can be considered as the
foundation and the correct interpretation of the vector model w~chis still
very useful in discussing experiments. It can be seen.. from tJiese e,quations
how the vector model can be used practically, i.e. how. much they a,re
wrong and how one must correct quantitatively the qualitative exact conclusioll to which they lead us. For example M~J) behaves like the projection
of a vector M(}) on the z axis, the angle between these being fixed (space
quantization) but the maximum of this projection is, in the vector model,
not/equal to .j(M(J)2 :; .jjU+ 1) but to j. The usual geolIletrical n9ti9ns,
as for example the theorem of PYthagoras, have to be modified if we want
them to apply to the eigenvalues of the quantum vectors.

say

of

and

4. Transition from the Group !7A 3 to the Subgroup PA 2.

4.1. ZEEMAN EFFECT

An atom in the state j, Vtrith a (2j+ l)-fold degenerate energy E.J , is

subjected to an external magnetic field H. The sl,herical symmetry of th~
system is replaced by cylindrical symmetry. The perturbation lW arising
from this field is now invariant only with respect to the Abelian subgroup
fil2 of !!J 3 , the elements of which are the .rotations around Oz (Chapter 4,
3). The invariant subspaces of the group are then one-dimensional

148

ROTATIONS IN 3-DIMENSIONAL SPACE

[eh. 5, 4

and, aocordins to the Wigner theorem, each of the levels En) is split in such
a field into (2) + 1) discrete levels.
From this
the reduction of D J follows at once. All the matrices of this
,
aroup reptesenting fetations Q)% around the z axis are in the diagonal form
(5.3&). As a result of the perturbation A. W they will be divided into (2j + 1)
one dimensional matrices without any change in the matrix elements and at
the same time the eigenfunctions are supplemented by terms of the order of A..
Under a rotation 0). these eigenfunctions undergo the following transformations:
"':,. II:

t/I_ exp (-imco,)

m == j, j -1, ... - j.

(5.28)

To each of them there corresponds a perturbed level E,.J".. If one counts

these levels one obtains the number j which is characteristic of the angular
momentum/state of the ,:"om.. This can be done by a determination of the
number of Zeeman components of the lines of the spectrUm in a weak
magnetic field (l is sma]), and by trying to fit this number to a certain
multiplicity of the initial and final states of the transition. By this method
Sommtrfeld htu fotlnd it wtU 1Ulcuaary 10 Jl8e odd numbers 2} + 1 tmd half
Integer, j II1I4m.
Thtf'solution of the secular equation (2.47) can be found as easily as the
reduction of DJ In the zeroth approximation we only take into account the
principal term Wo of the perturbation matrix, the elements of which connect
the (2j+ 1) states among themselves (cf. 239a) so we have
I

+J

Wo "'". ==

w".""

t/Im'

.' - - J

lfwe carry out a rotation 00. we have, according to the invariance properties
of Wo " and' the transformation formulas (S.28),
Wot/l ... exp (-imQ}.) ==

L wlllm,y,,,,, exp (-im'Q),).

".'

This relation can only be satisfied for all the values of the angle

00.

if

w..., == w",8 _III'.

Then the matrix Wo is diagonal and the perturbed energy levels can be
written

E,,)m == E.J+Aw
This formula embodies all the information which can be obtained by
group theory. The perturbing function and consequently the value of the

Ch. S, 4, S]

PRODUCT OF TWO REPRESENTATIONS

149

constants w". depend on the way in which the field influences the atom. For
the first attempt we will try to apply the classical formulas of the Lorentz
electromagbetic theory to the Bohr-Sommerfeld atom model: i.e. to the
angular momentum L z of the electrons of the atom around the z axis there
corresponds a magnetic moment (i.e. its projection on the z axis)

(5.29)
and an energy
lW = -HvII%

-H

- L%
2ntoc

where mo is the mass of the electron, e its charge, c the velocity of light
(H plays the role of l). Since L% is in diagonal form JI. and Ware also in
diagonal form, i.e. to each element mit of the matrix L:z there corresponds an
element of the matrix A. W

(5.30)
is the Larmor precession frequency.
The experiments did not verify this formula. To include the anomalous
Zeeman effect, the second member of (5.30) must be multiplied~by a certain
number g, the Lande factor which changes value from one spectrum to
another. The simple Bohr model which would hardly have explained the
fractional numbers j, must be modified. This statement is the origin of the
theory of spin.
B~fore treating this problem it is necessary to come back to the theory of
angUtar momentum and to establish their quantum addition formulas. In
this way the classical vector model of the atom will be completely justified
and perfected.
COL

5. Product of Two Representations. Reduction Formula

S.1. KINEMATIC COUPLING OFTWOSYSTEMSWITIiSPHERICALSYMMETRY

Let us first consider the problem from a physical point of view: suppose
one adds the last electron to a monovalent positive ion in order to form an
atom. The state of the ion is known. First we neglect perturbations, i.e. we
suppose that the action of the ion on the electron is described by a central
field. The electron forms a second system in a state that is easily defined and

150

RO'fATIONS 11'1 3.. DIMENSIONAL SPACE

[eb. 5, ~

calculated by the Scbrodinger theory. Then the ion and the electron are
further coupled by taking a certain perturbing function into account. The
problew is to find the state of the complete atom formed in this way_ That is
what levels may result from a certain initial level as a result of this coupling.
A detailed dynamical discussion will not be undertaken here; this discussion
would be difficult and in fact has been done only in some simple cases.
We shall only give a preliminary purely kinematical study leading to a
classification of the energy levels.
Two groups are essential here: tIle rotation group and the permutatioft
group. Only the first one is actually of interest to us but the method is
general. We couple two systems with a spherical symmetry whose angular
momenta are res~tively il andj2 and we look for the angular momentumj
of the total system submitted to a coupling with spherical symmetry_ A
similar problem arises concerning the spin and orbit.aI momentum.
S.2. PRODUcr OF TWO REPRP..sENTATIONS

Two systems for which the Hamiltonian is invariant tlnder the operations
of a group (I are brought together. Their interactions are supposed to allow
~C)- same group. They are respectively in the quantum states correspondin.g
to;,the irreducible representations ~1 and ~2 of <!J (cf. (4.13) and (4.14)).
How can the resulting states of the total system be classified?
.According to our hypothesis on the interactions this classification is
i1irJqe1llltmt of their magnitude; this classification depends only on the represdntatioD of G in the function space and can be found in the limiting case
~here the dynamical coupling is infinitely small (but different. from zero).
. rhe gradual strengthening of the coupling alters the positions of the levels
themselves without any change in their distribution over level-systems and
their degeneracy.
Let Xl' X 2, " ., ..'%7., be the coordinates of the particles of the first system,
Y1, Y2' ~ ... , y" those of the second systerrt, n1. an.d n2 the dimensions of the
representations l 1 and { 2 t I.e. the degeneracy of the corresponding levels.
'fh:~ first system is in tile energy level El t , the eigenfunctions of which are
arbitrary.. linear orthogonal
- , combinations of 11. basis functions, Le.
l

'" =

"1

L qmy,m(Xt , x~! . . ~ x~J~

m~l

The "components" q", are the variables of the' representation I t and the'
t/lm the basic vectors of the sp~Jce ~f"

Ch. 5, S J

PRODUcr OF TWO RBPRESENTA'rIONS

151

In the same way the second system is in the energy level 2' the eigen-

functions of which are

112

lp

L q~ tpl'(Yl , .v~ Yt).

1&-1

The q~ are the variables of the representation ff2' the CfJp are the basis
vectors of 812
It is known that if the coupling is loose the SchrOdinger equation is nearly
separable in two independent equations, the eigenfunctions of which differ
little from the zeroth approximation eigenfunctions tp = t/llp i.e. linear
basis functions 'P.JI. == l/I".tpp.. The perturbed
combinations of the
levels differ little from the level E = El + E2 the degeneracy of which is
nl n2 for coupling of strength zero. All these results are well known. We have:

"1"2

'1'

:=c

~JJ'
t/lcp .. L,
q", qp. "'m lP"
"',.

:: ~
~ Q",,, 'I'm".

( 5.31)

1ft.. "

The " are vectors of a n1112-dimensional space which is spanned by the

basis vectors 'P_" and which is designated by the symbol
-1

Bl2

It is the direct product of the two spaces

tll1

(5.32)
and al 2 and its coordinates are

(5.31a)
r

i.e. all ~ossible products of the coordinates of ut1 and 1l2 .. This is equivalent
to builtling a tenaor of the second rank/rom two vectors. (Compare Chapter 1~

1.2)~

BY! iteration one can obtain tensors of higher rank.

Under the operations of the group f the components Q.,. undergo linear
transformations which form. a group ~, a representation of t6. This representation is the direct product of the two representationa f/J 1 and <62 "

(5.32a)
From
111

t/Im -+ sf/!m == 1'-1

L "',. C"".
aDd

r
"2

CP. -.. 8qJtn ==

one gets:

pal

CfJ,7pp

152

ROTATIONS IN 3..DIMENSIONAL SPACE

'l!

-+ S tp

[eh. 5,

= L: Qml'S 'P

mp.

L: Q;pWrp
rp

with
"1112

Q;p =

L: 1Crm Ypp Qml' = L: Crm, PI' Qmll

m,

{5.33}

1'=

a transformation whose matrix can be written in a symbolic way

C = C 1 XC2 ,

= ,(Crm,PI')'

C1

= (Crrn ),

C2

= (YPJl).

(5.33a)

If 'lJ 1 and l{} 2 are two representations of C!J, ttl is also a representation
of C. This can be easily verified. But generally this representation is not
irreducible even if f!ll and '62 are.
Conse,quently, if we want to use the Wigner theorem and determine in how
many distinct levels E will be separated under the influence of the interaction
between the two systems and what will be the degeneracy of each of these
levels, the reduction has to be performed further. Our physical problem is
then t~1ated into a purely mathematical problem: .How to ,educe the
product t6 == 'lJ 1 X ~2 in its irteducible components. To each irreducible
compon:ent, that is to each invariant subspace of fJl 1 X 81 2 , there corresponds
a level whose degeneracy is equal to the dimensions of this invariant subspace.
~.3.

lU!nucnoN OF THE DlRECf PRODUCT OF TWO REPRESENTATIONS.

G~OUP fl. AS AN EXAMPLE. CLEBSCH.OOR.DANFORMULA
~

Thel problem which has just been formulated is easily solved in the. caso
of ~ unitary unimodular group tfl2' or what amounts to the same .in'th6
of the-rotation .&roup ffl 1 From these one. can obtain the ~. for:
rePresentations ~f higher dimensions. The reduction formula finally arrived

ca

at is called by mathematicians the Oebsch-Gordan formula:

(5.34)
First let us consider rotations CDs around Oz. The matrices SJ(J)z)~ and
Sr(OJz ) are diagonal and have.tIJo form (S.3a): one of them contains the
elements exp (- im wz ) with m- j~j- 1, ... , - j; the other one the elements
exp( -im'wz ) with m' . j', .. _, -j'. To the operation ())% there corresponck
in the product DJXDjl a matrix wbiclt remaIns diagonal after (5.33) and t~
elements ofwbich are the (2}+ 1)(~' + 1) exponentials: exp-i(m+m')ro:r;' =
em+m'(e = exp(-iro~) among these [2U+j')--+-1] only are distinct .

Ch. 5, 5]

153

PRODUcr OF TWO REPRESENTA'fIONS

They can be classified symmetrically with respect to a horizontal line

which contains the terms of exponent zero, the positive exponents are above
the line, the tlegative ones below; if we suppose that j > j' one obtains the
following table in which we have written all the exponents which have the
same value on a given line. It is sufficient to read this table through the
vertical columns in order to verify the above formula
m+m'
i+j'
j-l+j'
j-2+j'

j+j'-l

i-l+j/-l

j-2j'+j'

"

. . . . . . . . ..

-j'+j'

...

~.

j-f-j'--l

j+j'-2

i+j'-2

. .,

j-.1' I

....

i"
. .. i J

}-]

--]-I

. . . . . . . . . . I . . .. ......
-U-2j'+j/)
-(j-2+j')

-U-l+i'-l)

-(j-l+j')
-(j+j')

-U+j'-l)

::::::::::: I::::
~.~:!~ -:.~). :::: I

.=

i+j'

-- ( }.- )")

-(j-j')

'" . . . . . . . .. ."... I

,0,
( ]-t-)
-

2)

-- (i+j' -1)
-(j+j/)

We will be satisfied here with this hint Wllich is not a proof. It mainly
serves to illustrate that the number of basis functions is sufficient to obtain
(S.34}.1
The formula (5.34) is equivalent to
DjXDjl

= Dj+j,+Dj-iXDj,_t

(5.34a)

since ($.34) can be obtained by successive applications of (5.34a).

To establish this equation rigorously the space iRJ x ItJ , must be decompose~ in an irreducible subspace IlJ +j , which corresponds to the representatio"n D j +J, and one other reducible subspace IlJ-iXUlJ'-ie This decomposition can be obtained by looking for the basis vectors of Dti+jl (cf.
Appendix III).
The formula (5.34) is the symbolic translation of the addition rules of
angular momenta into representation theory language. According to H.
Weyl it is the fundamental formula of the classification of atomic spectra
as well as of the theory of chemical valency.
5.4. TOTAL ANGULAR MOMENTUM

Let us come back to our example: we bring an electron and an ion together,
t\VO systems both of which have spherical symmetry_ One of them is in state
1

This can be completed. using the theorems about the chrtracters (Chapter 3, 11.2).

154

ROTATIONS IN 3-DIMENSIONAL SPACE

[eh. S, '

j corresponding to the representation D J of 913 , the other is in state j'.

We want, to calculate the components of the total angular montentlmt

The representation of ~3 which determines the states of the total system
is DJxDJI in the space 8tj x8tJ with the variables Q(mm') = rl..q~:, q~ and
having the form (5.80).
As a result of (5 . 19a) and (5.20) the component of the total angular momentum alODg a given axis, which we will call M, is obtained by writing the
equations of the infinitesimal transformations of the variables Q(mm/)
induced by a rotation dB around this axis
.}

t?,

dQ(mm ' )

=-

L M mm ,; mtm

'1

Q(ml' m;)d8.

ml,"' 1

I.e.

(5.35)
Or -using the notation defined by (S.33a)

M = (M(J) X 1)+(1 X M<J ,

(5.35a)

an ;equation which expresses the additivity of momenta.

The properties of the angular momentum as well as the vector model of
thel atom will be considered more rigorously in the following sections. To
dOl this it is necessary to take the interactions between the two- partial
s~cms into acoount and to reduce the representation D J x D J according
(5.34) in order to separate the levels which were mixed before.
(

to

'.5. HELIUM ATOM WITHOUT SPIN

Let us consider first two electrons in the Coulomb field of a nucleus. If we

neglect their interactions and if we do not take into account their spin then
their states are described according to the Scbr6dinger theory by wave
functions of the type (S~l):
, (m)
d ,./~(.')
",,,,

an

'f' .'1'

Here the integer I plays the role of j and the values I =: 0, ], 2 .. ~ corre..
spond to the so. .called s, p, d~ .. " states of the eleetron. In the zeroth approx-

Cb. S, S J

PRODUcr OF TWO REPRESENTATIONS

155

imation, i.e. when one neglects the interactions between the electrons, the
wave functions of the total system are linear combinations of the products
"'~f) t/I~~:) the coefficients of which are the variables of the representation
D,XD . The levels E := E",+En'I' depend neither on m nor on m'.
If we" take into account the Coulomb interaction between the electrons
the representation D, X D must be reduced in its irreducible components
as in the formula (5.34). To" each of its components there corresponds a level
Ea'L and an angular momentum operator M(L) (where L .. 1+ I', I + l' -1 ...
11-1'1) i.e. a well-defined state of the atom, L is the total azimuthal or orbital
I[tID1Ztum number.

To give a geometrical picture of this decomposition into 21' + 1 distinct

states (/' < I) attribute an angular momentum equal to iii to each electron,
i.e. a vector M(l) == I, and add the two vectors' and " provided that they can
take only such relative orientations \vhich give a resulting vector of integral
length. When L = 1+/' the two vectors are parallel, when L == 1-/! they are
antiparallel.
It can be seen from this simple example that the equation (5.34) is really
the exact basis of tho vector model of the atom. All the formulas from
(5.20) to '(5.27) remain true provided that j be replaced by the integer L;
in particular (M<L)2 == L(L+ I) and not L2. One labels the states L of the
total atom (L == 0, 1, 2, ... ) by the symbols S, P, D, ....
Let us give some examples:
10 Tvto s electrons; I = /' == 0; DoXDo = Do; the atom is in a state S
(ground,' state of helium).
2 Otte s electron and one p or d .. electron; Do X Dl == D t ; Do XD 2
== D 2, ... the atom is in a P or D - ... state.
3 'Two p electrons; I == I' == 1; Dl XD 1 == D2+Dl +Do- As a result
of their interaction a single level is divided in three levels: S, P and D.
(Compare also problem 5.3.)
These rules are general, more complex atoms can be built up in the same
way step by step. The use of half-integers j does not change anytbing essential
except that the letter L must be replaced by J ( 6).
We must mention an important consequence: If two systems are coupled,
the first of which is in a j == 0 state i.e. a closed shell of any degree of complexity and the second one in a statej', one always obtains following (5.34)
only one resulting state J := j'. The simplicity of the spectra of the alkali
metals which consist of a closed shell plus one electron is thus explained at
least if one neglects the spin .

156

ROTATIONS IN 3-DIMENSIONAL SPACE

[eb. 5, IE

6. The Electron Spin

6.1. UHLENBECK
AND GOUDSMIT HYPOTHESIS
" ,
..,

For each ele.tron the Schrodinger theory uses only three quantum numbers n, I, and m (cf. 1). On the other hand, the study of the anomalous
Zeeman effect has led us to consider half-integers j which playa role similar
to the numl1er I and which correspond to the representations of the group
3 of even dimension. Then it is necessary to consider a fourth quantUDl
number in order to complete the hypotheses of wave mechanics and to set
up a connection between 1 and
j"
-"
Let us consider as Sommerfeld did the experimental results concerning
the spectra of alkali metals. We have just seen at the end of the preceding
section that these atoms have completely filled shells for which j is zero,
plus a single external electron the state of which is defined by the total
quantum number n and the orbital quantum number I playing here the
role ofj'. In this case the formula (5.34) shows that the levels resulting from
th~ coupling must be simple. The experiments, however, show that they are
double. The two states of this doublet are distinguished by a fourth quantum
number- and this number can only assume two distinct values. We foresee:
that these values will probably not be integers because the Zeeman effect!
for the alkali metals is anomalous in a magnetic field; the spectral terms arel
divided in an even number of components (2j+ 1) so thatj is ahalf...integer,l
The *umbers I are known from the study of spectral series and selectionl
ru1es~ The experiments show thatj is always equal to I+!.
Consequently the two states of the electron which form the doublet must!
be ,distinguished by the quantum number s = +t with j = I+s.
.
-In the more complex spectra involving many electrons such as the alkaline~
earths, these hypotheses and the vector model were able to give an explanal'
tion of the experimental facts. This work was done by Sommerfeld (192()..~
1923).
;
As a result of a precise discussion of empirical results Lande discovered mt
1923 a remarkable relation between the splitting factor 9 and the numbenl
J, L, and S which replace j, I and s in atoms with many electrons. TbPI
relation can be written, in the case of alkali metals
i

_ 2j+l
g - 21+1

( 5.36)

The theoretical explanation of these results stayed somewhat confused

Cb. 5, 6]

THE ELECTRON SPIN

157

until 1925. However, that time Uhlenbeck and Goudsmit had the idea to
connect these facts to some phenomena of a very different kind: the gyromagnetic effect, and thereby found the key to the problem.
Einstein and De Haas have measured the change in angular momentum
of a ferromagnetic substance when the magnitization is suddenly reversed. 1
Barnett studied the inverse effect where a certain magnetization is created
by rotation. These experiments which were improved more and more
showed that the formula (5.29) connecting the angular momentum L z of the
atom to its magnetic moment Jt % (which comes theoretically from the existence of the electron orbits) is not true for the elementary moments of the
ferromagnetic materials. For these substances the right-hand side of this
equation must be multiplied by the factor 9 = 2

(S.29a)
It is sufficient to substitute in equation (5.36) I = 0, j = t in order to
obtain this factor. 2
All these facts suggest the following hypotheses which we can express as
follows by using the vector model:
The quantum number 9 is related to a fourth and last degree of freedom of
the electron which can only be a rotational degree of freedom. The electron
then has ~ angular momentum or spin, the projection of which on a given
axis Oz Can only assume the values

(5.37)
The corresponding magnetic moment is given by (5.29a). Hence we have:

en

Jt=+-z
-

(5.37a)

2moc

according to (5.29) and (5.26) this moment is equal to the magnetic moment
of a p-orbit (/ = m = I) i.e. a Bohr magneton. 3 The ratio of the magnetic
1 A similar effect can easily be observed when a gyroscope is suddenly turned over .
s'In experiments SUCKSMITH [1930] succeeded in measuring the Lande factor g for
certain paramagnetic ions directly by gyromagnetic experiments and he found a value
equal to the spectroscopic number.
8 According to tho correspondence principle one .finds in classical elcctromasnetic
theory the formula (S.29) for an electric charge moving along an orbit. The formula
(5.29a) for a spinning electrically charged sphere correspOnds to considerations in the
theory of Relativity (compare for instance M0LLER [1952]).

158

ROTATIONS IN 3-DIMENSIONAL SPACE

[Ch S, 6

moment to the angular momentum is twice as large for the spin

as for the

orbital motion.
As the ener.gy difference between the components of doublets and multipleu:,
is always smalI the secular equation of Schr<Sdinger is a good first approximation and the dynamical interactions of the spin (interaction with the
orbit or with the spin of an other electron) can be considered as perturbations. 1
6.2. TRANSLATION IN QUANTUM THBORY (PAUll)

Let us consider a system with only one electron: the Schradinger wave
function y,(x, y, z) must be replaced by a function of four variables l/I(x",
z, s) where s can only assume the values !-.
If we know exactly how the spin is oriented (for this we must use an
external magnetic field to define a certain direction Oz of the space as in the
Stern and Gerlach experiments) we can fix the value of s. If spherical symmetry is preserved, which is usually the case, two values of s are possible, each
one with a certain probability. It is necessary to use two functions at the
same time,
I/!l(X) == t/I(x, y, z,

+i) and "',,(z)

"'1

=c

t/I(x, y, z,

-t)

where y,! d't and l/I; tfi2d-r represent the respective probabilities. for the
two values of a in the elementary volume dt
dxdydz. These two functioDl
can be cPnsidered as the two components of a vector in a two-dimensional
space _~ or spinspace. This vector is sometimes called a spinor. The state of
an atomjc system with one electron is then represented by a two-dimensional
vector "or spinor:

(5.38)

"2

and
are t\VO orthogonal unit vectors; to each of them there corresponds a well-defined state of spin orientation: they are pure spin ju;nctio1l8.
Their orthogonality means that an electron cannot at the same time be in
the two states of spin +; and --1=
U1

+t) == 0;
Ut( -t)U2( -!)+Ut( +t)U2( +1)

Ut(

-1) == 0,

U2(

o.

If we go back to the definitions (S.3l) to (S.32a) of the direct product of

This last hypothesis is justUled by classical pictures. The maanetic forces comlq
from the rotation are small with regar4 t.o electrostatic Corce$.
1

Ch. S. 6]

THB ELECTR.ON SPIN

159

two spaces and of the direct product of two represea~tion8 we see that
tlte space of the junctions y, i.e. the total/unction ;"cludingtlJe spin U:11te
product lpaee II. X \Jl. This is exactly expressed by the formula (S.38} and
this is ~ starting point for the theory of the spin.
It DOW remains to establish by some hypothesis how the. vectors of the
space 81, behave under a rotation of the system. We shall assume that
rotations in ordinary space induce in the space fJl~ transformations which
form an irreducible representation DJ of the group Is.
This assumption seems so natural that it is difficult to make a di1fQrent
one. In fact -since It. is a two-dimensional space an>d if the represen~tion DJ
were reducibl~ it could be deco~pOsed in two one-dimensional representatioDs which would not tell us anything.
Since the space IlJ which corresponds to D j is (2j+ l)-dUneDsio~ j must
be equal to 1- and the group ~3 induces in the space It. X \Jt the representa-

tion DtxD.
6.3. APPLICATIONS

It now remains to consider some consequences of this result:

10 Consider an alkali atomc Let us neglect the spin and hence the simple
Schr6dinger theory can be applied. The energy level E", and the angular
momentum state are well defined by the integer I and the irreducible representation D, of !!J3.
We *ow take into account the spin and the perturbations which arise
from this. To the energy level Elf there corresponds now the representation
D .. XD, which can be reduced according to (S.34)
I

Di.'XD,

:=t

L D J == D,+t+D,-t*

(5.39)

We obtain a splitting of the levels corresponding to the two internal quantum.

numbers j at I+t, j' == 1--!-.
20 Let us submit this alkali atom to an infinitesimal rotation p. In the
space It.. X 8l this notation induces an infinitesimal linear transformation
the matrix M of which represents the component along the rotation axis of
the total angular momentum. After (S.3Sa) we have

M == (Mc.A-) XI)+(I X Mi'.

(5.39b)

The momentum M is the sum of the orbital momentum M<l) and thC
spin momentum Mii). The first can be defined by (2.18) (apart from a
factor Ii) but the definition of the second one can be given only with the
more or less explicit help of the theory of the rotation group.

160

ROTATIONS IN 3-DIMENSIONAL SPACE

[Ch. S, 6

3'0 The components of the spin momentum MCt) along the three axes are
given ~y the Pauli formulas. Only M~i) is in a diagonal form with eigenvalues :hi. These are the two observable values of the spin momentum
projected on a fixed direction of the space Oz. Whenever the last one is
determined then ~i) and M~*) cannot be determined simultaneously
because they are non-diagonal matrices.
6.4. COMPLEX ATOMS

These results can be generalized to an atom where r electrons participate

in the emission of light. It is sufficient to build this atom by attaching these
different electrons one by one. We neglect first the mutual interactions and
suppose that the first electron is in the orbital state 11 , the second one in the
state 12, , and that the wave functions have the form:

To each representation of the group

D

~3:

== D,.XDbX ... XD,,.XDiXDi

....

there corresponds a state i.e. a system of unperturbed levels with a (211 + 1)

..... (2lr t- 1) . 2r-fold degeneracy.
. Iiwe now'take the electrostatic interactions and the spin interactions" into

~t e,lCh of these levels'i$ di~d~d in a set ~t dlstmct ~ttieil~~: ~f

whkfh is equal ~o the number of irreducible representations D J in D.
The reduction is done step by step. But in practice one will take into

l/The 'Wave functions tpl~d fIJI of (5.38) can be calculated in the zero~ approximation
by scalV~hr6d.inaer th.eC>rY and they will have the form (~.l) aftet at ~Qvonient normalization. Consequently in an atom with r electrons tp- can be written 'in tbe -same approxi.
matio~ as
(a)

1p\.Xl' ., Sf'

) _

'PI

',1 ... 1

(I)

'1'_Ut) "PI (r) U" (1) Uk (I'

U,

Cr'

where the upper indices label the. electrons. The indices i, k, I . can only assume the values
1 and 2 corresponding respectively to
and $ = -t.
If'there are interactions between the electrons the product of individual space functions
is ~.,.~ b~ a ~in"e 'function and we havo
8==

(b)

+i-

Y(Xl, , S,.) == ~ tp,lt ... ,(XI z,.)u,(l) 1l1t(IJ

1,1 .. -1

'

u,(r).

If the perturbatioDi from the SPin are small tp,~ . , can in principle be calculated by the
scalar theory (with only the Coulomb interactions).

Ch_ S, 6J

THE ELECI'RON SPIN

161

account the order of magnitude of the different perturbations which may

change from one atom to another.
Generalir when the order of magnitude of these terms of the perturbatkm
Is normal the Coulomb interactions play the main role, then come the
interactions among spins and finally the interactions of the total spins with
the total orbital angular momentum. This corresponds to the Russell-Saunders coupling.
First one reduces the representation:
Dh XD X. .. XD,,. =

'2

L DL

This operation gives one term for each possible L-value: L is the total orbital
quantum number; to the values L = 0, 1,2, .. correspond the states S,
P, D. .. of the atom.
Then one reduces
Dj-XDtX XDi

==

I: Ds-

To ea,.ch ~:Ipnber S there corresponds a state with a given value of the

total spin; for example:
DiXD*

= Dl+Do

D.XDi-XDi

= Dl XOt+DO)(Pi

= llt+Di-+P.i

(S.~)

S is tp.e quantum number of total spin.

Fin~lly when Sand L are known, one reduces
J

= L+8, L+8-1 .. . IL-81,

(~.41);
... '-:,~

J: is the total angular momentum quantum number.

It is easy to translate these results in the vector language: first addition of

orbital momenta into a single total orbital momentum, then addition of the
spin vectors and finally the coupling of Sand L into a total angular momentum 1.
This last coupling gives rise to the multiplets because it is the weakest.
The multiplicity of a level is, following (5.41), 2S+1 if L ~ S, 2L+l
if L < Sand 1 if L = 0 (a singlet or S state).
The examples (5.40) show that the numbers S are integers or half-integers,
the multiplicities 2S + 1 are odd or even according to the number , of
electrons being even or odd: the alkali metals have doublets, the alkaline
earth metals have singlets and triplets and so on.
In certain atoms there exist other kinds of coupling, particularly the j j

162

ROi'ATIONS IN 3..DIMENSIONAL SPACE

[Cbs S, t 6, 7

coupling in. which the coupling of the spin of each electron with itl own ,orbit
is more important and gives a resulting angular momentum j; the momenta J
of the different orbits are coupled according to the equation
DJI XDJ2X ..

XD J,.

= L DJ

7 Selection Rules
to

These well-established rules can easily be confirmed by using group

theory. We will consider only the case of atoms in. this section.
Generally 1 the radiation is determined by the electric moment II which is
a vector in the ordinary three-dimensional space. Its vector nature shows.
itself by
characteristic properties:
1 it has three components P;e, P" J1.. or more conveniently:

two

Jl. +

==

fJJC +iJJ, ,

J.l-

=-:

JJ~ - ip,

and JJ:

~o ~f ~ s~~~

undersoes a rotation s these components undergo the

same linear tninsformatioDs as the coordinates of a point, x+iy. x-iy, Z,.
i.e. one of the transformations of the representation Dl (cr. 2.1 and 2.2).
In quantum mechanics the components of the momentum are considered
as operaton applied to the wave (lUlctions. Let us suppose tha~ the function
space a decomposed in certain (2j+ I)-dimensional subspaces III which are
invar~nt and irreducible with respect to the rotation group- Each orthese
subs~s is spanned by a set of "orthogonal axes" "'JM which constitute
a co~plete system with m := it j-l, ... - j (cr. 2.2 and eq. (S,.3a). With
thesel axes each component of Jl is represented by a matrix defined by the
eQ9ations (1.25) which can be written as
p+ t/I jM

Il-..pm

==JIm'
L Y,F.,(P+)J'''''I J== .;

J.l~ "'WI

= .

(5.42)
H

For example each of the constants, (p.~)J"'.J'm" raised to the second power
is proportional to the transition probability from the state Jm to the state
j'm' and to the intensity of the corresponding spectral line (with a polarization alona Oz). Without magnetic fields the levels are independent of m
hence tlUSindex caa be suppressed and one can investigate the selection rules
for transitions j ... j' only.
We will no&lect here, as in Chapter 2. 6, the radiation of higher order muttipotes
(quad!upolfjs. etc..)
1

SELECTION RULES

Ch.. 5. 7]

163

If we submit the system to a rotation s, the left-hand sides of the ,three

equations (5.42) are the products of a component of a vector II belo,,"ging to
the three-ditDensional space \Jl 1 and a component of a vector belongina to
the (2j+ l)-climensional function space DlJ Hence they are the components
of a vector belonging to the product space 91 1 X atJ This vector will
transform under the rotation S according to Dl XDJ which are reduced
with the formula (5.34)
Dl XD J == D J + 1 +DJ +DJ - 1 The righthand side of the equations (5.42) in which the matrix elements

are constants, is a sum of terms which transform like the components of the
vectors belonging to the spaces DlJ" i.e. according to Df, .... , with)' == i,
1, i. ..
Since both sides must transform in the same way under the rotation S,
we find that all the terms of the right-hand side must be zero except those for
which i' a=)+ I, j, or }-1.
Hence we have the following selection rule (the arrow shows the possible
transitions)
(5.43)
j-+j-l, i, j+l.
In case) -= 0 we have, however, D1 XDo s= D 1 ; the only possible transition is j r:.:: 0 -. j == 1, i.ell 0 -+ 0 is forbidden.
One ~nds by the same method the selection rules concerning the transitions of lthc magnetic number m in a magnetic field. The operations of the
group 91 2 i.e. the rotations. around Oz through an arbitrary angle co are
the onlY allowed operations: p+ is then multiplied by exp (ico), l/I}", by
exp ( ....jjmw), "'J'm~ by exp ( - im' Q. 1
For a totation with an arbitrary 0) one has
p+ "'Ja exp

(-iw(m-l) =

L "'J'm' exp (-im'w)(p+)J.,J'.'.

J'm'

AU the terms of the sum in the right. .hand side are zero except those for which
m' =- m-l. By a similar argument for IJ- and Jl6 one obtains final)y tho
following selection rule: the only allowed transitions are~
", -+

nt--l, m, In+ 1.

(5.44)

The first and the last one give circularly polarized light in the xy plane with
two opposite directions of rotation. The transition m -+ m gives linearly
polarized light with a polarization plane parallel to the z-axis.
1

For the

minus sian, see 1, equation

(S~3)~

164

ROTATIONS IN 3-DIMENSIONAL SPACE

[Cb. 5, 8

8. Parity or ReflectioD Character. Approximate Selection Roles

8.1. PARITY; THE RULE OF LAPORTE

..

The group of pure rotations is not the only one that is implied by the
spherical symmetry of atoms. This symmetry is also preserved under
refiections, which are all the products of rotations and the only operation
connected with "symmetry with respect to a center", i.e. inversion of the
axes:
X

= -x y

-y z

-z.

We shall label this operation by the symbolIC; it satisfies the equation:

1(2

= E.

The group ~3 is extended by this operation, which commutes with all the
rotations, to a' group !;. As a result of the preceding equation this inversion operation will have representations in the form of a diagonal matrix
containing either + 1 or - 1. The eigenfunctions or basis vectors are multiplied by the factor b = 1 as a result of this operation. This factor b is
called the parity I, signature or reflection character of the representation.
The.. parity-of a wave function "'(Xl, Yl' Zl, X2, ; 91,32' . ) depends only
on; the space coordinates of the electrons and not on their spin s. The spin
can be considered as an axial vector (.gular momentum, magnetic moment)
the 90~nents of which remain invariant under inversion of the coordi~te
axes. ~e operation" does not operate on the vectors of the space'It.. ,
but dobs
OD the vectors of the space It.
f '
In the one electron problems the orbital wave functions (." 1 and "'2 in
(5.3~J) are as a result of (S.l) homogeneous polynomials of degree I in x, y
andl z, 1 being the orbital quantum number, hence they are multiplied by
( -1)' by inversion of the axes. The spectral terms have then alternatively
the refteCtion characters b = + 1 and 0 = -1, i.e. they are positive or even,
negative or odd in the following order: s+, p_, d+ ,1- ... which is the most
frequent order even in very complex atoms. This order is characteristic of
those terms which are called normal by the spectroscopist~.
The parity of the states of an atom with f electrons can be calculated
a prlDri whenever one can give to each electron a well-determined orbital
quantum number I, particularly in the case of Russell-Saunders coupling.
Recent developments in the theory of elementary particle interactions again foc~
att~tion on the subject of parity. For thiS development
refer to the literature quoted by
La P960] .. A group theoretical classification was given by MELVJN [1960].
1

we

PARITY OR REFLECI10N CHARACTER.

Ch. 5, 18]

165

The zeroth approximation eigenfunctions are then the products t/11

of the individual wavefunctions and their parity is

"'2 .;,
(5.45)

A perturbation will arise from the coupling of the electroDi; this perturbation can be important, nevertheless it always preserves the spherical
symmetry of the atom and changes neither the representations of the
group ~3 nor the parity ~- The value of b is unchanged according
to (5.45).
In the helium atom the first electron is generally in the state 8(1 0) and
is completely determined by the quantum number 1 == L of the second
electron: the terms are normal.
This is not generally true. For example let us consider the. atoms COntaiDjDI
two electrons in their non-filled shell! and let us suppose that both are iD a p

state (/ == 1). The formula D1 XD. =: Do+Dl +D2 shows that three
kinds of states may occur: the S statea(L =: 0), P~8"~(L - 1) or D states
(L = 2). These three kinds of states are known in Alg; they are tile so-called
primed tenD.$ and they all have the same parity

8 == (_1)'1+'2 - (-1:f =- 1.
The experimental importance of the parity lJ is shown by a seteoeiDa rule
discovc?red empirically by Laporte, Itullell md Sa1lDden.
The ~sn of the comPOneDts of the electric ~ II .........0 every:
polar ..ector are
UDder invenion of _ : the ro..... iadaceiD!
ordituaryspaoethe traDaformatiollloftheJrGUP
with the l*ity 8 - --I.
Let us come back to the equations 5.42 and replace the ia.dicelj and", by;
11' 12 , I if the IY*B' U1ldergocs the operation x: tbe~. of P, is
~ .... -~-.. ...
18- ."",....t'tipJied by 8 _ (_1)" +la+ +'1 - ... ~J,
In.
C;11&1.1.""",
" , . " 11
.......U1
.uv ifJ'l'lr:a "1 uz
~ == (-1)"1 +"2+ +1'1. Since the matrix elements (P+)" .. I't_. _1'/ are
constants it is necessary that

cha."

-1

f.

't.

8' ==

-~.

(5.46)

Consequently: in, an allowed tra:nsititnt the .." of the orbittJl ~_~

numberl 4+12 + ... +If CQ1t 0111, be chaaged by lUI odd.1IIInJber (selection

rule of Laporte).
The completely filled shells do not cootrjbute to the total orbital munber L. This is a
consequence of the Pauli principle.
1

ROTATIONS IN 3.. DIMENSIONAL SPACE

166

reb. S 8,9

8.2 APPROXIMATE SELECTION RULES

The selection rules for the numbers Land S as well as those of j can be
obtained by considering the rotation group, but they are only approximate.
The operator of the electric moment p. + , Jl- , 1'. changes only the orbital part
of the wave functions (1/1 ilc " in the formula (a) page 160 footnote) without
any action onth.e pure spin functions (u~ 1 ) u12 ) .... ). As long as the perturbations
resulting from the spin are weak and the distance between the components
of the multiplets is small, these functions t/I can be calculated by the scalar
SchrOdinger theory and they haye an exact total quantum number L and the
spin functions have an exact total quantum number S. Hence when one writes
the expansions (5.42) of the components of the electric moment the pure
spin functions are the same in both members ofeach equation and the expansions
01 the second members. can only be done in terms oj" space eigenfunctions
"'ft .... '. An argument identical to that of the above section leads to the
rules!

-+

L+ 1,

L~

L-l

or AL = 0,

1 S -+ S or AS = O. (5.47)

The inversion of axes forbids the transition L -+ L.

These rules are the basis of the classification of lines in the series. For one
electron atoms they can be obtained directly by the theory of spherical
harmonics.
But,; while the rules (5.43), (5.44) and (5.46) are rigorous the last ones are
only approximate and cease to be valuable as soon as the perturbations due
to the ~in "b,ur" the sharp values of the vectors Land S. In fact, there are
manYi"exceptions for heavy atoms, where the series disappear almost complet~y.

sInce the

parity rule is violated. for strong interactions (large deviation

from th~.Russel1-Saunders scheme), at first it seems slightly contradictory that
in elementary particle theory the rule is violated for weak interactions. The
term "weak interactions" is, however~ purely generic; it indicates interactions
that are weak compared to the nucleon . . nucleon interaction .

,. Stadt Effect. Anomalous Zeeman Effect. Line CompoDeDts IateasIty.

Lande Splitting Fact9f.. Paschen Back Effect
The results obtained at the end of Appendix III allow us to complete at
certain points the theory of the .Leenlal1 e.ff,;;ct ( 4) and to say something
about the Stark effect.

.:
Ch '):7
<

Jt. .... )
~;

Sl'ARK E;FFECr

'::J

From th~, ,group t11eoretieal point of view the essential difference bet\veen
these t,vo pnenomena COJnes frorn the different symmetries of the magnetic
a.nd electric fields~ The first one has the symmetry of a rotating cylinder, it
admjts only the Abelian group formed by the rotations !!tJ1 aronnd the~field
and a reflection \vith respect to a plane perpendicular to the field~ 'Tht' second
one h.as the synlrnetry of a cone of revolution whickl al1o"~vs the' iO!f.. Abelian
group ,g{:; of fntations and refleGljons~ tJbvio"Csl:-;,v ;;:4? is a subgroup of ~}~ .
If there is a degenerate level corresponding to the irreducible representation
!!tj of the group '@.3' a nlagnetic field will split the level into 2j + 1 cOlnpo'"
nents (In :.~. j,j ..-l" . ~ . ~ ._j) according to Chapter 4, 3.2 and 4 . The splitting
in an electric field will be less conlplete. 'rhe two values ~ 1'1'1 of t!le tUagnellc
quantum number (ofvve could say the electric quantum (i1.uuher Iml)
correspond to a unique level which cannot be split by the electric perturbation (essential degeneracy). Only the level m = 0 contains i\\'o ternls} one
posi tive~ th.e other negative with the parity ,+ 1 and - J. (cf.. 4). A~Jtogetltel'
the splitting by Stark t~ffect will consist of j+ 1 separate levels"
(yenerally thisdecorflposition is of second order except for an accidental
degeneracy (hydrogen ca~e), the expansjon of the perturbation. begins
terms proportinnal tr; the S(H~are of' the .field .
fUllction is
....1..

where ,'% is the field and Ji,z. = Leizi is the projection of the electric m()ment on this field. l<he diagonal elements of tIle perturbation rnatrix are
",",,.,,1,.
1
t S f~1',
q (.1~I (J)
.1 (J)')
-'. 0 ..~....
'-',10 .... H 't'l...,. .;.(J)
th
- e S(c.d.aI
pJ. Ouuc
M ' flz t.//.\l;
Wlh;
uC h are a.;.11 Ze!
ttt..("3.1J:.-ii.
~,,\.! ~ He ,) ~i
have the sam.e parity for a given J and the total result 1~: antisynlmt!tric
due to the odd parity of the operator J.l:z.1
TIle selection rules (5.43) and (5~44) remain exact but (5.47) are violated
as soon as tbe electric tieid is strong enough to create eigenfunctions differing
fronl thelT zeroth approK~mation. by a significant amoullt .. i.(~. they start ti)
sho\v the symuletry of the new problem. 'The rule of Laporte is unnl0dified
by a magnetic field which is represented by an axial vector and hence gives
no change in parity. This rule is, however, not verified in a strong electric
field (a polar vector) since the perturbed wave function does not have the
''I"'

parirl of the unperturbated wave function .

Any integrand anti-symmetri<: in the coordinates will give zero when integrated over
the total space~
1

168

ROTATIONS IN 3-DIMENSIONAL SPACE

[eb. S, 9

9.2. INTENSITY OF COMPONENTS

Let US come back to the expansions (5.42) of the components of the electrio moment. Tho left-hand sides of these formulas are the products of the
components of two vectors. The first are p+ , p._ and p.. which transform under
a rotation as x+iy, x-iy, Z, i.e. as the basis variables of the representation
1
9 1 or more precisely (cr. 2) as q~11,
) and 2- t q~l). The second arp the
functions ';Jm which transform as the set q:.f> wit4 m =- j, ... , --j. At the
right-hand side the constants JlJ'.' t J"", which will be written with a clump of

-q1

the indices as (P)JAI,jnt, are multiplied by the fUDctions y,1M which transform
as the basis variables of the representations D1 - .
The equations (5.42) are essentially the same as relati011 (S.A26) in Appendix III~ provided we substitute in this last one j' ::Ill 1, m,' =: 1, 0, -1,
J =)+ 1, j,/j-I, m+m' == M == m+ 1, m, m-l. Referring to the remark
following the equation (S.A26) the matrix elements (p.+)J-.JJI, -(p,-)P..IJI
and 2*(p,S)Jm,JM have the form PJC~ where the coefficients C are given by
Table S.A.l in Appendix III, the PJ remaining arbitrary. This is the first illus-

tration of the Wiper-Eckhart theorem (compare Chapter 4 and 6).

Thus we obtain, apart from the factors PJ' the components of the electric
moment i.e. the "amplitudes" of the different components of a given spectral
line in the Zeeman or Stark effect, the squares of which represent the radiation intensities. These components are, neglecting some numerical factors)
for j

-to

=:

i+l,

(P+)IfI,m-l

-(Jl-)".. _+1
(p,)., ",
for j

-+

J = j,

(Jl+)".,m-l

= PJ~(j+mXj-m+i-)

(P-)".,m+l

= PJ.JU-m)(j+m+l)

CU%).,m
for j

-+

J == i-1,

= PJ+fJ(j-m+1Xi-m+2)
== PJ+ IJ() + m + 1)(J+m+2)
== PJ + 1 J(j+ m + 1)(j + m + 1)
(5.48).
....

= p)m

(Jl+)",.m-l == PJ-l~/(j+mXj+m-l)
-(p-)... JIJ+l := PJ- 1Jlj- m Xj-:"'=i)
-(p..)".,1fI = fJJ-l J(j-+ m)(j- m).

All the other matrix elements are zero. Hence we found more precisely
the selection rules concerning} and m.
,
These relations have been checked experimentally in vieak fields. We notice

169

STARK RPFEcr

Ch. 5, 9]

that the components of the j -+ j transition are of exactly the same form as
the components Me.{!, M!! and ~J) of the angular momentum in the
state J (section 3.4). This is not surprising.. In both cases the axial or polar
character of the HelOT does Dot make ~y difference in the pure" rota~ions.
For the same reason the formulas (5.48) are valid in Stark effect as well as
.,

in the Zeeman effect.

9.3. LANDE-FACTOR

In the same way as we have explicitly developed the electric moment

matrix i.e. by using the formula (S.A26) in AppendiX III, we shall give
some f11rther attention to the Zeeman effect and complete it with the help
of the preparations of 4.
Let us remember that in the perturbation matrix we must in the first
approximation take into account only the matrix elements that determine the
mutUal relations among the (2)'+ 1) states which form. the descnerated unperturbed level Enl These elements are related to the representation DJ of
the rotation group. We have found that these ele;lents form a diagonal
submatrix A(w.) l1l=j,j-l, ... -j. We have now to calculate the terms
lw", after completing the form of the perturbing function with the hypothesis
of the spin ( 6.1, 6.3 and 6.4).
Let H be the magnetic field. The angular momentum. L of the atomic
orbits and the angular momentum of spin S (infinitesimal rotations) operate
first on the space ilL of the orbital wave funptions and secondly on the space
Its of the spin functions. Their sum operates on the' vectors ~longing the
space It == fRsxBt and can be written as a result of (S.39b)

to

M == [MiS) X (1) +( I)" X ML)]~

The unit matrices (1)8 and (1) complete the matrices

(5.49)
M(8)

and

M(L)

in

the parts of the space It where these do not operate. The above equation
can be decomposed in three equations of the same form if we use M + , M --
and M .
As ,,'e saw in section 6<'the angular momentum due to the spin must be
counted twice in the calculation of the total magnetic moment. To the vector
M corres~nds a magnetic moment .A and its projection on the field H
can be written as a result of ~S.29) flDd (5.29a) as

.A% =- /l[A(S)XI+~I)XI+ IX~L~

where Pis the Bohr magneton

(ft

:=

;II:

P[M~S)X(I)+M%)

(5.50)

eh/2mo c). 'This formula is equivalent to

170

ll0T.I.\TIONS IN 3.. DIMENSIONAl., SPACE

the assumption that the total magnetic moment is the sum of two vectors
PAf(l.) and 2j)"W,:sl.
1o. At 1irst let us suppose that the field is weak with regard to the spin
orbit coupling (L S) which we suppose to be of the Russell-Saunders type.
The energy differences of the Zeeman components are small with respect to
the splltting anlung tIle different components of the multiplet, which wjll be
split separately.. The spherical syrrtmetry of the atom is hardly modified.
If one makes a &uitable choice for axes, the space Dis X DlL will be reduced
to subspaces 8lJ irreducible with regard to lj)3. The operator M of (5.49) is
decomposed into a. sum of operators M(J), each operating on one of the
subspaces UlJ and each being an infinitesimal operator of the different
representations g; J" rrhus
M

~=

I: l~,,(J) =

[}\-f(S)

XI

+I

M(L)],

In this coordinate system the perturbation function can be written as a

result of (5~50) and the preceding equation
AJV =:= - H viiz

::: --

P1I[M~S) X 1+

L M~J)].

(5.51)

The SI)litting of a terrrl .1 is, in tIle iirst order ~ obtained by considering only
that part of the operator, or that part of tJle correspo nding matrix,. which is
related to the space fft J That is only the part related to the (U + 1) basis
func1jons of the representation D J is considered:

(5.52)
\Ve kno\v the operator 1"17: which has. been calculated in 4 . It is defined
in;the space of total anguJar momentum. by the equation
+J'

Mzt#Jm

12 m'=12
J'

311d 'l;NC

tfrJ'm,(M~)!;~

-J'

knov,,' that this fnatrix is diagonal:

m .:::: J, J-l .. ". -J.

The matrL~ (J.\1~S) X I) which we shall designate for simplification by Sz is
defi!led by a similar e~qllation:
+J'

and these

t~'()

equations, together with those related to the component!

Ch. 5, 9 )

171

STARK EFFECT'

/ttl + and M _, have the form (S.A26) in Appendix III ("vhere one has to take
j = J, jf = ),' ,T = I') m + m' = m').1. Then Sz =
S~J). lChe lIlatrix elements
(M~J and (~~J are proportional to the constants C:"ltl' i.e . proportional to
each other \vhere the constant of prc,portiollality P.; is independent of m and
is the same for S~,J) and S~J):

2:

S~J)

= PJM~J)

(5.53)

== pJ(rn).

Finally we obtain
(5.53a)
a similar forlnula to that of section 4 but with a coefficient g == 1 + PJ \vb.ich
is called th.e Lande splitting factor: the energy difference between the Zeeman
com.ponents is g times the normal differences.
This factor can be very simply calculated as follo\\'s: Let us consider the
part of the operators fri operating in the space _(J) and let us put
(I X l\tjL)(J) =

t J ).

Then (5.49) takes the fonn

,,~(j)

S(J)

+ 15J )

or

.en ==

lYl(J) -

stl )

which expresses the addition rules of the vector model. This equation shows
that (J), 8(/) and L(/) commute among themselves since S<J) and
operate
on dMerent spaces . (Spin space and the space of the functions tJ!(x).} Hence

.n/)

vie

can "~vrite

(ni))2 = (M(J)2 + (S(J,)2 -2M(J)

.. S(J),

or according to (5.53)
2M(J) S(J) = 2PJ(M(J2 = 2p.1 J (J + 1)
= J(J -t-1)4- S(S + 1)- L(L+ 1)~

and finally

9 = l+PJ = 1+ J(J+l)+S(S+1)-L{L+1).

(5.54)

2J(J +1)

This is the formula discovered empirically by Lande. 2.

1 The summation ~m' contains for each component only one term according to the
selectiOt:l rules and formulas (S~26).
I

FOl'the geometrical meaning of Land~ g . . factor see Mayer B.nd Mayer p. 344 6g.15.2.

172

ROTATIONS IN 3-DIMENSIONAL SPACE

[eh. S

20 In a strong magnetic field the coupling (L S) disappears. The system

allows only the rotation group g)2 around the field direction. The space
8t8xll~Dnot be split in subspaces ItJ As 1)2 is Abelian, the matrices
M~L) and M~S) can be reduced entirely to the diagonal form and we obtain
aa;ordins to (5.50)

AW == -pH(mL+2mS)

-pH(m), mL

L, L-l, ... -L
ms == S, S-1, .... -S,
aD

the eneray differences between the split levels become normal again: as
they are large with regard to the energy difference between the multiplet
components and as the selection rules remain valid, the field gives rise to a
normal triplet. It is the Paschen-Back effect.
The intermediate cases can be studied without difficulty.
Appendix I
THE CONNECTION BBTWBBN A FOR.MAL SBT OF BASIS FUNCTIONS AND
THE SPHERICAL HAIlMONICS

Becauso we have used a formal choice of basis functions, the qf", we must
now consider how to deal with actual physical quantities. It was mentioned,
particularly in 2 and Appendix III, that the results obtained for the represe~tations
are independent of the basis. Specifically some of the " 17
,
co~ be replaced by 1]* and thus simplifying the basis functions to mo. .
nonfiats. As a result we do not have a one-to-one correspondence between ~,
f/ and x, y, z anymore. It is necessary to consider the inverse process. That
is if there is a set of normalized monomials
rI+ l8 as in (5.80) and if j
il"an intesert what is the corresponding description in x, y and z:J
The connection between the ordinary variables x, y and z and the formal
variables ~, " are given by the following set of formulas:

,*

,i-In

x+iy == 2P1~

x-iy ==
Z

I"V

+'12

2~". ~

,2

== CC*-",,*

JIv

(S.AI)

+C'1-

The first part of the equation. express the original results obtained in 2.
The seco~d part indicated by" ,...," shows the way these quantities transform.
Solving this for the separate variables the equations become, ."

(S.A2)

APPENDIX

Ch. 5]

173

Using the fact that these three quantities transform in the same way
consider a scalar defined by the internal product
,>

:z

,1,,2
x - +y -

~2:2

-z~".

(S.A3)

This invariant can be rewritten as follows

K == -(x+iy),,2+(x-iy};2-Z~I1.
(S.A4)
Nothing new is obtained since this expression says essentially the same thing
as formula (S.Al) or (S.A2). However new results can be obtained if we
raise this invariant to an integer power, I,
X, =:
e'-"",,'+m<p~(x, y, :z)~
(5.A5)

L
In

In the right-halld side terms of like powers in , and 11 have been grouped.
Each coefficient is a function only of x, y and z. Both the monomials in , and
11 and the polynomials in x, Y. z transform. according to a certain representation. Hence knowing that
form a basis for an lrlYJducib/e represen"'
tation we conclude that the qli'(x,)', z) do the same. It turns out moreover
that the functions, constructed in this way are actually the solutions of the
Laplace "equation. This is not surprising since one of the properties of the
Laplace equation is that it does not change under a rotation of coordinate
axes. Its set of solutions, forms a representation of the rotation group and it
is possible to choose the solutions in such a way that they form an irreducible
represe~tation. Hence we conclude that the solutions of the Laplace equation
should: be a linear combination of the polynomials we have constructed
above. f
If we went into more detail we could show that the polynomials are
actuany, up to a mulplicative constant, the well-known spherical harmonics~
(See KttAMBJlS [1937], page 170.)

,'-_,,'+m

Appendix

CONSTRUcnON OF THB I1lR.BDUClBLE IlEPllBSENTAnON

OF' THE GR.OUP .

The following argument is of a type that is commonly used in quantum

mechanics. It does not differ essentially from those which ODe finds, for
example, in Dirac. 1 It goes actually back to Lie and Cartan.
1 DIBAC [1958] Section 30. See also BoltN and lomAN [1930) Ch, IfdioD 27, which
we follow closely. This book it full of aroup theoretical araumcmta: the authors 4f.d DOt
find it necessary, however, to put much emphasis

OD

this fact.

174

ROTATIONS IN 3-DIMENSIONAL SPACE

[Cll.. 5

Since the rotation group is continuous, the representations consist of

matricc;& that are continuous and differentiable with respect to three parameters the group. Hence they are completely determined by three infinitesimal basis transformations, tllat is by three matrices Rx , R" R. or which comes
down to the same M+ 1 , M-l' Mo (compare (5.19), (5.20) and (5.24.
We demand only from these representations that they are unambiguous
in tlte neighborhood of the identity. We know that they have to satisfy the
commutation relations (5.24)

of

M+ 1 M_ 1 -M_ t M+ 1 =
M_ 1 Nlz -Mz M_ 1
M s M+ 1 -lVl+ 1 M:r

2M~

(S.A7)

= M-l
= M+l

(5.AS)
(S.l\9)

to which we add (compare (5.27a)

M2 =

M;+M;+M;

= M+1M_l +M-IM+l
2
== M+IM_I-M.+M;.

+M;
(5.Al0)

As--we have seen, Al2 represents the square of the total angular momentum.
This is an invariant 9f the rotation group, which consequently commutes
with aU the operators of this group; in particular with M", M, and Ms.1
We have now to find all the possible systems of three matrices M + l ' M - t
aJ1d Mo that satiSfy the preceding conditions and form an irreducible linear
group.
, Let ~ be an arbitrary representation of )3 and Dto the corresponding
'representation space~ Ito is a subspace of the function space 81, invariant
with respect to the rotation group~ We can choose in this subspace ~l
system of basis functions t/l1mt' finite or infinite, (we explain in a moment
why we use two indices) for which the abelian subgroup ~1. of rotations
around the z-axis is completely reduced. In other words the basis func~ions
.dre chosen in such a way that a rotation 8%, considered as a change of coor. .
dinate axes (compare Chapter 4, 2.3) induces the transformation

therein. The jnfinitesimal rotation is

dt/lmil = - iml/l".,. dOz
1 Independent of all physical interpretation this commutability can be verified by a
direct calculation with the help of equations (S.A7), (S.AS) and (S.M).

APPENDIX II

Ch.5]

where m is an integer of the representation and is unambiguous for all

values of 0.,.,. We have seen that this condition is not necessary if we are
content with unambiguity in the- neiahborhood of the identity (compare
Sec. 1). Hence it is sufficient to allow m to be a real or imaginary number.
From (5.19) and (5.20) we see that the corresponding matrix M% is diagonal
and contains a sequence of eigenvalues m, m', m", ... which may each occ;ur
with a certain degree of multiplicity hence a second index n. The t/lnm are
the eigenvectors of M z.

I. In each irreducible representation of pjl the numbers m, m' ...

are all smaller in absolute value than a fixed number K.
Indeed, from Schur's lemma we know that M2, which commutes with all
THEOREM

matrices of this irreducible representation should be a multiple of the unit

matrix
M2

= .[(2 I

where K2. is a constant. On the other hand,

M!+U: == M2_M: is, in our

system of axes, a semi-definite diagonal matrix, whose eigenvalues cannot be

negative. Consequently we have

K2_m l ~ 0

Iml

IKf.

(S.AI1)

The result is that in our search for irreducible representations of P'J 3

we ~n restrict ourselves to representations ~ which satisfy the condition
(S.A~l).
l~t

the operators M + ds and M _ dl act on the basis vectors "'rna;

the t;ransforrtled vectors M + "'nan ds and M _ "'mil dt stay ill \1 0 ' which is invariant. le !nay suppress the infinitesimal numerical factors ds and dt.
If we

If. M + 1 tPmll and M -1 ifJJfUI are eigenvectors of M:z and c()rrespond

respectively to the eigenvalues (m';-l) and (m-l).
The verification is done easily: (S.AS) and (5.A9) give:
T}lEORE~f

M:;M -11/1"." = M -1 (Mz t/lJfUI-- ifJ'M) = (m-l )ki--l t/ltIUI

MsM + 1 t/I".,. == M +l(M% tfJ.. + "'".,,) = (m+ l)M + 1 "'mit
M+l

(S.A12)

t/I,.. is hence a vector in the eigenspace of M. which corresponds to the

eigenvalues (m+ 1): it is a linear combination of the basis vectors

"'m+1,'

where (m + 1) is constant and n takes all possible values. As the last are only
defined except for a unitary transformation, we can take M + 1/Imn itself as
basis vector under the condition that we multiply it with a convenient conHt~~"n.t to n.orma.1ize it. The same considerations hold for ~f -1 V1"". in the

176

ROTATIONS IN 3-DIMENSIONAL SPACE

(Ch.5

eigenspace (m -1) of the matrix M Hence, if we now suppress the index "
which became
.Interpretation:

superfluous, we can give to our theorem the following

~ ~

If we begin with an arbitrary eigenfunction l/I_ of M. corresponding to the

eigenvalue m, we obtain, by iteration with the operations M+1 and M_ t a
sequence of eigenfunctions of the same operator: "'.~ 1 ';",:1:2 t/t_%s
corresponding to a set of eigenvalues which differ by a factor onc.
Thls sequence is obviously limited in both directions since the space Ito
satisfies the condition (S.Al1); hence there should exist a maximum valuej
and a minimum value k for which one has
M+l y,~

:a

0; M -11/1"

= Q.

(S.AI3)

For the other values of m we write

M-l

"'lit == a,... t/!..-1;

M+l t/I.. =- P.Y,_+l;

M.t/I", == mJ/!m

the ". and p. are numerical factors. Thls, together with (S.A7) and

(S~A13)

gives:
1

M+ 1 tPJ-l == -M+1M_ltPJ == -(M_ 1 M+ 1 +2Ms )t/lJ ==

-l/IJ ==

PJ~l"'j

1
'
1
M+.tJiJ-2 == -(M_ 1M+f.+ 2M,,)y,J-l = [PJ-tt!J+2(t-l)]Y,J-l
(l,j-1

ai-1

== PJ-2t/!J-1

From which. by putting

Pm == P.OC,.+l' we

find the recurrent formula

= Pm+ 2m

(S.A14)

{J",-l

Pj == 0, PJ-l

=:

2j,

p/_" == 2U+U-l)1, ... , Pm 1U+ l)-m(m+ I).

:II

But for m - k the minimum. value of m we know from ('.AI3) that

M+1M_l "'Ie - O. Hence Pk+2k = 0 and flnally

JU+ l)-k(k+ I) == 0

(S.AlS)

an equation that has two solutions. One is impossible: k =J+ 1, the other
k ;: - J is allowed. Hence, the sequence of eipnvalues of M., which we
have extracted in this way with the help of the operatOR M + and M _ , can be
written as j, j -1 t

U -1),

-j. They are symmetrically arranged around

Ch. S]

In

APPENDIX UI

the value m = O. The sequence contains 2j+ 1 terms and since this numlx!r
should be an integer, j is either an integer or a half-integer.

The cort'e'sponding eigenfunctions .p1' "'J-l,. 'tfi- j span a 2J+ 1dimensional subspace UlJ invariant lvith respect to the rotation group, since
to

the functions transform. among each other under the infinitesimal operators
M +1, M-l' and M% of the group. The space is irredu.cible since the operators
M+l and M-l iterated in a proper way, will transform each arbitrarily
chosen axis I/Im into any other arbitrarily chosen axis !/1m apart from a multi. .
~licative constant.
Hence the representation D J which we have extracted from D and which
is determined by the infinitesimal basis transformations M+ 1J M-l' Ms is
an irreducible representation of tJ} 3. The representation turns out to be
identical to the one that is constructed in 2 and which was designated by
the same symbol.
It is sufficient to put in (S.i\14)

Pta

==

(Xm+l

= JjU+l)-m(m+l)

that is

am =

.JiU+ 1)~(nl-l).

W"e again have equation (5.26) and we are sure that the basis functions are
normalized in the same way_
There is no other irreducible representation ~3 since for each representation D one can, by the preceding method, extract at least one representa...
tion D.i and if D is irreducible, it should be identical with Dj
Appendix

PROOF OF TliE FORMlJLA

DJ1 XDJl

(S~34a)

-= DJ1~J2+DJl-*XDJ2-t

(5.34a)

We have the following problem: The vectors in two unitary spaces are
subject simultaneously to the aame transformation of the unimodular unitary

group.

e~

= ~1 + Pf/l

'1~ = -P~l+a,*fJl

~2

= cx.~2 + P'12

'12 =

-fJ*~2+Ct'12

We consider the representations. D j1 , D J:). and

rx,a.* +

tJp* ...

Djl X DJ2

1.

of this group in

178

ROTATIONS IN

3-DIMENSIO~TAL

SPACE

[eb.5

the three spaces Dilt' DlJa and mil x Uiil . In these three spaces we take as
coordinat~~, respectively: the (2j 1 + 1) monomials
..

the (2j2 + 1) monomials

~Jl + m2~j2 - ml

_
q (i2)
2,JIlI2 -

",2

'0

Y (j;z + ml)!(il-m'l)!

and the (2jl +J)(2j2+1) products

Qml m l

q(}t) q(i2)

1,

rill

2,

~{1 +ml '1{! -in! ~~.1. +ml tl~2-m2

nil -

'

-V(il + 1tll)!(il. - m 1)!(i2 + mZ)!(j2 - m2)!

We have to decompose tIle space 9t)1 X laJ1 in nvo subspaces invariant under
the transformations (5.4), suell that the first: Dl)S+Jl is irreducible. To obtain
this result we will make a change of coordinates which replaces the Q"'I"'l
by linear comhinations of these quantities. Let us calli1 +j2 = J. We have
to find 2J~}- 1 independent linear combinations, which. transform 81nong
each other, without intervention of any other linear combination, under the
transfornlation (5.4). They form the basis for the representation DJ1 ';'j:"
If we take el = e1 =
'11 = tl2 = 11 the last system should reduce to the
functions

e;

qJp
Th.~s

is

~j,d.l:=~t Vle

eJ +P 1'/J-P

(P = J, J - 1., . , - J).

::-=:=-,--_.
~v (J + P) !(J' -- P) !
-J----==-.-

'.viII call tIle condition A . .From this we find that

(P.l

-=

y.,7,jt j:2.h,
'::-7

"'>}

---/.=:::::.:--

"VI~J
,&,

should necessarily makf.; a part of the basis of th.e ne'lN irreducible subspace,
since this is the only one whicll reduces itself to

c. 2J

i;;;

'\/21

if

= '2

and

"1 = 11

Let us suppress the normalization factor in the denominator in order to

simplify the formulas. We can insert it easily at the end of the calculation.
Since we know one of the variables of DJ 1 +J,. let us subject it to a transformation (5~4) and then we must find a linear ftmction of all the variables

.A.PPEt~D I.X

Ch.. 5J

179

III

of this representation. 1"'h'us in the usual notation

with
'PM = (2it)!(2j2)~

(2J)!

,_._--------------

1/~___ (J +~)!(J -M)!

. (5.Al7)
ml +m2=M f (j.1 -!-- ml)!(il - ml)!(i2 +. m2)!(i2 - m2)!

We state that the polynomials qJJ) l(JJ-l .... CfJ-J are the variables or basis
functions of the representation D j1 +i1 It is evident from (~.A16) that they
satisfy the condition A and are consequently llnearly iIlde~ndent. 1 Let us
now verify that these basis functions transform among
other under an
arbitrary transformation of the group %' 2

each

e~'
PJ~

ae

+ b111
= -b*el+ a*"l
==

(S.A18)

We have actually
with

= aCl.-b* {1;

B = bct+a* p.

(S.A19)

Hence
(ae~' + p,,~)2jl((Xe~ +p,,~)2J2

== (Ae 1 +B'11)211(A'2 +B'12)2Jl

or as a result of (5 . 1\16) and (S.A17),

4.4 J + M.aJ-"\(

a,J+MpJ-M

L -7}'I

"-

- _.

"(J + A-f) !(J -l\t/)!

XM =

where Xii is the transformed. of

L: ---- ..:.......::..-:'.:.=::::::::.':':'~
hl

;:.. ...

!p 'AI

.J(J +M)!{c/-M)!

by the operation (S.AI8)_

If we take (5.AI9) into account,~4J+M.BJ-M becolnes a polynomial of the
degree 2J in (X and p. Tllese two nurnbers are arbitrary. lIenee it is sufficient
({JM

This condition made us choose the particular

equatiom (5.1\.16) and (5.i.. 17).
1

group~ng

of the roefficients in the

180

ROTATIONS IN 3-DIMENSIONAL SPACE

(Ch.5

term

to identify in the last equation the twQ sides

by term in order to obtain
XJI as a linear combination of !pJl. Tltis verifies that the !PM form the variables
that are th~ basis of the representation Di, +j'J,
As for the variables of DJs-iXDJ2-i' these are obviously expressions
p ....~ analogous to Q"1~2' but of degree 2jl- 1 in ~1"1 and 2j,.-1 in
Hence these arc not linear combinations of Q"'l1f1,. since these are of higher
degree. In order to reduce the first to the second, it is sufficient that

'2"2-

~1.2 == ('1112 -'11 '2)P,"1"''-

II

(5.A20)

These functions are of the dearee wanted and transform under a substitution
(5.4) exactly like the quantities P1II1"''j since the first factor of the right-hand
side is an invariant. We have indeed

'1"2 -"i e;

== (~l '12 -'111 '"X + Pfl*)

-= C1 '12-"1 ~2'-

In this way we have found the basis functions of D J1 +JJ and DJ,-,.XDJ2-i.
We still have to show that each linear combination of expressions
Q -2 can be expressed in an unambiguous way in a linear combination of
ipJl and Q':

(S.A21)
where t/I is a linear combination of P",.m2' that is a. homogeneous set of
polynomials of degree ('i1 -1) in '1' '11 and (2j2 -1) in ~2"2.
Tb.tr necessary number of arbitrary coefficients, are provided easily since
we have,
(2Jl + 1)(~2+ 1) == [2Ui +j2) + l]+2Jl 2}2*
He~ce it is sufficieJit to show that the different terms

of (S.A21) are independent, that is that 4> cannot be identically zero unless all coefficients in the
risht-hand side are equal to zero. Let us first take ~1 ::II '2' til == 112; the
last term in (S.A2l) is zero fP == 0 gives aJ := == a_J == 0 as the CPA{ are
independent. The QJI being zero, let us make ~1 :;.: (~t '11 ;' '12; tP == 0
implies necessarily '" == 0; hence the decomposition is unambiguous.
The representation D}t-.. XDJ2-i can be reduced in the same way as
DJI XDJ2 We &raue as before; taking (S.A20) and (S.A21) into account
we see that it is suftlcient to look for the coefticients in the development of
(~l!lZ -'11 C2XrX~t +"'1)2Jl-1(<<~a+J"JJ,h-l

(S.A22)

in order to find the linear combinations of Q.1JW2 which serve as basis oftb.e
representation Dji XDJ2

181

APPENDIX III

Ch . .5]

If we operate this way, step by step, we succeed in expressing the decomposition (5.34) of the representation Djl XDj2 in an explicit form:
(we suppose' ~that j1 < i2)
tl> = l/Ijt+iz+(elYJ2-f11'2)t/!jl+Jr-l +
8

.....

+('1112-"lel).t/lll+J2-J. ....

(5.A21a)

which is immediately clear by comparing it with (5.A21) Let us put J =

jl +j2 = A. The linear combinations qJ~ of Qmlm2 which serve as a basis for
the representation D)l +j2- A can be obtained by developing the function
(~1 '12 -111 e2)).(a't + P112)2jl-l(CX~2 + P1I2)2 J2 -A

(5.A22a)

and since we have

(e

AI
171 ~2Y = l'~O ( -1)" v !(J. ~ v)!
A.

1112 -

e:-""I e; 17;-"

we find, by a calculation similar 1 to the one by which we arrived at (S.A17)

<P~=PJ

}2

1Hl+mz=M

= flJ :}2

}2(-1)"
v

C~lml Qml m2

(At! = J, J -1,

. ., -J),.

(5.A23)

1nlt-m2=M

PJ is; a normalization constant which depends only on J, that is on

it,

i2

and A.. The index v varies in principle from 0 to .A., but one has to put
all ttfrms that contain a negative factorial equal to zero.
The coefficient C. t m2 is called the Clebsch-Gordan coefficient, the Wigner
coefficient, or the 3-j symbol. To calculate its numerical value is a tedious
task; fortunately tables are available (see references in Chapter 6).
The formula (5.A23) solves completely tIle problem of the decomposition
of the product Djl XDJz " It takes the siInple form (S.A17) if A = 0 and it
similarly simplifies in the opposite case: A = 2j2) J = il -j2 Ul > i2)
1 The binomial coefficient of
('"I.:':

,,'-"1

(OC~l

-t-/3'Y!/1 )2.1 -"A

+{J'YJ1)2 J - A contains
-- "

-..:....I

(2j k-.;,)

a summation index, say k

('ft.): )2.J-A -Ie

,=,1

(R1]
)k
P 1

in order to have the total power of';1 equa.l to j~tm we take k = j-'P-m and restrict the
summation to positive or zero values of this quantity., Similarly for ~2' 'I'J,-

182

[eh S

because" can only take the value

Cit ~;i2= (-1 )i' +m,
ml1n2

v :.-.:. jz + r1'l2 and one has

,--"

--

~,_~_.(j l-=!:,n 1~ I,j 1 -: mll!__.____

(i2 + "~2)!(j2 -- 'n2)!(J -to M)!(J - M)!
_.> __

"

-We can now give the equations (5.A21) and (S.A21a) a definite form, dis.
playing clearly the decomposition Dit X D i1. in its irreducible eienlentS., This
is 'the so-called Qebsch. . Gordan series
IJ1- Jll
L.

" "J
') PJ CimJ m Qmlml =
=J=Jl+J214=-J
t;.", aM
ml+ml:;aM
J

JMtl

~"
J
J
LJ /-J aMCf'M"

(5.A21b)

...f .M

The vector ifJ is an arbitrary vector of the space 91il X 8fJ:, that is a linear
combination of Qrnl m2" The 0-1- depend on the fornl of this combina.tion~
The PJ are normalization factors and the C;'tm:t are the constants determined
by equation (S.A23). The tpi, are the orthogonal vectors that span the space
It}I X ,all' those which have the same index span a separate invariant
subspace which transforms according to the irreducible representation D j
Their mutual orthogonality rt~sults from the orthogonality of the different
irreducible representation of the saIne group 1 (.1#3 for the indices J, !?d2 for
the indices kJ). We can write

But the Qm lm l = qf.l1 qf.~ are also orthogonal, sin(~e the fa\~tors q~\ and
q!:,. are;. Hence one can, keeping M = ml -}nml constant, consider (5 . A25)
with Q""ml. The transformaas a ~tarytransformation-connecting film!
tion Imatrix is B ==
C!t_a)' where J labels the rows and ml the columns.
W~'can revene this transformation by expressing Q"Hml as a linear cODlbina~~~9f . ~~. Since B is unitary and real the inverse matrix is equal to the
transposed matrix, hence we have

--2

Y'J

(5.A26)
The preCeding calculations have t~.en made V\tith the help of a partkwular
choice of basis functions lJ'!.:,., q!,.~ and QJftJm% of the representations DJ1 , Dh
andDJl ~I!h. But the results obtained are all linear hence they de.pend only
on the representations and all stay true no matter ~'hich functions we use
as a basis. In particular, we could repiac;e certairl ~ by 11;" and c.e~rtain 11 by
11;bS ltatementis only true in case every rel'resentation oo...'UfS only onco, a C".ondftioil

fulfilled for the fuJI rotation

grOUPt

but. not always for a 11nite ,roup,

183

APPENDIX III

Ch 5]
4

-,*, which do transform in the same way (compare 2.2). We can give in
this way a simple geometrical or physical meaning to the basis functions.
They could be
.. for example the Laplace functions, or more generally the eigenfunctions of an atomic problem. The formulas (S.A17), (5.A21b), (S.A23),
(S.A24), (S.AlS) and (S.A26) always apply_ This statement forms the basis
of the proof in Sec. 9 about the selection rules and the calculation of the
intensities of the Zeeman effect. In the majority of the applications i2 == 1

and

'
DJXD t == DJ + 1 +DJ+DJ -

&

The possible values for m2 arem2 == -1,0, + 1, henceM=ml + 1, mi' ml -1.

Eq. (S.AI7) (S.A23) and (5.A24) give the cottstants in the following table:
TABLE

S.A.l

J+l

yu+m+2~U+"'+1)

-V2U+m+l)(j-m}

)-1

'vU

-m}<-;,-m+l)

y.(j-m+2~(J-m+ ~

VU+m+"l)U-m+!)

V2U+m)(i-m+l)

2m

I-vu+m)(i~m)

l/U+m)U-t m_"

1)

For further tables see the Systematic Bibliography 7.1.

aa~ng once established the Oebsch-Gordan formula for the decomposition ot the basis into its irreducible components it is very easy to perform
the dicomposition of the representation into the irreducible representations
it cdntains. We pointed out that the coefficients CJ , for fixed mlm2' could
be' looked upon as a unitary matrix. Hence if we want to transform the
matrices with respect to the new coordinates we have to apply the similarity
transformation equations (1.7) or (1.27a). Explicitly we have

This formula forms the basis for the derivation of the Wigner-Eckhart
theorem and will be used in Chapter 6, 3, equation (6.17)~

CHAPTFR.

CONTINUATI.ON OF THE THEORY OF THE ROTATION GROUP

1. Irreducible Tensor Operators

At the end of Chapter 2, 3 it was stated that if angular momentum is a
good quantum number, then polynomials of angular momentum operators
can be constructed. These polynomials will undergo certain transformations
among themselves if the system undergoes a finite rotation. The angular momentum components form. a vector, so the polynomials are tensors, since they
are products of vector oomponents (or linear combinations of such products).
The reason that we prefer linear combinations of products, instead of the
products themselves, is obvious. These tensors must be defined in such a
way that they transform like irreducible representations in configuration
LxL" LyLx ... contain one
space. For ~xample, the nine products L;,
invariantL;+L;+L;, as well as three linear combinations that transform
like a vector (LxLy-lj,L~J -+ L z , ... , hence there are actually only five
oper~tor polynomials of rank 2.
Mfer this preparation we introduce a set of angular momentum operator
pol~~s w~ch are usually called the irreducible tensor operators. They
wer9'ongmally tntroduced by RACAH [1942, 1943] (see also F ANO and RACAH
[1~9]), with the following properties:

L;, L:,

Dop

r: =

..t

D pp' TIl~

p'= -.1

(6.1)

T;

that is under a rotation of configuration space, the component transforms

like a spherical harmonic
The "quantum number" ). labelling the irreducible representation is always an integer. This is because the ope~a,tors T
related to physical observables which do not tolerate the ambiguity in
sign inherent in half-integer representations.
It is,oot necessary that the operators T be Hermitian. The conventional
gives a basis analogous to the spherical harmonic basis
choice for the
functions. The spherical harmonics are of course a specific choice out of
many possible equivalent bases. They happen to be complex, although

Y;.

are

T;

184

Ch. 6, 1 J

IRREDUCIBLE TENSOR OPERATORS

185

equivalent to real ones. The practical advantages of complex exponentials

over separate sines and cosines are well known in mathematics, physics and
electrical engineering. The counterpart of non.. real numbers are nonHermitian operators. F'or instance, x+iy will correspond to JxiJy , a
so-called escalator matrix, which has only zero elements on one side of the
diagonal.
Explicit construction of the irreducible tensor sets on the basis of the
definition (3.1) can be done in different ways . rfhe simplest would be the
Schmidt procedure used for the construction of orthogonal polynomials.
Start with a constant and the tl1ree tensor operators of rank 1, J 0 and 1 1,
and successively construct polynomjals that are orthogonal to these four.
The orthogonality is determined by the condition
Tr PiPj

= Jtj

where PI and Pj are any two operator polynomials. 'The orthogonality

provides independency. But if the polynomials are independent, that does
not insure that they form a basis that is irreduCIble. To avoid this uncertainty,
start the construction in tIle same way as Legendre polynolnials are constructed and gf!neraliz-e- to spherical harmonics, i.e. first take the p, =' 0 case.
They do not contain 1 l ' only 12 and J %" (The non-operator case contains
and Z, but not x+iy.) For successive A-values we find

,2

,-

A = 0 : constant
A.=l:Jz
1 = 2 : 1;-!J2
A. = 3 : SJ;-3J 2 J z +lz ..

We see that for ). :::; 3 the polynomials are different from the spherical
harmonics. The spherical harmonics are homogeneous in x, y, z of degree A..
The operator polynomials contain-terms of the degree A. and lower. Actually,
the concept of "degree" does not have meaning in this case, since with the
insertion of a commutator J z = (J xl y - J yJx) the degree can be increased
in a rather arbitrary manner.
Instead of pursuing the Schmidt procedure for It =F 0 a second, heuristic,
but sometimes more effective way of constructing the p~lynomia1s will be
indicated. Take a spherical harmonic and replace a term like
/

2xy
i.e~7

by

(.1 x ly+J yJ x )

take all possible arrangements and divide by the number of arrangements. 1"'his process leads to lengthy expressions which can be shortened

186

THEORY OF THE ROTATION GROUP

[Ch. 6, I

and. systematized through repeated application of the commutation relation.

The third method (compare EDMONDS [1957]) for the introduction of
irreducible *,nsor sets that transform like spherical harmonics, is based
directly on their transformation properties. It does not actually give the
operator polynomials but their commutation relation, as well as their matrix
representation. Since we are, for all practical purposes, interested in the
latter, this method is the one that is most widely used.
'
The derivation is based on equation (2.19), which ca.n be "Written as

(6.2)
Its formal integration leads to an expression for the results of a finite
rotation (J) on the wave function "'(0):

e -ill-l / . CI) 1/1(0) == t/I((J).

(6.3)

Using Uris unitary operator for the transformation of the tensor-operator

under a finite rotation, we have:

sTs- 1 :: e-'I-lJ4C1)TeiA - 1 / - W = T-ili- 1[J, T] . . ..

(6.4)

On the 9ther__hand, the right-hand side of (6.1) can be series developed in a

similar way (compare Chapter 2, 3)

(6.5)
Comparing coefficients gives us the relation
;.
[It, Tp]
=

.LJ (/:)I'~I T""

""..

(I == 1, 0, -1).

(6.6)

Fro~1he explicit expressions for (1)p.p" well known from the Zeeman effect

(eq.lS.48), we find a set of commutation relations:

[/lt TI']

= T;l IiJ(p=F..iXlIt+l)

(6.7a)

[10' TJ = T;hp

(6.7b)

which are equivalent to the set of equations originally introduced by Racah

in 1942_,to define the tensor operators.
.
In the case of finite groups, the tensor operators introduced above are in _
many cases no _~9naer irreducible. There will be, of coursc, only a finite set of
representations. A general formula cannot be indicated, For the different
finite groups, the irreducible operators were determined by KORRINGA [1954]
and KOSTBIl [1958].
Since the irreducible operators T art:! determined by their rotational be-

Ch. 6, 1, 2j

REPRESENTATION OF TENSOR OPERATORS

187

bviour only, it is possible to indicate their matrix elements with respect to

an irreducible set of wave functions without any further specification, except
for a part,- USually referred to as the "strength" of the interaction, which
does not depend on the geometry. This part will stay undetermined in the
form of a number of constants, which do not depend on the axial quantum
numbers m or Jl.
2. Representation of TeDS8r Operators
The use of group-theoretical methods has been most beautifully demonstrated by the so-called Wigner-Eckardt theorem (WIGNER. [1931, 19S0],
ECKARDT [1930]). The theorem states that if there is an invariant coupling by
a bilinear form, such as spin orbit coupling:

L S == LxSx+L,S,+LzS.
(6.8)
the matrix elements can be calculated by group theoretical methods, ex~pt
for propotHonality factors. In ~ther word,s the ratio between the matrix
elements of such an invariant operator can be determined.
This separation of matrix elements into angular or geometrical versUl
radial or physical parts, is found in many different forms and the Wigner. .
Eckardt theorem is actually an example of many similar applications of
the following idea:
Suppose an operator 0 transforms according to a given (irreducible)
representation of the rotation group, say l and we want to determine its
matrix elements with respect to two sets of basis functions. One set transforms
according
to the irreducible representation j 1, the other according to i2.
,
This ;idea is not quite new, siIlce in Chapter 5 we treated this problem for the
special case of 0 being a vect<>r (i.e. ~ = 1). Given these irreducible represen~Uons the matrix elements will of course transform according to the reprentations contained in the direct product of the three rC'presentations
I

ril xr~xrJl = L, r..

(6.9)

Usually some of the r /8 will occur more than once. 1 The general idea is
that the unitary transformation which performs this reduction is given by
group theory (actually the theory of group representations) except for a
1 In finite IfOUPS a multiple occurring representation may already show up in the direct
product of two representations, in the rotation &rOUp they are sinsly occurring
(Oebsch.Oordan series). With the direct product of three irreducible representations or
the rotation group, there are always multiple occurring representations, except in trivial
cases like it =- ,0. Compare 4 on Racah coetllcients.

188

THEORY OF THE ROTATION GROUP

reb. 6 2

certain ambiguity due to the mUltiple occurring irreducible representation

In order to determine the matrix element, we take the integral over the
left-hand side. The integral can be written in polar coordinates and hence
co~sists of an integral over the unit sphere and a radial integral. If the integrand undergoes a rotation in configuration space, the integral will have the
same value (provided of course that we transform both operator and the
wave functions; a rotation in configuration space "induces" a rotation in
function space - compare Chapter 4, 1) since such a rotation only amounts
to shifting the intervals of <p and 8 on the llnit sphere. We conclude that the
only representation on the right-hand side which will give a non-zero result
is the unit representation r 1 All the others will average out under spherical
integration, and the only one left, the unit representation, needs no angular
integration. The integral for this term is 41t (the surface of the unit sphere)
times the radial integral.
Let us consider th~ left-hand side of the symbolic formula (6.9). It stands
for coll~tion of products of wave functions and tensor components. There
are (2m + 1)(2J.l+ 1)(2m' + I) possible products. Instead of integrating, we
write them first as linear combinations:

(6.10)

such that, they transform like irreducible representations. That is, hi. is a
~Jc~~ ~ the represen,tation is one-di~ensional), o~ a certain number of
bt,a,./qrm (in case the representation has more than one dimension) a basis of
an"'irJ/educible representation. If we integrate now, we find that all integrals
ov~/b, are zero, except bi which corresponds to the dftit representation.

'l:

pmm'

a!!,.,(ml0I'Jm')

= 41tc5 1i Jb 1 r 2 dr

01i

B t

(6.11)

If we take the inverse of the matrix a~~m' (i labels the columns and the triple
index labels the rows) we have the matrix elements expressed as a product

{6.12}
of'the radial integral Bl and a factor which depends only on the reduction
procedure". Hence this factor is entirely determined by representation theoryl
We went through this argument in order to stress two points. The first
the, question how many constants Bl are involved. This is now easily a~ .
~wered. The proof above was given for fixed values Ofjl,j'J, and A. We have ~Q
repeat it for every other
of values 11 and j2. How many possibilities

is

pair

Ch. 6, 2, 3]

WIGNER-ECKARDT THEOREM

189

does one have for a given A? That answer is given by the triangular rule,
which is of course the same condition as demanding that (6.9) will contain
the unit representation. For instance, for A == 1 the vectors i1 and j2 may
differ by 1, or o. Hence tlfere are three different constants Bl for every
given value of il (or i2). These three, or in genera121l+ 1, constants depend
on the radial shape of the wave functions and on the radial dependence of
the operator.
Since it is possible that there is more than one wave function with the
same jvalue, another quantum number t% may be added in order to distinguish
between the different (radial) wave functions. Hence B depends onj1 ,j2' ill
and (X2.
The second point is the remark often overlooked in importance that for
given a1 and/ (X2 the unit representation occurs once (or not at all). The
possibility of multiple occurring representations has already been mentioned
in a footnote. This is not the case for the unit representation. It occurs only
once, since the direct product of two representations contains each irreducible representation only once.

r). xrJ1

Lr

(i = l+ j1 , .. , lA-ill)

(6.13)

and the only way one can form r lout of the direct product of r j1 X
by taking i = j2. 1
This is not true for finite groups.

r i is

" 3. Wigner-Eckhardt Theorem, Redaeed Matrix E1e...ents

If the wave function and the operator all refer to one type of function,
the/theorem in the last paragraph is rather trivial, since the only invariant
operator is the operator with A. = O. In physical examples it is nearly
alwaY'S the Hamiltonian or parts thereof. The theorems tel Ius that the energy
states are, characterized by certain constants (the eigenValues of H) which do
not depend on m or m'. From ,the triangular rule one concludes that j = j'.
The only thing that is of al:lY interest is the statement that for j = j' =F 0
several matrix elements have the same value: a (2j + 1)-fold degenerate
level. This case is already adequately coyered in Chapter 4. The Hamiltonian
was sup~osed to consist of two parts, the kinetic energy operator and a
potential energy, a static field essentially. It is, however, well known that the
1 This last step has to be more carefuJly formulated if the representations are of the
second kind U is half-integer). One has to take the antisymmetrical product of r i and the
spin conjugate of l ._

190

THEORY OF THE ROTATION GROUP

[Ch. 6, 3

Hamiltonian many contain other terms, which could be called dynamio.

For instance terms like spin-orbit coupling or similar vector couplinp
well knoWll~in the theory of the atomic spectra (CONDON and SHOR.TLIY
[1935]).
These products are scalars and hence invariants, but they are made up
of vectors, that is three operators that transform as the coordinates. af a
point. The complication arises from the fact that one vector operates on ono'.
set of functions (for instance the angular momentum operator components
act on. the orbital wave function) and the other vector on a different set of
functions (the spin functions). Hence the wave functions have to be the
product of two different functions, say the space and the spin functions.
Generally speaking, this does not have to be literally true, the total wave
function has only to trfJllSjorm like the direct product of two sets of wave
functions. In case the total wave function is n-fold degenerate we againfiad
that n levels are degenerate, according to the Wigner theorem.. One can,
however, be more demanding since it turns out to be possible to calculate the
matrix elements of each of the six components of the scalar product in
the n-dimensional function space without any knowledge beyond the transformation properties. This procedure will be described below and is based
on the repeated use of Clebsch. .Gordan coefficients. Instead of considering
the vector coupling first and generalizing this later to tensor coupling we
may ju~t as \vell give the description in terms of tensors right from the
beginni~g.

The tensor coupling expression is very similar to the scalar product of

two ve~tors~ For practical reasons we write the scalar product
,~

bx +a,b,+'Q7,b= ==

b - l +aobo+a-tb+l

(6.8')
where al == 2-+(a.ia,). In the same way we can construct invariants out
of -tensor components provid'ed
take irreducible tensors and label the
components in a proper way. The construction ofirreducible tensor operatob
a priori 'has been considered in the previous section. It is only necessary to
define them ~ere as quantities that have the same transformation propertlel
Q3 the spherical harmonics. The general form of an invariant coupling will be
Dx

Q+l

we

01=0

== H coup1 ==

p.c - l

T:(1)T~p(2),

(6.14)

where the first operator acts on one type of wave function, the other on another type of wave function. These were originally the spin and orbital wave
functions; we call them 1. and 2. They may have irreducible representations

Ch. 6. 3]

WIGNBR..BCKHARDT THEOREM

191

'Characterized by i1 andj2, (and ml and m2). These representatiollS usually

are reduced,of \ i.e. written as a set of wave functions with total j U = i1
+i2' .. OJ, 1ii'-j21). Hence the matrix elements of (6.14) depend on many
labels:jlml,j2m2,ji, mi,j2 m i and} .. )', m == m' as wen as It and A. AU
unprimed quantities refer to the initial state, all primed to the final state.
Again, we can show that these matrix elements are a product of '~constantStt ,
that is, factors which depend on the different j's and A, times expressions
containing the m's. This m-dependence is completely determined by the
representation theory.
We will give a .proof slightly different from 2 as follows: A trans~
formation of configuration space of an irreducible tensor operator gives,
according to Chapter 1, eq. (1.7):

<ls- 1 T;I> == (18r,;'S-11).

(6.15)

This equation states that a transformation in co~guratiOD'Space is . eq1(liva-

lent with anotIt,r wana!orma,tioa in function space indicated by t4c; ,oper,.~()rs

S. The second one 18 the inverse of the transformation induced in function
space by the tirst(oompare Ch4lpter 4, 1.2). The proof of equations (6.7)
was also based on the same idea.
If we write this in components we. have

<jkIT"lj'l) == (kIS-1fm) (mIST"S-lln) (nISll)

== (kIS- 1 Im>(mls- 1 Tp Jn><nISII) ==

L
AD;./ JD;;n1 j'Dnl(mITyln).
wnn

(6.16)

SQ f we have done nothing but explicitly expressed the fact that the bases
and the tensor. components transform according to irreducible representations. We can now, using again the fact that all rotations are determined by
the 'ame set of Euler angles, apply the Clebsch-Oordan reduction formula as
expresse4 in,Chapter 5, Appendix III, eq. (S.A27),that is we .can~~ethe
product ~AflJt by choosiq a linearcombillation. such thateach.Gf~these
transform according to the representations j" == J.+j, A+1-1 .. _IA-jl:

(6.17)
and after this substit~tion we,let the parameters (Euler angl~) contained in
S run through all pOssible values. We integrate over theSe variables and
apply the orthogonality relations between the matri"t elements (see Chapter
3). As a result (6.16) reduces to:

THEORY OF THE ROTATION GROUP

192

.;

,
'"

[eb. 6, 3

A
= Ci',i
<5,,+m, n(jml7:IJ.'n).
bIl + i, I "L.J Cl',}A
pk
\pm
2 ., 1
"

'J

mnv

(6.18)

The last sum is independent of m, n and v and we denote it by

L ct~j). ~,,+ ....

VIITIlj') =

2j' + 1

m, ..l'

VmIT"Ij'n).

(6.19)

The result is that the ratio between the matrix elements (jkIT"lj' /) for different k and 1 values is completely determined by group theory. The proportionality constant depends only onj,j' and whatever other quantum numbers
are used to describe the system. The double bar matrix element is called the
reduced matrix element. It is at this moment that we see the full importance
of the Clebsch-Gordan coefficients in physics.
Introducing the notation 1 used by CONDON and SHOltTLBY [1935]:

(j1 m llr;Um ) =

l)A-it +j

(AJljml.tjit m l)VlIT).fj).

(6.19a)

2i1 +1

This is the Wigner-Eckardt theorem in its original form and by using

equa~ns (6.14) and (6. 19a) we have for the scalar product Qftwo irreduci.ble
.tensor! operators:
,

(rxjli2,JMI

T;(l) T~"law'j~j~JM>

I'=-A

= L cJ , J1J2CJ, l'tJ'2(j1 ml1 T;(1)lii mi> 02 m2'T~Il(2)ljl m;>

= L c" J1J2CJ, l't}';z cl't , itAC}'::, i:: A(j1I1T(1)lIji> V2UT(2)lIj;).

(6.20)

We will come back to eq. (6.20) later in 4. Eq. (6.19) has a wide range of
applicabilities; for instance transitional probabilities, that is, matrix elements
in which j' rI= j. One can, of course, also deduce transition probabilities from
(6.20) by taking as one of the tensors the field that causes the transition.
Besides the application of (6.19) to the calculation of the dipole moments"
1 Compare EDMONDS [1957]. This monograph contains a very practical table that
indiCates the connection with all other notations. Our Clebsch.. Gordan coeflicit"ftt is
related to Condon and ShortIey's by:

J
CmImi

==

(jlj"JMlil mJ.m,,).

Ch.6,3,4]

RACAH COEFFICIENTS

193

which will lead us again to the expressions (5.48) it is also easy to determine
the corresponding equations for the higher multipole moments.
Another application is made in crystalline field theory where the potential
energy is replaced by a linear combination of irreducible tensors that have
the same symmetry. The matrix elements of the energy and, after diagonalisation, the eigenvalues can also be determined with the Wigner-Eckhardt
theorem. The result contains a number of undetermined proportionality
constants. This is not surprising since the replacement Hamiltonian fulfilled
the necessary condition of symmetry, but is not sufficient to describe the
physical situation adequately.

4. Racah Coe8icients
The Racah coefficients were originally introduced in polyelectronic atomic
spectroscopy in 1943 and subsequently found a widespread use in nuclear
spectroscopy_ They are also useful in solid state work. If the crystal symmetry
is introduced through an operator-Hamiltonian, sooner or later in a physical
calculation the coefficients in front of the different terms of the operator
expression have to be calculated. It was realized that these coefficients were
similar to the reduced matrix elements and hence connected with the Racah
coefficients. In the following pages we will follow Rose's treatment.
The idea ~f a Racah coefficient is as follows: If two electrons are coupled
the qebsch-Gordan formula states which irreducible representations are
contailned in the product representation, and that each of these representations is contained only once. However, if three electrons are coupled the
irredncible representation of a certain finalj-value (total angular momentum)
may occur several times. This indeterminacy is shown by couplingjl withj2
into j' and coupling j' with i3 into j. As a result there is a set of mUltiple
occurring irreducible representations. However, if i2 and i3 were coupled
into, say, j" and subsequently j" and il were coupled, then the resulting
bases would have been different. Of course the product representation will
contain the same set of representations each with the same multiplicity.
Since any linear combination of equivalent irreducible representations is
again an irreducible representation of the same type, the two different
coupling schemes may give different results for the bases with respect to
these multiple occurring representations. The Racah coefficient gives the
relation between these:

t/!J, m(viaj') = L Rj,)"t/lJ(viaj").

j'

(6.21)

194

THEORY OF THE R.OTATION GROUP

[Ch. 6, 4

For instance take jl == 1 and i2 == 2; then jt = 3, 2, 1 and combining

this withj3 == 3 we havej == 6,0 (once) 5, 1 (twice) 4, 3, 2 (three times)..
HoweverJi+13 == j" := 5, 4, 3, 2, 1 and combining this withj2 gives the
same set. A certain possible value of j say j == 1 comes in the first case from
j' == 3 and 2 and i.n the second from j" == 2 and 1" If we also specify the
value of j' from which j == 1 is derived, the set of functions thus obt$ed
will be a linear combination of the functions coming from j" == 2 and from
j" == 1.

.- -4
Plg. 6.1. The coupling of four ansuiar
momenta.

Fig.. 6..2. The matrix elements of a scalar

bilinear operator compared with tho
previous scheme.

By rworking out both coupling schemes the coefficient R can be expressed

in tetins of a product of four vector coupling coefficients. Some of the summati9ns can be worked out explicitly; but one still.remains in the final. expre,sion for the Racah coefficient. It is convenient, although to a ~rtain.
e~nt misleading, to use a vector diagram. Fig. 6.1 expresses tIle fact that:.
j' ==Jl+j~; j" =i2+j3 and j==il+j2+j3 (the third possibility j'll ==
il +i, does not lead to anything new).
The coef6.cient.~" also be used in describing the coefficient in front of an.
invariant couplQJ.a operator (6.14). Suppose we have a wave function cl1arlc>;
terized by-j and suppose we break this j-value up into i l l and
The,
tensor operator will take j into ilf' or in other words, the~ tensor opera~j
has non-ze.t:omatrix elements between thesestates. The same holds for,~~;
part of
bilinear invariant that acts on J2. The final result is a set Q(
~. ~
ii/ andj2/ which will be brought together to jffnale Since the tensor cou.~~
chosen was invariant, i.e. T'(l) transforms contragradient to TA(2), the
resultinglfhlal will be the same as the initial one. However, the way in whichj
I

l'

tlic .

J",.

Ch. 6, 4]

RACAH COEFFICIENTS

195

was decomposed intojl' andjli is not uniquely determined, and an the different values of jl1 (to each of which belongs a certain i2') will give a contribution to matrix elements of the bilinear operator. Each of these contributions
has a coefficient which is completely determined by the rotatiow transformation properties. They all have the same proportionality constant: the
"strength" of the coupling. The result of the decoupling and recoupling is
equal to the product of this "strength" constant, denoted by (IIT"(I)II)
(11 TA(2) II ), and the Racah coefficient. Hence, if we take the contracted

product of two contragradient tensor operators, like (6.14), the dependence

of this invariant on the quantum numbers j is detennined by a Racah
coetlicient.
The Racah coefficient can be expressed as a sum of products of four
Clebsch-Gordan coefficients, following the recoupling diagram of Fig. 6.2
The first coupling it +i" == j' is expressed by

(6.22)
and coupling the resultant with j3 gives:
'"Jm(j') = ",tL C~''''3 "'i'm' 1/1JJ m3

(6.23)

m3

(m' + '"3

=fII)

Substituting (6.22) into (6.23) gives,

"'JII.(}') ==

L L C~'J m-".,C~1t m'-fIIt"'},,,.! tJ!Jlm, ... ",tfrJsm-m'

(6.24)

"' .....z

In the same way "'J,.(j") can be found and (6.21) will take the following
form

(6.25)
Multiplying by JjJ )11'1 t/I J2Jj;J "")IIA' will give a set of ~ functions at the lefthand side which make ml =: JJl; m' -m:t == Jl2 and m-m' == 1'3. i.e. there
are only two summations over the ",'s since the third is determined by
P1 .+- fJ,,. + Pl == m t-tnd similarly on the right-hand side. The result is:
li c'
. . ,~ R ii" CJI'lJj3
" "..J
e"'~"lJA1;
Pl+Pl.l'l =: 4.J
\';,J1.tl'::+Ill "
1"

'II

orthogon:?~tity

of the C,.,..G. -:;oefficients (Chapter 5, Appendix III;

note that orthogonaJity is \vith respect to the superscript and one of the

Using the

196

THEORY OF THE ROTATION GROUP

[eb. 6, 4

subscripts) we find

(6.26)
~.

'"

The resulting coefficient depends onjl,j2,j3,j,j' andj", but not on m,

since the position of the tetraeder of Fig. 6.1 in space is irrelevant. The Raca.h
coefficient is usually defined as
21J.. 3 ; J.,J.") = [(2J + 1)(2"
W(111
(6.27)
'J + 1)] -tR j' j"
The equation (6.26) and the two summations can be reduced to one by some
complicated algebra. For this we refer to the original papers of Racah.
[t has become customary to replace the W-coefficient by the so-called Wiper
5-j symbol:

{ ~1J3 ~2] J~:,}

= (-1)il+i2+J3+iW(i1i2jj3;j'j").

This notation is convenient because it can be easily generalized. The 3-j

;ymbol or symmetrized Clebsch...Gordan coefficient and the 6-j symbol are
the first two members of a family of 3n-symbols. They refer to the coupling

Fig. 6.3. Addition of vectors of given length and given z..component.

Ch. 6, 4)

RACAH COEFFICIENTS

197

of 2 angular momenta (n = 1), 3 angular momenta (11 = 2), 4 angular

momenta (n = 3), etc.
The W-coqf6.cients have a number of symmetry properties, which reduce
the number of entries, but their calculation is nevertheless rather tedious
particularly for higher angular momenta. For very high angular momenta
they approach the classical vector formula. Fortunately a large number of
tables is aVailable.
Finally we WOUld like to make a comment on Figure 6.1. Instead
of thinking of angular momenta as classical vectors, an inheritance from
pre-quantum mechanical days, we can imagine them as cones produced by
rotatin, vectors. This picture helps the imagination since it stresses the
fact that J% and 17 are not determined, but that 1 and J. aro. It
only
necessary to add the "numerolo8Y" that is the fact that J. is an inteser and
12 == 1U+ 1) to complete the situation to a correct description. It is obvious
from the picture that for given total j and m the orientation of the vectors
jlj2 and i, is not completely fixed. So that for a particular i' == 11 +il,
a number of different values for j" can be assigned. The main conclusion
from Fig. 6.3 is that Fig. 6.1 has to be regarded as an tetraeder instead of a
two-dimensional-diagram.
-In finite groups the Racah-coefficient becomes quite complicated. The
first coupling may already contain multiple occurring representations, that
is the aebsch-Gordan coefficient needs more indices than the rotational case.
The mafn practical difficulty, however, is the lack ora closed formula for the
basis function of the different irreducible representations.

is

CHAPTER 7

SPACE GROUPS

1. Outline
In tJUs and the following ,chapter we 'will display those groups that are
used in !lon-relativistic physics in general and in quantum mechanics in
particular.
The most important group is the full rotation' group.;,The exte~sive treatme~t qf the r,~presentations' of this group was gi~eIl.in two of'thd previous
chapters~Th~~fulrrotation group is usedin the description ot.isolated,atoms
and ion~. Itaiso pJaYS a role in the ~asic description of tp.e wave function
in general, irregardless of the symmetry, since the representations of the
r"II rotation group serve as a starting point for almost every calculation in
quantUJll,. mechanics.
.
In dea.liJ1g with crystalline solids, however, another type of grouphasto
cOnsi1ered. ,The most strikin~ property of a crystal is the re~~rity of its
structjure.
. If this structure is perfect, i.e. no errors, infinite in siZ .and free of
temMrature agitation, this property can be expressed 'by saying that certain
tran~ations are allowed. It is to be expected that a number of physical propet1ies are invariant under the operations of this group and so it would be
interesting to study the group properties.
. Besides these translations, rotations of the crystal that will bring the lattice
into complete covering with itself are possible and the group of translations
may be extended to a larger group, the so-called space group, which contains
both. The meaning of this synthesis in group theoretical terms will be discussed in 4. First the space group will be considered as a totality and then
analyzed in rotational and translational parts .
The space groups contain elements which represent pure translations,
elements which represent rotations around a center a~d elements that represent both. The elements representing rotations will be shown to form a
subgroup. These subgroups are called the crystalline point groups.
The third set of groups of interest in quantum mechanics are the point
2

'

198

be

POINT GR.OUPS

Ch. 7, 1, 2]

199

groups, i.e. th.e groups representing a finite set of rotations around a center.
These are ~~ed in molecular problems. The crystalline point groups are a
special case of the general point groups . For instance, the five rotations over
~1t around a certain axis form a point called C s . Thisisnotacrystallinepoibt
group since it is impossible to construct a periodic structure that contains
a five-fold axis, since one cannot cover a plane completely with pentagGn<s.
Most molecules happen to belong to the same point groups that form th.e
crystalline point groups. There are exceptions and nature seems not even to
"abhor" five-fold symmetry~ Th.ere is a substance called Ferrocine that
contains a 5';'carbon ring.. The a(,'tual shape of the molecule is somewhat like
a sandwich of two of these rings with an iron atom in the middle such that
the symmetry around an axis perpel1dicular to the rings is ten-fold.
The relations between the sets of groups are as follo\vs:

1. Full rotation group

2~

~
3. Point groulls

Space group

~
4. Pure t.ranslation

2' Crystal1ograph.ic point groups

where each arrow indicates a subgroup .

1 to 4 deal with the classical space group, 5 and on \vith quantum
mechanics.
,

l~

Crystallographic Point Groups versus GeaeraI Point Groups

In order to find out which point groups are crystallographic and which" are
not/it is necessary to translate the restrictions imposed on the crystaUOgraphic point groups into mathematical terms. There are t\VO different
ways to do this. The first is to construct geometrical bodies, of.a certain
symmetry and to see whether these "building blocks" can be stacked in such
that they fill infinite space. The~e objects are, of course, the unit cells which
fill the space of the total l~ttice.
A different and much simpler approach is to make use of the properties
of representations. This will be demonstrated for the simple case of a twodimensio~ lattice and will serve merely as an illustration for the threedimensional case. As mentioned before the defiaition ofa pomt gr~Up'is:
a set of transformations such that the lattice goes over into itself; Under the
restriction that one point is kept fixed. This group contains a finite number
of elenlents. Consider a certain point P (not idefttical with 0) under all

200

SPACE GROUPS

operations of the group (rotations, refiections). 'This point will remain the
same distance from 0: OP = OP'. If the group would be infinite all the
points P' wDMd cover a circle. From consideration of the ~pace group it will
be seen that any point 'in the lattice can be described by a set of integers.
Hence all operations can be described by a representation which uses only
integers. This is again true for the point group contained in the ~pace gro\1P.
Any representation of the crystallographic point group is equivalent to a
finite group of linear substitutions among integers. This important conclu. .
sion, which is fully described in the next section, leads in a very simple way
to the restrictions on the order of the group (compare BUltCKHARDT [1947]).
An arbitrary orthogonal rotation in two dimensions
X' = x cos cp - y sin tp
(7.1)
y' = x sin qJ - Y cos <p
can be conveniently expressed with the linear combi,nations x+iy and x-iy,
which gives rise to a transformation matrix
iqJ

e
( 0

0)

e- iql

As stated above, every representation is equivalent to a representation with

integers. Hence the trace of the above matrix, which is an invariant, should
always be an integer,

(7,2)
Since

(7.3)
the ~ible values of n are n == -2, -1, 0, 1, 2. Let e f., == e then multiplyin, (7.2) with 8 gives,
2
(7.4)
8 -ne+ 1 = 0,
8n"equation which can be solved for the five different values of n:

-2 gives
n = -1 gives

n:=

n
n

=
=

8
8

= -1
== i(-1i.J3)

0 gives 8 = i
1 gives e = t(1 i.J3)
2 gives e = 1.

(7.5)

Thus the"-(j,yclic groups can only be of order 1, 2, 3, 4, and 6.

It is possible to explore by similar considerations the number of possible
crystallographic point groups in three dimensions. For this analysis we refer
to the literature {compare BURCKHARDT [1947] 11 and 12}.

Ch. 7,

21

POINT GROUPS

201

'There are two conventions in use for the indication of (crystallographic)

point groups. The 1ir~t one was originally the Hennan-Mauguin notation,
now caUedthelnternational notation; the second is the Schoen1iiess notation.
The last one consists of a capital and one (or more) subscripts. The capital
letter is either C, D, S, T or 0, standing for Central, Dihedral, Screw,
Tetrahedral and Octohedral. The subscript n refers to the order of the~axis;
for instance C s refers to an abelian group of order five, the five rotations
over in. Dihedral is the symmetry of a "two-plane" that is a body from
which we consider back and front side equivalent; for instance D3 is the
group of 6 elements consisting of rotations around the 3-fold axis and the
three flip-over operations around the bisectrices which we considered in
Chapter 3, 7.4. It is the custom to consider the main axis placed in the
vertical positions and hence the two-fold axes are called horizontal axes.

Fig. 7.1. A tetrahedron can be imbedded in a cube showing that its rotation group has to
be a subsroup of the cubical group.
!

Addi1ional indiCfS are attached to C and D if we include the improper

rotations: i for the inversion, v for the retlection with respect to a vertical
pl~ne and h for the reflection with respect to the horizontal plane. S
indicates a screw axis, a combination of translation and rotational motion.
This possibility is only mentioned for completeness; it does not belong to the
point group operations. The groups T and 0 are the only two point groups
in which the axis of highest symmetry occurs more than once, that is, in all
<lther crystallographic point groups there is always .one dominant axis
which we took vertical. If we do not restrict ourselves to the crystallographic
point groups there is still one more group. The dodecaeder group, consisting
1
of 10 pentagons. This group has 60 elements.
1 Tho octahedron is obtained by connectinl all midpoints of the f~ of the cu~,
honce has the samo II"QUp. A similar reciprocal relation exists for ~e l~sabedr~n WIth
respect to the dodecahedron. This, tog~er ~ith the t~trah~ron (whIch IS self reciprocal)
completes all five possible regular bodlos In three dltlleUSlons.

202

SPACE GROUPS

[eh. 7, 2, 3

As a memory devise geometrical objects can be constructed which have

the proper symmetry. Some pictures are indicated in, this section . In parti-

cular Fig. 7.1 servesto,illustrate that T is, a subgroup ,oro. Another method
is the stereogr~phic projection. Table 8.2 gives all the' crystallographic
point groups. The extension to non-crystallographic point groups is easy to
establish and can be found in the chemical literature. 1 A third method
using pieces of cardboard has been described by SCHIFF [1954].

3. Space Groups
After the intr9ductory remarks in 1 a notation must be ~sta.blished for
the elements of th.e space group. These elements consist of a combination of a
rotation and a translation. By rotati,on we mean a P1l!e rotation, a reflection.
an inversion or a combination of these . It should also be pointed out that an
element may be a product of a rotation without translation and a translation
without rotation, but that this is not necessarily so.
In certain space groups there are screw elements, that is operations consisting of a translation and a rotation which is not a product of a translation
without rotation and a rotation without translation where both are elements
of the groupthemselves . The screw elem.ents do occur in nature, for instance,
we know of the existence of optical effects, such as the rotation of the
polarization plane. This occurs if the index of refraction for left circular
polarized Jtjght is different from the index. for right polarized light.
The mo~t general translation allowed in a three-dimensional crystalline
strtlcture is described ,vith help of three primitive translation vectors a, b, c.
These ale three non..coplanar or independent vectors~ that is
/

a(bxc)~O

(7.6)

chosen in such a way that they represent the smallest possible translations.
We will see later that there is a certain 'ambiguity in the choice of these
vectors altho1.1gh not quite as large as the choices of basic vectors in a
vect()r space.
The general translation is

(7.7)
where n l ' n2, n3 =s 0, 1, 2, . ., ..
The"pure translations are not the only possible operations allowed in a
t

E'i:~lNG! VVAVI'!.R

and

KIMBALL

[1944J, page 379.

Ch. 7, 3]

SPACE GROUPS

203

crystalline structure. There may be certain rotations and reflections possible

too. We will indicate these in general by a matrix (Rij). Hence the operations
of the space group can be written as

Xl

= R 11 X 1 +R12X2+R13X3+TtX1
= R 21 X l +R22X2+R23X3+T2X2

x~
x~ = R31Xt+R32X2+R33X3+T3X3-

(7.8)

Or in a more convenient notation:

x' = (TIR)X = Ax.

(7.8')

(This differs slightly from the convention used by SEITZ [1934] where the
rotation is written on the left, the translation on the right in order to preserve the usual order of operations: from right to left.)
A possible representation of A can be obtained as follows: The operation
A on any T should result in a general translation. In particular if we operate
on a primitive translation vector

Aa. =

n~a)a+n~a)b+n~ll)c

the integers n)ll) considered as a three-by-three matrix form a representation

of A.
The main characteristics of the elements A are as follows:
1) The Unit element is
E

2)

the

= (Olb ij)

or

(OlE)

(7.9)

product of two or more operations is given by

y

Fig. 7.2. Illustration or the product rule for space group elements.

SPACE GROUPS

204

[eh. 7, 3

(7.10)
The product rule can be illustrated with a simple diagram (FiS- 7.2).
A certain" Point taken initially at 0 and a certain direction in the crystal
taken parallel to the x-axis undergo a rotation R1 and translation T 1
This is followed by a second operation (T2IR2)' The translation axes are
fi.xed in the crystal, hence the second rotation will rotate the vector 11 ,
or to say it differently, the second 'operation takes place as if the origin had
been at 0'. Hence the total translation is not T2 + Tl but T2 + R2 Tl .
3) The inverse element is
(TIR)-l

= (-R- 1 TIR- 1 ).

(7.11)

This rule can be derived from (7.9) and (7.10).

Pure transl~tions are characterized by

(n1a +112 b +n3 ct<5)

:=

(n141~}(n2 bI6)(n3 cl<5).

(7.12)

Due to their Abelian character, a group element of this type can be unambiguously written as the product of three group elements, each represent.
ing a translation in a certain direction.
4) The associative law holds: This proof will complete the demonstration
that the combination of rotations and translations indeed form a group.
Let the operator (T3IR3) act on the product of two group elements,
(T2IR 2 )(T 1 IR1 ) =- (Ta +:a.,l T1 IRz R. 1 )

Then the result is,

(TIR)
No~

= (T3+R3(Tl+R2Tl)IR3(RlRt).

(7.13)

consider the product of 3 and 2 acting on 1,

(TIR) = T3 +R3 Tl)+R.3B.2(Tl)I(R3R.l)Rl)~

This gives the same result.

The space group will contain elements that are rotations only. These
elements form a subgroup: the multiplication of two of these elements
(OfR 2 )(O:R 1 ) = (OIR2 R 1 )

gives an element without translation. This subgroup is not invariant, since

(TIR)(OJ:a.')(TIR) - 1

= (TIR)(Ofa')( _ll-l Tla -1)

1

= (T-RR'R- Tln'R.-

(7.14)

using (7.11) and (7.13). This is clearly not an element of the point group

Ch. 7, 3]

SPACE GROUPS

205

(except in the trivial case that R' = E). This result is independent of the
presence or absence of screw axes.
The 1triplc product (7.14) can also be illustrated with Fig. 7. I. The second
operat~on will transform 0 into 0' and the third operation will translate P'
backwards to 0', instead of to 0 with the result that the triple product does
not represent a pure rotation.
On the other hand, if pure translations are considered it is immediately
seen that they form a subgroup (Abelian), but this subgroup is invariant,

(TfR)-l(T'IE)(Tllt)

= (R- 1T'IE)

(7.15)

using (7.10) and (7.11). The result follows immediately from (7.13) since the
total rotation is
R3R.2Rl

= a-lEa =

E,

\vithout considering the translational part.

The factor group ~/~, (t represents the space group and :T the invariant subgroup of pure translations) is isomorphic with the point group if
there are no "screw"-elements in the group. This can be easily seen by the
following argument. Consider all the elements of the space group and dis~
regard the translational part and set up a many-to-one correspondence
between these simplified elements and the elements of the point group.
This leaves a number of residual elements. These are elements that have a.
rotational part (a rotation proper, an inversion or a refiection) that does not
belong! to the point group. These are the screw elements; if they are absent
the iso~orphism between t/fT and the point group holds.
The 'following is an example of a structure containing screw elements.
That,ls an example of an element which has a translation, that is not part of
the pure translations and a rotation that is not a part of the point group.
(Both conditions have to be fulfilled simultaneously.)
Consider the so.. caI1ed diamond or zinc-blende structure, the crystalform
of germanium. The conyentional way of describing this lattice is to con~ider
it as two face centered lattices displaced over a distance i, i, t. T'he tetrahedral bond arrangement in one lattice can be obtained from the other by an
inversion. The "screw" element consists in this case of a translation over -1.J3
along the body diagonal (which is not an allowed translation by itself)
combined with the improper "rotation": inversion. If we carve a WignerSeitz unit cell around these two (non-equiva1ent) positions by constructing
planes halfway perpendicular to the bonds with the nearest neighbors we
see easily that this was the only screw element possible. In a zinc-blende

206

SPACE GROUPS

rCh. 7, 3, 4

crystal, which has the same structure except that one type of site is filled
",~ith a Zn and the other with an S, the screw element is excluded.
A ted!()"tIs analysis shows that there are 230 different space groups. They
are a combination of one of the 23 point groups and one of the 14 different
translation patterns. A translation pattern is illustrated by a space lattice or
a Bravais lattice. These Bravais lattices are the space groups in which there is
one atom per primitive unit cell (i.e. the unit cell spanned by the three primitive unit vectors). In order to bring out the system (a system is characterized
by the number of 2-, 3-, 4- or 6-fold axes) to which these Bravais lattices
belong, the so-called conventional unit cell is used. This cell has a volume
twice or three tinles the smallest unit cell (compare PHILLIPS [1956]). In
quantum mechanics it is convenient to introduce a more symnletrical unit
eell, the Wigner-Seitz cell mentioned in 6.
\

4. Structure of the Space Group

The synthesis of two groups of order 91 and 92 into a group of order
glg2 can be accomplished in different ways, depending on the nature of the
elements. of each group. This problem has its counterpart in physics since it
l1appens in a nunlber of cases that a physical meaning can be assigned to the
elements of the first group and a different physical meaning to the elements of
the second group. For instance the rotations in space and the permutations
of el~ctrons are both connected with a group. The system is invariant under
the qombined group of spatial rotations and permutations (compare note on
pagt) 120).
111 many, but by no means all, physical cases the elements of both groups
cqlnmute whlch makes them absolutely independent. This type of synthesis,
where all elements of C 1 commute with all elements of (2, is called a
direct product of two groups. The product group t( contains f 1 and rg 2
as invariant subgroups and the factor group with respect to one is isomorphic
to the other:
t

C/~1 '" ~.2;

f/~2~. ~1.

The converse statement holds also. If (i) c 1 and ~.2 are invariant subgroups
of <#, (ii) the elements of t!J consist of the products of the elements of t[lt
and the ,elements of (12 and (ill) the elements of < 1 and l 2 have only the
unit element in common, then the group C is the direct product of ~ 1 and
!l. That is a given element of t can be written as a product:

Ch. 1, 4]

STRUCTURE OF THE SPACE GROUP

where both G 1 and

47.)

02

are determined by

G. (VAN DER WAERDEN

207

[1937]

There.are less restricted ways to form products of groups: the weak direct
product and the senlidirect product. Both are generalizations of the direct
product mentioned above, but their nature is entirely different.
The weak direct product (MELVIN [1956]) has instead of commutability,
the requirement that the product of an element of c 1 with an element of f' 2
should give an element of < such that

/here G; is another element of t2 and G~ another element of 91 1 , requirements (i) and (iii) are maintained. The weak direct product is the most
general product, because it does not require any subgroup to be normal
(compare problem).
/\.. third type of direct product is the semidirect product (McINTOSH [1958a].)
Let fJl and ~ be two groups and suppose fR is an operator group for f.
That means that an element of (Ji operating on an element of rJ results in an
element that belongs again to ~. A group can be defined by stating that its
elements WIll consist of a combination of one element of fJl and one element
of ! such that the product is defined by the rule
(G'iR')(GIR) = (G' R'(G)IR' R)

(7.16)

wher~ R'(G)

means the operator R' works on G. In the application of this

chap~er G is the group of translations, an Abelian group, and hence the
first "product on the right-hand side is usually written as a sum.
~ general, that is whether the group Cd is Abelian or not we can state that
the subgroup t is invariant and that the subgroup f~ is not, except for trivial
cases .
The requirements for a group to be a semidirect product are (i) the exist...
ence of a non-trivial invariant subgroup. If all cosets of this invariant sub..
group are formed it should be possible (ii) to select an element from each
coset such that the chosen elements plus the unit element form a group.
E~ples of semidirect products are:
(i) The <;iouble group (see Chapter 8, 3). (ii) The space group without
screw elements. This group is isomorphic to the semidirect product of the
group (TIE) and the group (01:&).
An example of a combination which does not fit any of the categories is
the combination of Lorentz transformations without rotations and the

SPAC~

208

GROUPS

(Ch. 7, 4, S

rotations without Lorentz transformations. In this case the first set of

operations does not even form a subgroup_
5. The Qa'W.m Medlanics of SoH. State
In nearly all quantum mechanical problems we want to use wave functions
that are adapted as much as possible to the symmetry of the problem. The
space groups provide a way to study the transformation properties of the
wave functions both. with respect to rotations and translations. The structure
of the space group provides us not only with some ideas for the best eigenfunctions with which to start in a given perturbation calculation, but with a
well-defined set of quantum numbers as well. In such a study two questions
arise immediately. First, what kind of space groups are possible, a. question
considered briefly in 3; and second, what are the irreducible representations
of these groups in unitary space, a question to be considered in 6, etc.
Before we do this we would like to dwell a moment upon the most reasonable choice for the zeroth order wave function in a solid state problem. This
question is directly connected with the type of binding 1: Ionic, Covalent or
Me~c~ In the ease of ionic binding the wave function will be such that the
electron from one atom spends most of its time around another atom. In the
case of covalent binding the picture is more like an even distribution over the
two atomic centers with an increased charge density in between. This density
is the re~ult of the exchange integral and responsible for the binding effect
of such.i wave function. In the case of metallic binding the wave function is
spread out all over the crystal. This is called, in chemistry, a non-localized
bond, itl (;Ontrast to the first two which are pair interactions. The famous
example of a non-localized bond is the benzene molecule.
The most extensive use of the group theory of translation is made in
considerations on metallic binding, since here we are really dealing with
wave functiQns spread out all over the crystal. There are, however, situations
:n which th~- application of group, theory is of importance to covalent solids;
"'excitons" or non-localized excitations . As a matter of fact the distinction
between these two is vague, since on one hand the metallic binding is associated with bands, but on ~ other hand covalent pair bond will give rise to a
narrow baJ;lD-as well if we take all possible pairs in the crystal. illustrated
belo'Y is a'simple example where there is no fundamentafdistinction between
the different types of binding or to formulate it differently all degrees of
.,(,

1.

See for instance SLATEll [1953].

Ch. 1, 5]

THE QUANTUM MECHANICS OF SOLID STATE

209

covalency, ionicity or metallicity are possible. This is no surprise if it is

realized that each of these binding types are related to a certain type of zeroth
function .
order :wave
..
Consider two electrons 1 and 2 and two centers a and b;

((Ja(r) = q>(r-RIJ),

(7.17a)

If'(, - Rb)

. (7.17b)

lfJb( r)

:=

Fig. 1.3. Two center coordinate system used for diatomic molecules.

The two wave functions are the same functions if taken with respect to their
centers. The most general zeroth order (i.e. product type) wave function
~ch can be set up is
I

: t/J

= C 1 9'a(1)cp.,(2) + C 2 fPb(1)<Pb(2) + C 3 q>.(1)tp,,(2) + C4 tp,,(2)lPb(1)

(7.18)

~here C 1 , C 2 , C 3 and C 4 are arbitrary coefficients.

/ In the ionic case both electrons are at one center, say b. Hence C1 <<:: C1
and, since exchange is disregarded, C3 == C4 == o.
In the covalent case take C 1 == C1 == 0 with C3 == - C4 , the only possible
combination that is antisymmetric. The binding is the result of a spin-zero
state and since the total wave function has to be anti-symmetric, the orbital
part has to be anti.symmetric too. This type of wave function gives rise to
exchange integrals and as a result of that the charge density at 1(2" + R,,)
Yll11

increase.

In the metallic binding we deal with a product of molecular orbitals. If

these are taken as a linear combination of atomic orbitals (L.C.A.O.),
which is a simple and convenient choice, we have
(7.19)

reh. 7, 5, 6

SPACE GROUPS

210

and
or
C1

= C2

= C3 =

c4

To summarize this the following list for the two electron two center case is
given:

Ionic
eonvalent
Metallic

C1

Cs

Ca

C,

large

small

It is assumed that the N-electron N-center case shows the same characteristics. This undoubtedly is true for the ionic case, to a lesser extent for the
covalent case. The metallic case needs considerable improvement and it is
here that the fullest use is made of the space group.
6. Pure Translations

Retprning to our main problem, i.e. to what representation of a certain

space !group does a certain wave function belong, consider an element
s (-fIR) of the space group ~,

Sr

=,.' =

Rr+T,

(7.20)

which induces in the wave function space a transformation that can be

written as a linear combination of basis functions
(7.21)

If irreducible representations are sought, it is noticed that the invariant

subgroup of pure translations Tn = nl a + n2 b + n3 c is Abelian. Hence
the irreducible representations of this suberoup are one-dimensional. The
result of '~a translation is
(7.22)

where

l'

is a number, in general complex. If the three primitive translations

PURE TRANSLATIONS

Ch.7,6]

211

are introduced, the vectors are considered as operators,

(7.23a)
= aop V1(r) = tXt/J(r)
(7.23b)
To 1 0 t/I = bop t/I{ r) = PlfJ( r)
(7.23c)
Toot t/I = copt/l(r) = rl/l(r)
and then we can calculate the three cQefficients Cl, pand 1. If periodic boundT 100'"

ary condltions are involked, which' is not necessary but convenient, we find
x,

(aop) N1 t/1(r) = ~Nlt/1(r) = tfJ(r) -+ t?l = 1

(bop)N2y,(r) = pN:J.t/I(r) = 1fJ(r) -+ pN2 = 1
(cop)N3.p(r) == yN3tfJ(r) = ",(r) -., yN3 = 1.

(7.24a)
,..

(7.24b)

(7.24c)

etc. means that the operation. of a primitive translation is repeated Nt

times 011 the wave function. The distance N14 is the distance over which'we
consider the lattice periodically repeated in the a-direction. Similarly
N 2 band N 3 C in the b- and c-direction.
The solutions of these three equations are,
(tNt,

= eiltl/Nl

(hi = 0,1,2, ..., N t -l)

(7.2Sa)

p = ei"'21N
1 -+ ')' = elh3/N3

(h2 = 0, 1,2, .... , N 2 -1)

(7.2Sb)

= 0, 1, 2, . . . , N 3 -1)_

(7.2Sc)

a,Nl

=. 1 -.

pN2

= 1

yN3

-+

ex

(h 3

NoW we want to describe the result of an arbitrary translation: say III

steps ih the II direction, n2 steps in the b direction and n3 steps in the c
direction, i.e. a translation:

The result is easily obtained from. the preceding considerations: it produces

eigenvalue

an

1"

= exp i(h 1 n1 /1v 1 +h2n2IN2~h3n3IN3)"

to

This is n;>ta'very practical'way make this description since (7.26) acJtually

describes,. with respeCt to an obliq'ue set of cootdinate vectors 4, b, and c.
Although nl , n2 and n3 were originaUy thought of as int~gers, this restriction
can be omitted without any trouble and hence they can be considered as the
/

componttifs of"rin' this coordinate system whichjs adapted''to'thecrystat

Howevert r usually js expressed in rectangular' coordinates
(7.27)

SPACE GR.OUPS

212

[Ch. 7, 6

Comparing the oblique components (7.26) with the rectangular components

(J .27) we find
(7.28a)
nlQ.x+nlbx+n3C,x =: x
(1.28b)
nlQ7+ n2 by+n3 c, == Y

(7.2Sc)

nlQ.I+nlb.+nsc. == z.

The inversion of this formula, i.e. the oblique components expressed in

the orthogonal components, gives:
nl == y-l{(6xc)xx+(bxe)"y+(bxc)sz}

(7.29a)

"2 == V-l{(cxa)~x+(cx.),y+(cxa)%z}

(7.29b)

n3 == V-l{(O x b}~X+(4 X b),Y+(1I X b)zz}

(7.29c)

v == II <b x c).

(7.30)

where

The coefficients form. three vectors

* ==

II

bxc

,. (bxc)

b*

==

eXIi
II

(bxc)

axb
; c* == ----.11,.

(bxc)

(7.31abc)

which have the following properties as is easily verified by equationi (7.31 abc)

11* == 1 4* b == 0 a* c == 0
b* == () b*" == 1 b*. e == 0
c ." ::: 0 c* b = 0 c* c == 1.
Hen~

(7.32)

the expressions (7.28) can be written as:

(7.28')

and the eigenvalue of an arbitrary translation can now be expressed as,

"' == exp i(h 1 1l,*/N 1 +h"b*,N2 +h 3 c*/N 3 ) r.

(7.29)

The numbera hi, h2 and hs could be used as labels for the different irreducible representations. Usually it is preferable to use:

" == ht.*/N 1 +h 2 "*,N 2 +h3 C*'N3.

(7.30)

Obviously this vector " can attain only a limited number of values. This is
usually expressed in a geometrical way, i.e. by stating that Ie can only
1

The asterisk caDDot be confused with complex conjugate, since these vectors are real.

Ch. 7, 6]

P1JRE TRANSLATIONS

213

occupy the points of a certain unit cell in an abstract lattice. The lattice
is called the inverse lattice, and is generated by multiples of the vectors
a*, b* ana~c* in a space called the inverse space since the unit-length is the
inverse of the conventional units of length, like em -1 etc. The unit cell is
mapped out by the condition for hi' h2 and h3 mentioned in (7.25abc).
Technically speaking this unit cell is a half open point set, i.e. all points at
the boundary at one side are included, all points at the opposite boundary
plane are excluded. The reason that one should not take both ends is that
they differ by a distance a*, b*, or c* depending on the pair of planes one is
considering and hence are equivalent points.
Instead of the condition (7.25 a.bc) any unit cell ill the inverse lattice will do.
The most elegant choice is the unit cell vvhich was introduced by WignerSeitz in their ca1culations in the direct lattice. This cell is obtained by considering a lattice point and its neighbours and erecting planes half way,
between each atom and its neighbours, and perpendicular to the connecting
line. This unit cell is called the (first) Brillouin zone and would have been
established directly if \ve had used the following range of h values:

= 0, 1~ ... , !(N-l);
-- 0 ,...!..
-l- 1,

.J,-""
..1.- "', .. ..

t(Ar+, l)-N, ... (N-l)-N

-l_'-"!1 (N.- 1)

(7.25')

obtained by subtracting N from the second set of values. 1

The shape of the Brillouin zone is entirely determined by the symmetry of
the o~iginal lattice and Figs. 7.4 give an illustration of the different lattices.
The general Brillouin zone is the locus determined by the points k, such
that,
(7.31)
where K is a vector connecting any l)air of points of the reciprocal lattice.
7..

Bloei.~

'lneorem

The basic point in sytnmetry cOllsidc"rations of quantum mechanical

problems is that although certain operators and their expectation values must
be invariants under the operators of the group, the wave functions are not
necessarily invariant. Only tIle absolute value of the wave function (or in
case of degeneracy the absolute value of a linear combination) has to be
invariant. In the case of a non-degenerate "rave function this leaves a certain
freedom on the phase of the c0I11plex <~a ve function .

214

SPACE

OROU~PS

[eb. " 6

z.
I

7.4(3)

7.4(2)

7.4(5)

7.4(4)

,
I

~A

P-r ---- -Ll

I

X
-",CY'
,;

-0- - .- -

7~4(6a)

'(

-.v

7.4(6b)

Ch. 7, 6]

PURE TRANSLATIONS

215

7.4(7)

7.4(8)

7.4(9b)
z

,
f

A~

I
I

i
l

,
I

rf4

A--o-

/ / ?'-M

710)

r~_..g_

/~;
;x

'c-'

216

[eb. 7, 6

SPACE GROUPS
z

7.4(11 b)

7.4(12)

z
z

7.4(13)

7.4(14)

i~il. 7.4. The 'Wigner-Seitz' or symmetrical unitceUofthe fourteen spacelauices; (l)Tridi.pic

(not illustrated); (2) MonOclinic, simple; (3) Monoclinic, base centered; (4) Orthoromhie, simple; (5) Orthorombic, base centered; (6) Orthorombic, body centered. (a) Hciaht
Jarger than the diagonal of tho rectangular base. (b) Height smaller than the diagonal of
the rectangular base; (7) Orthorombic, race centered; (8) Hex8aonaI; (9) Rombohedral,
.also called trigona'. The three primitive translation vectors are on a cone around the
z-axis, making equal angles with one another. <a> Height of the con. larger thap "\1'2 times
-the radius of tho circle. (b) Hoiaht of tho cone smaUerthan '\1'2 times tile radius of the circle.
If equal, the lattice is face centered with z-axis along body dfal9n&l; (10) Tetrqonal,
.simple; (11) Tetragonal, body centered. (a> Height !arpr than the diagonal of tho square,
(b) Heiaht'smaller than the diaaonal of the square. If equal, tho lattice is race amtered
cubic;<tho diagonal is one of the edges of the cube; (12) Cubic. simple; (13) Cubic, body
-centered; (14) Cubic, face centered.
If we have one atom per unit cell tho corresponding Brillouin zones are the same except
.for the interchance of 13 and 14.

Ch. 7, 7, 8]

REDUCED WAVE VECTORS

217

After these introductory remarks it is obvious that a non-<iegenerate wave

function can be written as

(7.32)
where u(r) has the same translation symmetry as the lattice that is if
V(r) = V(r+ T)

-+

u(r) = u(r+ T).

(7.'33)

Since the wave functions in quantum mechanics have to be basis functions

for the representation(s) of the symmetry group of the quantum mechanical
problem under consideration, the results of the last section can be applied
immediately and hence the phase factor cp(r) is equal to k r where k is the
label of the representation. The result is that, in full generality, the wave
function can be written as

(7.34)
where u,,(r) is periodic with the periodicity of the lattice and. is in general
dependent on k. For a given value of k there may be more than oneperiodie
function u" hence one has to add another label to distinguish. these different
"branches".

8. Reduced Waye Vec:tors

In t;his section we will show how the symmetry adapted wave functions
look for a solid state problem of given symmetry.
If~e consider the Bloch Theorem [eq. (7.34)]~ which indicates the general
shape of the wave functions ilr the periodic lattice we could ask ourselves
what the wave functions are for an empty box, that is to say infinite space
wjthout any potential energy wells. This question is not quite so trivial as it
sounds, since these eigenfunctions will turn out to be the symmetry adapted
wave functions for the problem. in which the potential energy is not equal
to zero. The picture which we are going to describe is e&Sentiallya 8foaP
theoretical idea. Empty space. with infinite boundaries or with periodic
boundaries corresponds to a group which consists of an infinite number of
translations,e . lntroduce into this empty space a set of potentialWe~.whieh
are centered around lattice sites, then the corresponding group will consist
of a finite number of translations. The Bloch wave functions form the basis
for a representation of this group. Hence we are dealing with the problem
which we have already mentioned many times: the perturbation requires

218

SPACB OROUPS

[eb. 7, 8

that the oripal group of transformations be replaced by a SUbsrOllp of

transformations. It is helpful to introduce a set of'wave functions that aJready
form a ~ucible basis for this $ubjroup.
Hence, the replacement of the free electron wave functions by the following
set of Bloch fUDetions

(7.35)
is the expression of the above-mentioned procedure. We also see clearly the
'JDe&nina of the reduced wave vector emerge, that is the wave vector inside
the Brillouin zone.
The periodic function

u(x+T} == u(x)

(7.36)

is independent of It, but ~ the label K which numbers the different

branches. K is the distance between any two points in the reciprocal lattice.
Now if we plot the eD.er1Y as a fuactio1\ of Ie we seo:that for the wa.efunctiOll
described above the single parabola is,teplac:edby a large set -of parabolas
(Fi,. 7.5). 'Ibis description forms the start of our perturbation calculation.
E

.,It

Pt.. 7. 5.' J!aav u a ftmctioD of the \lDJ'eCluced wave vector (or exteIlded Brillouin zone)
.for a tree pardcJ.e.

TIle introcluccion of Don-zerO periodic' potentials will slightly deform the

parabo~ri1l ~ethepcrt1lrbatiOJ1ia very strona
maybe cottsiderably
dofotmed. In both cases, however, the group theoretical argument which
we uaed stays valid. Suppose for -convenience that the perturbation-potential
is relatively small. In this case the main shape of the parabola is maintained

*"

Ch. 7, 8]

219

REDUCED WAVB VECfORS

~ in the:pointa,neat the Brilouin zone.

At these 'points the eDerg curves

intersect and.,hence we have to refer to the problem of a degeaerate or a nearly
depnerate'teteL Aa we have seen in Chapter 2, 7.4 this mayor may no~
live rile to the so-called Don-crossing rule, depending on whc~ the, offdiagonal elements are zero or non-zero. If we suppose, for the moment,that
the non-diagonal elements in the secular matrices are non-zerot we will .ve
Figure 7.5 changed to Pigure 7.6. It is interesting to notice that as a result of
\

...
,. If

Pia. 1.6. BaeraY curves in tho UDreduced wavo vector space, for particles in a periocIic
potential.
t

Iymme~ considerations both curves

wm have a taqeD.t which is horizontal

at the t"Brillouin-zone" (that il the point k - 1(,/0 in the ono-dimensional

case). ;this leads to two important conclusioDL Firat there will be~a gap near
the ~uin zone aDd second the energy aa~afuDctioD of'k can beapproxi.
mately' descnW by a parabola in this neiJhborhood. This, means in the
1aftauage.'ofaolKt state phyaica that an effective mass can be assigned to the
electron. or holes in the nei8hborhood. of the gap. 111 many cales,of,. . . .
this simple picture may be,~ by one or more of,:tb,e fo1low.iDa
three facto~. Firat, the problerrl is actually not _ o n a t aaclhence
the enerlY lines are three-dimellsional'surfaces in f o u r _ apace~
Second, the wave function with which we are dea1ift8 is not always sinale,
but may be, for instance, three-valued like a p-functionin a germanium. atom.
Third, the-conclusion about the parabolic behavior is destroyed if there is
spin orbit coupling in the neipboJhood of the pp.
The usual Sroup theory in solid state physicsean be extended beyond this
point. Besides the group of lattice translations wb.k:h are expressed by the

220

SPACE GROUPS

Ch. 7, 8 9)

fourteen Bravais lattices there are also point groups in the reciprocal space.
These point groups mean that certain sets of k vectors may be the basis for
a group. elf. the same or of lower order. This idea was worked out by
B:ROUCKAER.T, WIGNER and SMOLUCHOWSKI [1936]. A short review of their
ideas will be given in the next section.

9. Little Groups, W.B.S. Method

9.1. LITTLE GROUP THEORY

An irreducible representation of a group ~ can be used as a representation

of one of its subgroups ~ by omitting all matrices that do not belong to fl.
The representation of .Ye obtained in this manner is in genetal reducible.
This is expressed as follows:

(7.37)

r,

where
represents the jth irreducible representation of~, S, the representation of .Tf' after omission of the elements of ~ that do not belong to ;F
and '14 the irre~ucible representations of :Jft'. The coefficients ca are either
zero or integers. This pattern is the major theme in many applications of
group theory in physics and the general case will be considered again in the
next chapter.
In $is chapter the special case where ,;'I' is a normal subgroup will be
consi*red. This case is, as we shall see below, of principal interest in the
appliq,tion of group theory to solid state problems. From the purely ma~
matiaU. point of view, the demand that .Tf' be normal is interesting for its
own/ sake since it leads to a number of important considerations.
Let usrecaIl that if the (abstract) elements B belong to I, than ABA- 1
wjI1 also' belong to :Jft' (definition normal subgroup). Consider an element
A that does not belong to :Jft'. otherwise the statements would have been
trivial. \A different, way of characterizing the normal subll'oup is to say that
any "outside" elementwillindu.ce a permutation of the or4er of the e1~ments
in the subgroup .Tf'. The same statement will hold for the representatiou:
l

3(B) ~ Z(ABA -1)

(7.38)

which gtVes rise .to anot'her possible representation. We can maiDtainth~

notation since we are dealing with the same matrices as before; the difference
is that they now correspond to different abstract .elements.: The important
point is that the representation obtained this way (called the conjugate

Ch. 7, 19]

LfITLE GROUPS, W.B.S. METHOD

221

representation) is not necessarily equiv~ent to the original one, since

~,

S(ABA -1)

matrix

= r{A)r(B)r(A -1) ==

r(A)E(B)r(A -1),

(7.39)

and the
r(A) does not have to belong to the representation S.
If S is decomposed into irreducible representations with respect to ~,
a set of conjugated representations can again be constructed with help of the
different elements A available in ~ and "outside" ;t'. These conjugate
representations will again be irreducible (they are a permutation of an irreducible representation, hence if they were reducible, the original would
also have been), but not necessarily the same; i.e. non-equivalent;

1",(8) -.. ')',,(ABA -1)

= "",(B).

(7.40)

We would like to mention that 1", and 1" must have the same dimensionality and al$O, since all irreducible representations originating from r j are
each other's conjugate, we-:find that all representations ,),,, coming from one
necessarily have to have the same dimension. All this stems from the fact
that "I,,' was connected with 1" by a permutation of the ulinks" between
abstract elements on one hand and the representation matrices on the other
hand.
If the element A is considered as a variable which runs through <D a set of
irreducible representations conjugate to y" is generated, this set is sometimes
called an orbit. Conjugation with certain elements A of t will produce
representations
which are equivalent to the original one. This will definitely
1
be true if A belongs to ,;t' since in that case the representation remains exactly
the ,; same. Also if this condition, i.e.

r,

1,,'(B)

= ,),,,(ABA - 1) = 1,,(B)

(7.41 )

is fulfille4. for a certain element A, it is also fulfilled for all elements that are
in the same coset as A.
All the elements A that fulfil condition (7.41) form a group, the little
group ()f the second kind, .fRD. It is easy to show that these elements do
indeed form a group. For example the product rule: If A1 and A2 are any two
elemetlts, then
(7.41a)
y,,(A 1 BA;-1) == 1,,(B)

r,,(A2BA;1) == y,,(B)
and the product element Ai A2 belongs to !!en because

(7.41b)

/'

",,(A1A2BAiiAii)

== 1.(A1 B'A11) == 'Y,.{B").

Similar arguments can be given for the other group postulates.

(7.41c)

222

SPACE GROUPS

All elements of I belong to ,[JAI. All other elements of 9" can be

srouped in cosets, as mentioned before. The elemeDta Of'!l'D contain I as a
normal s1!b&roup and the factor group !ED/I is called the little group 'of
the first kind

:r.

If these ideas are applied to the normal subJroup ~ of pure translations

in the space group:

(7.42)

r.

the right-hand side represents another element of

If we now take the
irreducible representation of the subgroup
the element (TIE) will, according to 6 be represented by the following exponentials

.r

(TIE) -+ elt r e 7(k)

and the cotijusate representation of the element

(7.43)
T by:

(It -ITIE) -+ etlr r

(7.44)

accordin8, to the rule (7.41). If we replace R. -1 T by T we find

(TIE) -+ elk. Rr == eiRt r.

(7.45)

A1l conju,ate representations of

y(k) are the representations r(Rt). If

we take for instance a two-dimensional square latti.cc and if we ~. &,
repr~tation characterizedbya ~rtain k =ra (k;x:, k7) the conjup.te ~p~n
tation
be characterized by the following eight vectQrs 1 =: (k~, k7")
and <*k" k~) obtained by applying the aUo~~ operations of the lKlint
group iIJI consisting of rotations over in, 1, in and reflections with respect to
the Uis of the lattice. The prong thus formed is called a atar and is a special
casotof the "orbit" mentioned before. The k-vectors in the star represent the
se~ of all conjugate representations of '}'(Ic) that are non-equiVilent..
/Promthepreceding considerations it is cl~ that we find tlteset of vectors
that mutually form a star by taking a certain Ir-vector inside the Brillouin
zone and have it Underso the pOint transformations allowed by"the lattice
symmetry. The i-vector is inside the Brillouin zone since the "vector ~fers
to the irreducible representations of the pure translations and hence is the
reduced vector ( 6).
The el~llle11ts of the little group were those matrix elements oftl,le group t;I
that give rise to conjugate representations that are equivalent to 7(k).
These are given by i-vectors that fulfill the condition

rwt

Rk == 1c+2nK

(7.46)

Ch. 7, 9]

LI1'TLB GR.OUPS W.B.S. METHOD

223

where K is a vector of the reciprocal lattice

K. T; nl,,*+n2 6*+n3 c*;

n1' n:z and n3

:Ie

0" l, 2 ...

~ ~D~ the two-dimensional square .~ as ;~~<~haVe

mdicated m Fl,. 7.7 the points (r. M, X) and lines (Et A, Z) of sYriJmec:rYin
Icy
.------+----~M

Pia. 7.7. Brillouin zone of a 2-dimensional square lattice.

the BriUouin-~De. _' The little &rOUp belonsina to 4CO~~ of the u1li~
e1emontud diireftection with respe<:t to the klC~.l The
longing to r is tho same as the ,point group of tbelattJee':: D 4 Tlie'previOUI
two points wereex&mples of ,group elements that stay invariant,. urid8r tilt
condition (7.46) with K == O. The points M, Z, X su\.Y'not 'ofllyinVariant
unde~ certain rotations or refiections but also under a combination of these
with , translation. The irreducible representation of the special points or
lines ~ve been worked out for a number of Bravais lattices (see HEItMAN
(1958] and K()STBIl [1957]).

1iUJe,..,.pbe-

9.2.. APPUCATIONS OF THE THBOIlY OF THE UTILE GROUP

The first type of question that can be answered with group theory in band
theory are of the same nature as the questions answered ":~ crystal fiCld
theory. Suppose there are r wave functions all be10qing to the same energy
level in the unperturbed state, what splitting of these , levelS win take place
after the interaction is taken into account? The level &~, i:.e., the number of separate levelS -.u.d their depneracy, on tJlo.basis of,:these symmetry
considerations is called the uessential'" degeneracy. Whether separation
1

We mean

or course tho little poop or the Int kiDlL TIleUUle aroup of,.dle;'MCO~d

kind COD8iats of r tmd A', where It.. t, the reflectioB with l;apoct: to the x",ua.
that it, all pure traaslations plus all tranalatiODl combined with refteatiOlll arotmd the
x-axis.

[eh. 7, 9

SPACB GR.OUPS

really takes place remains to be seen from a detailed calculation, and if the
separation turns out to be zero, the additional degeperacyis called "acci.
dental" deaoneracy.
The names are rather unfortunate since there could be
.........
other reaSons, beyond the symmetry considerations, which made the
degeneraC1 not accidental at all. A famous example is the Fock-argument
for the energy levels of hydrogen atoms. 1 The essential or symmetl)'induced degeneracies occurring in the special points mentioned above
~ be illustrated by two examples.

The energy of a plane wave depends only on the magnitude of the vector k ..
Hence all the four points: (0, k.~) == X, (0, k,) correspond to plane waves
that have the same energy. This holds in particular for the two waves
corresponding to X. The little group belonging to X, however, has only
one-dimensional representations, hence the levels will split under a general
perturbation (the potential field of the cores). For symmetry reasons they
will have the same value at the points (0, ky ).
A second example is given by considering the four wave functions connected with M. They form a reducible representation of the little group
belonging to M. The last group is D4 and has four one-dimensional and one
~~~-:4~al~! representations. The representati9n resulting from the
four plane waves decomposes into two one-dimensional representations and
one two-dimensional representation (compare Table 7.1), hence under
TABLE

7.1

Character table: D. = M 1 +M,+M,

E

C.

C.

C.

C,

D4

.,wI
M,
M.

1
1
2

1
1

1
-1
0

1
-1

-2

C. : ~t. n around z-axis (1 e1.)

C. : rot. in around z-axis (2 el.)
C, : reflection with respect to x and y axes (2 el.)
C. : reftection with respect to the two diagonals (2 el.)

The 'wave equatioa of tile DOn-relativistic hydrogen atom can be written in a form
icloaticaJto the iatepal. equation for spherical harmonics in four-dimeDlioaal space. II).
Y8riaace UDder tile operatiODS of the last group leads to the ~Dclusion that the energy
1

levels with different L, but the same n, are desenerate. (Compare McINTOSH [1958b ].)

Ch. 7, 9]

LITTLE GROUPS W.B.S.. METHOD

22S

influence of the perturbation the four-fold level will split in two sinale levels
and one two-fold level. This is often expressed by statina that two of the
four surftLCes "stick together" at the point M. This is based on the fact that
r is a continuous variable of the function k and a slight increase in Ie, say
Ie -4 k + Ak, will correspond to an arbitrary point in the Brillouin-Zone and
hence the four energy surfaces will, in general, be separated if one moves
away from M.
Compatibility relations and "accidental" degeneracy. The special points
in the Brillouin-Zone fall into two classes. Points like ~, Z, and A form a
locus given by the lines rM, MX, and rx. The little group of 4 should be a
subgroup of the little group of r as well as of X. Suppose we know the irreducible representation of the degenerate wave functions in X. If we move
away from X but stay along A the wave functions will span a reducible set
and the representation of the wave functions in X will decompose into a
number of irreducible representations of A. Conversely, this implies that a
given representation of A can go over into a limited number of representations of X or r. These so-called compatibility relations between the irreducible representations have been worked out by BOUCHAERT, SMOLUCHOWSICI
and WroNER [19361 and for the two-dimensional (pedagogical) example we
give Table 7.2 taken from HEINE [1960].
TABLE

7.2.

Compatibility relations
Representation

Compatible with

r1rar.; XIX,
r.r,r:;; x.x,
r 1 r.r,; M1M.M.
r2rar,,;
Xl Xa;
XIX.;

M,Ma M ,
Ml Ma Ms
MsMfoM.

r.

reduces into ..41 +..4 1 or Xl +..E.

M" reduces into .E1+..E'1 or Zl+Zse

Group theory is not only helpful in degeneracies attributed to symmetry

as has been pointed out by HE1lRING [1937b].'This type of predictable degeneracy is most easily illustrated by an example.

226

(Ch. 7, 9

SPACB GR.OUPS
T,uLB 7.3

Character tablea

r.M

rM
riM.
l

raM.

r.,M,

riM,

.~

2. 4., 4:

1
I
I
1

1
1

mll
m.
1

1
1
-1
-1

1
-1

-2

-1

ot r, M, X,

m.

m.,

1
-1
-1
1
0

J1, Z

.r1.EZI

~l.&lZl
~.z.ZI

1
1

2.

m.

Xl
X.
X.
X,

I
1

1
1
-1
-1

1
I

~,

-1

1
-1

m m m.
1
-1

mil

1
-1
-1
1

Note: m. denotes a reflection in a line perpendicular to the x-axis.

CalcutatiODs have been made by

with the Wigner-Seitz

method on sodium. This metal has a body centered cubic -structure and the
corresponding unit cell has the shape of a cube from which all eight comers
have ~n cut off perpendicular to the body diagonals. The faces are hexagonal anb the point in the center is called G. The Wiper-Seitz method consists
of sol~g the SchrOdin~r equation inside the cell under the restriction that
it shQhld fulfill the Bloch condition at the surface. Slater was the first to
incll1de
wave functions with angular momentum (p, d, etc. wave functions)
/
besides the s functions used by Wigner and Seitz.
The wave function will be, in general, a mixture of the s, p, d, etc. wave
functions. There is no longer any reason to retain these labels because the
problem no longer has spherical symmetry. It turns out, however, in particular in the approximation used, that the wave functions are s-like, p-like,
d-like, etc. Calculations show that at the origin, the lowest level is s-liket
the next d-like, and at the point G the lowest is p-like, and the next s-like.
The lowest level at G lies higher than the lowest level at the origin.
This is 8.Iready sufficient information to draw the conclusion about nonaccidental degeneracy. The little groups belonging to 0 and G 8.re the same
as the Point 'group the lattice. Hence the s-like level has the unit representation, the p-like a three-dimensional representation, i.e., it stays degenerate,.
!

of

SLATER [1934]

Ch. 7, 9]

LITTLE GROUPS W.B.S. METHODS

227

and the d-like splits into a three- and a two-fold degenerate level. (Compare,
for example, Chapter 8, 2.2.) Using now the compatibility relations we can
state t~t ~from the s-like level originates a JJ 1 representation (along the 111
axis), from the p-like level ad; and a ~ 5 representation; (This last is two-fold
degenerate.), from the d-like level a three-fold representation giving rise to a
~5 and .4~ and a two-fold level giving ~1 and .d 2 If curves are drawn between the end points, making sure that every curve has always the' same
representation, we see that ~2 and.4 1 must cross each other. Considerations
tt

If2-4, +.da

Ii; -.d~+.d,

-+:k----...k

o--~-----

.Pig. 7.8. BnerlY versus k curve of sodium in the Ill-direction accordinl to HBiwNO (1937].

of this type apply only to cases in which a certain amount of information is

already available~ They are an extension of the idea.
compatibility. This
idea isparticularly helpful if calculations are,restricted to certain' poiatSin
k-spaco-," since they often ~allow us to connect the points in an ambipoUi
way_ From. the connections the conclusions about overlapping in a certain
direction can be drawn, one of the first questions' askectif ()ne studies a

or

~ar

solid.

The implication of time reversal on the energy band structure has been
discussed by HmuuNG [1937a].

CHAPTER 8
FINITE GROUPS
1. Rotational Crystal Symmetry

In this section we are interested in the so-called point groups. This is a

collection of symmetry elements applied about a point that leaves the structure under investigation invariant. The most common elements of symmetry
are the n-fold rotation axes about a point, that is the n rotations around a
lattice point over angles 21r,/n. Other possibilities are for instance mirror
reflections with respect to a plane or inversions with respect to a point.
A simple example was given in Chapter 3, 7.4. The example was a triangle
having a three-fold axis as well as three mirror lines, altogether six operations.
M!>~ophisticated examples will be referred to below. As has already been
pointed out in the preceding chapter the translation requirements of the space
group mean that only a restricted set of point groups can be dealt with in
solid state. This is in contrast to the situation in molecules where the only
resttiction is of internal consistency between the different symmetry elements
oft* group. It is only in the last section of.this chapter that molecules will
be cpnsidered again.
~ restrictions on point groups resulting from the translational requireI11;Itnts ~beiUustrated by using the two-dimensional case as a pedagogical
example. We ha:ve already seen in Chapter 7, 2 that in solids only one-,
-two-, three- anel six-fold axis are permitted, a proof which ,t;aD. be given
in aless formal way by considering the translational vector; b&ving it undergo
a certain rotati~Jl and demanOing that tile trallslation and the rotated ~a1ll
lations,c.Wfer by. qertain translational factor. This again leads to t~.result
that the possible .nales of rotation are limited to the sources mentioned
before (compare KrrrBL [19S7]).
In ~'Athese five ~sibilities are combined with a mirror line, five moro
possibilities arise. This is shown in Table 8.1, which indicates the . equivalent
points in a circle toge~her with the notation used in crystallography for the
two-dimensional point group.
228

Ch.

8.IIJ

ROTATIONAL CRYSTAL SYMMBTRY

229

TABU 8.1
Two-dimensional point lfOapI

08
O ea'
@#.'
O
. @,,
O
''
,
0

C,

1m

Ca

2mm

elY

C 2v

C3

e 3y

3m

"

C...

4mm

C.. v

6mm

The study of three-dimensional point groups brinp besides the rotation

axis mentioned ~fore, the possibility of a reflection plane, inversion
center. and the so-called rotation-inversion axis. The last arecombiDations
of a rotation., and an inversion, such that the rotation by itselfis not an allowed .tranSformation, ~ut the rotation is allowed if it is combined with an inversion. All possible combinations of these symmetry elements lead to 32
different tgree.dim.ensiona! point groups. In Table 8.2 Ute. stereosrama as weB
as ~ eqUlvat~t to the djagrams are indicated. An open cirde refers
to a point below th,plane of the r'paper, a. dot represents a point above the
plane of the papot.. The plane of the paper is usually considered to be horizontal so that the z-axis is normal to the paper. In case the plane of the .

230

[eb. 8, 1, 2

FINITE GROUPS

8.2
Three-dimensional point groups
TABLE

Triclinlc

o
1

Tetragonal

C1

J
C,.
Mon"ocllntc (2nd setting)

Q)

EBEB
EBffi
EBEB
\/f0
83 1....

Orthorhombic

f
\

C.

4'

s ..

4/m

C4h

I(T)\

. "'.....

422

CDCD CDEB ffi@

4mm

mm2

ffi@-,

-.
B

.. 42m
..

21m

mmm

02hJ

'D2d

'

""i'A
. !~\
.. \./JV
'-~

Ch. 8, t 1

231

( Contil7up.d)

232

FINITE GROUPS

feb. 8, I, 2

8.2
The thirty..two crystal point groups
TABU

---

International Symbol
Short
]

Schoen1iies Symbol

FujI
1
1
2

C.(C1A)

21m

nJ

222

mm2

mm2

i
4/m
422
4mm

42m
4/mmm
3

222

mmm
4

DIA(VM )

c,

i
4m

CIA

422

D.

4mm

S,

c,.

i2m

D ..(V.)

422

Du.

mmm
3

32

32

3m

3m

8m

!J~

6/m

C,(S.)

c.

222

mmm

C1

c.
c.,(S.)
D.
c..
Du
C,

c.
Cell

622

622

D.

6mm

6mm

8m2

8m2

C.1.'
D81t,

622
mmm
23

Du

6/mmm

23
m3

432

!S
m

43m

432
43m

m3m

~~

mm

paper is asymmetry plane, the circle is drawn with a heavy line. The stereograms are very useful except in the cubic case wh~re it is much. easier to
picture a cube instead of this rather artificial device.

Ch. 8, I 2]

CR.YSTAL FIELD THEORY

233

In case of a purely ionic crystal, that is to say the case where the electrons arv ~ completely localized and belong to a certain ion or complex, the
wave iunction of these electrons or complexes will differ from the free ion
wave functions in such a way that the environment, in particular the nearest
neighbors, will produce a perturbation on the wave function. This result
may be hard to calculate particularly if the perturbation is strong, but it is
always necessary that the final wave function should have the proposed
symmetry. If the symmetry of the crystal is really known the number
of unknown parameters in the perturbation calculation may be greatly
reduced. Hence if the symmetry symbol of a certain point in the crystal
is known from the literature, it may be possible with the help of the theory
or representations to set up the symmetry adapted eigenfunctions. This
will be demonstrated in the next section.

2. Crystal Field Theory

2.1. ANGULAR WAVE FUNCI'IONS UNDER. FINITE R.OTATIONAL SYM
METRY

Cons.ider a free atom or an ion, which has the symmetry of the full rotation
group, and place it inside the lattice of a solid. The allowed symmetry is now
restricted to the point group of that particular place in the solid. The physical
influence of the environment will be due to 1) electrical fields, 2) magnetic
fiet,ds, and 3) exchange interaction or covalent binding_
In general it is very difficult to give a complete calculation of these inftu~nces.
In certain cases, the so-called diluted paramagnetic salts, it can be
l
asSumed that the first influence, the electric field, is the most dominant of
the three. In this case a reasonable description can be made by assuming
electrostatic fields and ca1culating the influence of this perturbation on the
wave functions of the particular atom or ion. Fortunately, however, it is
possible to make a number of statements which are based on symmetry
considerations only and are completely independent of the actual physical
mechanism of interaction.
These qualitative considerations can be given with the help of grO\lp theory
or more correctly with the theory of representations. They were given the
first tinleby Bethe and simultaneously by Kramers.
Assume that a certain level in the free atom or ion was characterized by a
certain value of L or J and assume, in order to take a specific example in
mind, that the symmetry of the crystal is cubical, The Hamiltonian operator
H ul) will no longer be invariant under all possible rotations. The symmetry
j

234

FINITE GROUPS

reb. 8, 2

of the crystal ha~ reduced the number of rotations (which was 00 2) to only
those 24 rotational transformations which bring the cube over into itself.
Again lipply the theorem of Wigner, tb.at is, if cjJ -belongs to E then also
Sy, belongs to E, except that S now refers to a restricted set of rotations. ~ro
formulate it more precisely, the group of transformations S forms a sub..
group of the original group, the full rotation group. Again we state that
dimension of the reducible representation is equal to the degree of degeneracy
of the level, since the number of dimensions of the boxes of the reduced
representation is nothing but the number of linearly independent wave
functions which belong to that particular eigenvalue. However, the representation cJlaracterized by L or J is now in general reducible with respect to the
cubical group. lhat is to say, if there are no common unitary transformations
which are able to bring the infinite set of matrices of the full rotation group
simultaneously into a form with s~aller boxes, there may be transfonnations
which are able to do tIns for the 24 rotations which correspond to the spacial
rotations which bring the cube over into itself. At first we are only interested
in the nurnber and k:.ind of irreducibJe representations of the finite group
which are contained in the origina12L+ I-dimensional irreducible representation- which is reducible with respect to the cubical group.
This program can be solved with the help of the characters of the matrices~
One of the important conclusions from th.is ll1ethod is that only the symmetry
detet;mines the pattern of the splitting of the levels . TIns, however, will hold

only!as long as the splitting of the level is small compared to the distance to
the rlext higher (or next lower) multiplet. In such a case the wave functions
of stlch a neighboring multiplet will have to be considered before starting
t~ considerations formulated above .
Before considering the general theory we would like to describe an example. In how many levels will an F-state (L = 3) split jf placed in a cubical
field and ho"" large is the residual degeneracy? To answer such a question
the characters ofthe 5 classes of the cubical group in the (2[. . ,+ ))..dimensional
representation of the full rotation group have to be calculated . These are
easily- calculated with the help of the following formula:

Note that the formula is derived for a rotation. around the z-axis (compare
(5.28) but the result holds for a rotation w around an. arbitrary axis, since

bet trace is invariant under a similarity transformation (1.7) or (1.27a).

We have to choose for tb.e unitary matrix S (or U) a transformation that lets
the rotation axis coincide with the z-axis"
The five~lasses of the group of cubic symmetry corr'~'spond to the following
operations.
E Unit operation.
C 2 Rotation over 1t around each of the three edges (3 elements).
C 3 Rotation over -!1t around each of the three edges (6 elements).
C 4 Rotation over 1'(, around the face diagonals (6 elements) .
C s Rotations over 11t around the body diagonals (8 elements).
Since each class always refers to one and the same angle of rotation,
that particular m in (8 . 1) has to be substituted. The result is~ for general
values of L, as follows 1 :

X E = 2L + 1 ;
"'c3=(-1)[t
Tn this

s~cific

XE

X C2
L

];

Xc,=

X C4
{

= (-- l)L
1 (L == 3n)
0 (L=3n+l)
,;;.,}.
1 (L '~ ~n+""\

(8.2)

'--.,,;

case where L = 3 the equations becoIne,

== 7; X 2

= X ...

-1;

X3 = -1; '-s

= 1.

(8_3)

Now if trus set of characters is compared with the 5 irreducible representatiolls of the cubical group, then a linear combination of these five must be
fo+d such that each class has the values as given by equation (8.3): This is
do-q.e simply by solving the five equations in five unknowns. The result is
gi)kn in the following table:
X

C.

e,

c,

Cs

rs

1
3
3
7

1
-1
-1
-1

-1
1
-1
-1

-1
-1
1

1
0
0

--1

T,

If.

L=3

(8.4)

Hence the conclusion is that the F' level decomposes in two levels which are
three-fold degenerate and one level '''hich is not degenerate. From this
simple >~ example, which can be worked out in a much more sophisticated
way<with the help of projection operators, we see that knowledge of the symrnetry is sufficient to determine the number of separate levels as well as
their degeneracy.
1 [iLl means the largest integer contained in !L.

236

[eb. 8,12

FINITE GROUPS

A word of warning is necessary. It is, of course, not strictly necessary that

the three levels mentioned above are really separated. Itcpuld be for instaDce
that certa!a1ll&trix elements of the perturbation are zero and hence\ that the
perturbation does not give rise to a splitting. If this treatment had been used
for a level which \vas characterized by L, it could be asked now what happens
if the spin is taken into account (spin-value unequal to zero) and if a spinorbit coupling term is assumed in the Hamiltonian. In this case it is ne~
to start with the product representation of one of the afore-mentioned r's
with the 28+ I-dimensional representation of the spin space and reduce the
resulting product space. This again is done with the help of .characters of the
respective classes. The characters of the product representation are equal to
the products of the characters of the composing representations. For
example, if it is assumed that the afore-mentioned F-level has a spin
S == 1, that is a 3p state such as is found in Ni++ Consider the r41evel
wbich is contained in the F state. The characters of the representations D 1 of
the spin are found with tIle help of the formulas indicated above equation (8.2)"
-1
-1

-1

1
-1

o
o

-1

-1

-r.XD1

-1

(8.5)

as a result we find the character: 9, 1, -1, -1 and 0 and have to investig~te

~hich crmbinations of the characters of the cubic group will lead to this set
of num~rs. It turns out to be 1 + 3 + 4 + 5. Hence the conclusion is
that th~ three-fold degenerate ground level whic~ after multiplication by
the spili variables became nine-fold degenerate, dislntegrate~in a single level,
a twq-!fold degenerate level and two three-fold degenerate Jevels~ The situa...

r r r r

I
-L-

:2

--L-

-L.

f\lO nELD
NO l.'S

cue
NOLS

::
--L-

t1

--L..

~-

L'S

CUSLS !:S CUB

NOFiElD f

NO F"IELD

NO :5

variOllS

,i
f

_.1

Fig. 8.1. The splitting of a aF-level under

conditions.

Ch. 8, 2}

237

CRYSTAL FIELD THEORY

tion is sketched in Figure 8.1. All the unspecified numbers refer to

the representations of the cubic group. On the right-hand side is the
fictional o? Case in which the spin orbit coupling is dominant. The figure
indicates the proper order as found in nickel but the distances are arbitrary.
Before going over to general considerations let us mention one more specific
case. Suppose there are two levels which are very close together. In this case
it is necessary to start out with the product representations of this
particular set of levels. Hence take the direct product of, say" L 1 and L2
and calculate the characters of this product representation . '"rhe next step is
again to see which linear combinations of irreducible characters lead to the
set of characters found from the product.
2.2. EXPLICIT CALCULATION OF WAVE FUNCI'IONS

We want to give an example of how a reduction actually is performed .

Suppose an atom with an L = 2 (D-state) is imbedded in a cubical field.
It is known from the character table that D.2 = 3 + 5, i1ie. l~ :: 2 will
decompose in the two-dimensional representation and one of the thr'eedimensional representations. .
Before treating the straightforward method, with the projection operatoft
we will illustrate the guess-technique usually employed for the lower values

r r

of L.
The L == 2 level has a basis which can be described by X2, y2, Z2, yz, XZ, \ xy
exclu4ing X 2 +y2+z2. The three last products transform like x, y, and z,
except! for inversion. Hence they correspond to one of the two three-dimensionalrepresentations. The possibilities with the squares are restricted since
only two are independent x 2 + y2 does not work, since it goes over into +.:2,
using the rotation around the body diagonal which makes x -. y, y -+ zand
z -+ x.. If this transformation is applied again, the result is Z2 + Xl. These
three functions together form the combination 2(x2 +y + r) which had to
be excluded. If the combination x2 - y2 is tried, it leads to y2 - rand Z2 - x 2 ..
The third is a linear combination of the first two: z2-x2 .. _().2_z2)_
(x 2 - y2). After checking all the other matrix elements it is found that the
first two indeed form a (non-orthogonal) set of basis functions.
Slightly less haphazard is the use of spherical harmonics 1 (compare
j

Table 8.3):
In the following calculation the proportionality constant is of no importance. In
1:tpecific applications it should be DOuced that different conventions are used for the sign or
phase' factor of the spherical harmonics. These differences are lilted by EDMONDS [19571
1

on p. 21 ..

238

Flt~ITE

o - 3z 2 ....... ,2
Y2""'"
Y21

GROt.!PS

feb.

_ ~-...
~_2
x 2 """"Yu 2
......

-(x+iy)z

Y22

8~

(8.6)

= (X+iy)2 .

Yi -

The three-dimensional representation is obtained by taking,

Y2- 2 and
1
..t Y2-- . 1~he only choiCes left are + y;2 and YJ. These correspond to
the polynomials .~2 - Jl2 and 2Z2 - x 2 - y2 and these are orthogonal.
,
The straightfonvard method is to use the projection operator (Chapter 3,
12.3). For these it is necessary to calculate the sum of a class.

Y;

yi

o o o 0\
1 o o
E = f 0 o 1 o
\\ no 1
\0 o o o

/~

~)

o
o

~/

I -- i

I
I

\ i

001

o
o
o

000
000
o -2 0
o 0-1
-2 0

Cs =

0 0
-0 0
0 0
-2 0

0
-1

o
o

o o
o o

--1
0

\~

o
o
I

o o
o 2 o
o o o
o o o
-1 o o
1

0 0
0 0
-4 0
0 0
0 0

-~\

~ ).

(8.7)

-2/

If they arc used in

S(ll)

= n" L X (i)C.
hiP.

,~

which is actuallyeq. (3.32) except that the character is real, we lind for
J.L- == 3,5 (all others give zero matrices)
p (3)1-I-2) /+'
?JIV'
__
....l..
'~

(0IZ(3)!0) ~~ 1;

,+'> - .1 ..

/.,..~t~(3)f
,TkCt
,_-

2'

< l[e(5)1 + 1) = t;

(216{S'f2) = -<+2(8(5)1+2)

(8.9)

=-i,

In this case the state vectors are already orthogonal. We find after normali-

Ch.

8~

2]

CRYSTAL FIELD 'THEORY

i1\
! I'

(:2)
, 1'\)'/ I~-;'i
.. !
- f - , t . J '\,J ,.{;".

r\ i

-1>~

The symbols} 1m), refer to the spherical harmoni~s; for practical purposes
a table for I > 4 is included (Table 8.3).
TABLE

8.3

Spherical harmonics

'We want to include, although this is superfluous, a semj .. quantitative

method" 1~he reason is that on one hand guess work may fail and on the
other hand the projection operators although straightforward are rather
tedious to compute. Hence it may be \vorth,vl1ile to look at the secular
determinant directly . As an example take L ==- 3 (P. . state) and again a
ell bical fre ld~
A cubical Held can be described by
Reub

= ar- 2 (x 4 + y4 + Z4 - 3r2 )

(8.11)

if the four--fold axes are chosen as the x, y and z-L1tis4

240

FINITE GROUPS

(Ch. 8, .2

In general the potential is of the type

(8.12)
This satisfies Laplace's equation~ since it is assumed, that the charges that
generate the field are outside the atom. The constants, ~ are the parameters
which one hopes to determine experimentally.
The calculation of the matrix elements with respect to a set of angular
wave functions of given L-value is found by integration over three spherical

harmonics.
(L, M'IYl'IL, M) == (4n)t

f yf'ytyfd(J).

(8.13)

unit
Iplter.

The integral is only tIJI: 0 if; (i) M' -= M +P and (il) the triangular rule is
fulfilled: A. ~ 2L. This means that the power series (8.12) actually is of no
importance beyond a certain term. The coefficients (8.13) are similar to
(5.42) (where 1 == 1) and in general to (S.A24) the Wigner or OebschGordan c~fficient. They are tabulated in the literature, l but for low Land p.
they can be computed quickly by employing the following integral

(ax+by+cz)211

=(4n)-1 I (ax+by+cz)

2I1

dm

oil

'pltor.

(8.14)

C;omparing coefficients of ~bPcw the expressions are found for x"Y-ZI.

Tale for example L = 3, then for Hcub the result is
"

3
0
0 -7
0
0
0
0
JfS 0
0
S
0
0

0
1
0

0
0

Jf5

0 J15
0
0
0
5
0
0
0
6
0
0
0
0
1
0
0 -7
0
0
0

Jf5

0
0
0
3

= (MIHcubI M ').

(8.15)

If rows~ and columns are interchanged this matrix can be reduced to one
one-dimensional and three two-dimensional matrices. The roots are A = 6
(3 times); ;. = -2 (3 times) and A. = -12 (once) in accordance with the
1

See systematic bibliography at the end or this

boo~ .

Ch. 8, 2, 3]

DOUBLB GROUPS

241

decomposition in_ (2.1). This leads immediately to the irreducible set of

basis fu~ctions. However, this example was facilitated by the easy way in
which t1ie matrix can be decomposed. 1 This is generally not the case.

3. DoDIJle Groups
In Chapter 5 it was shown that representations of the full rotation group
could be obtained with even dimensional (irreducible) bases, corresponding
to halfinteger values ofj. This leads easily to the problem of manufacturing
a similar set of representations for a finite groupo The irreducible representations of the full rotation group are usually reducible with respect to a
finite group and with the techniques of the projection operations the problem
is solved in principle.
Before such methods are applied it is necessary to have the character
tables of the double groups. In case of the full rotation group it was seen
that the characters of the double group are different from representations
\Vith the help of spherical harmonics (if they were the same the representations would be equivalent!).
The s.ame is expected to happen for the finite groups. At first inspection
it looks strange that additional representations must be added to the
character table, since a well-known theorem states that the number of
representations is equal to the number of classes and the latter is given by
the !abstract group table. It follows from this that we are actually not dealing
with the same abstract group.
The reason is that the representations derived from the spinon are not in a
one-ta-one correspondence with the different rotations of the finite group.
A' careful distinction must be made between a so-called non-faithful
representation and the case we are discussing here. In the ,non-faithful
representations a one-to-many correspondence is dealt with in the sense
that to one representation matrix there correspond several elements of the
abstract group. Here the case is just the opposite. We have two representation matrices) one with the plus and one with the minus sign, which correspond to one abstract group element.
As a result the double group ?I' actually consists of 2g elements and the
number of classes is larger than the number of classes in the original group.
SinG'~ the abstract group is different, of course there can be a larger number
The fa(,~orisation of the secular equation has nothing to do with reducibility. The last
ternl only refers to a set of matrices. ,"n individual (normal) matrix can always be com. .
p]etely diagonalized.
1

242

Flt~ITE

GROUPS

[eh. 8, 3

of classes. However, it is not necessa.-ry that the number of classes be twice

as large. This is an important point which will be discussed belo\v.
1heunit ~ie.ment E forms a class by itself; in tIle double group it will split
in t\VO classes E and R = - E. Tllese two elements have the same character
in "those representations \vhichthe double group has in common with f,.
tlley wiU have opposite sign in the additional representations.. l"hese two
elements E and R form an invariant subgroup, hence < and R~ form a

factor group.
We will now show that the oppositeis not true. With the opposite \ve mean
that < is not an invariant subgroup of the abstract double group. l1his is
easy to show since t is not even a subgroup.. The two spinor variables ,
and '1 will transform under a rotation 17: around the x-axis according to:

(8.16)
(compare (S.7b) with lp = 1-n; '" = -in and :1 =m~ = x).
lienee if the elements of t;g are all the elements of C' witll only one of the
two signs, ~ inverse element of A cannot be included, wbjch shows that !JJ
IS not a isubgroup. This simple<illustration helps to stress the fact that tIle
double group is not simply a direct product of the single group and the group
consisting of the two elements 1 and -1 as is for instance the case if inversjon
.symm4try is added. toa group. Hence the structure of the do.uble group is an
intere~ing subject of study. The structure of the group 2t: can be described
as fOnpws. The elements Band R. form a norm~ ~ubgroup.Al'" of order 2;
the f,ictor group 2f/% is isoJ)lorphic to the ordinary point group. This
str~ture is similar to the structure of the space group [/'. There too a
notmal Sllbgroup exists, the group of translations !T, and the factor group
!Y/.r is isomorphic to the group of the rotations, The group of rotations
is not a su,pgroup of f/, similar to the fact mentioned above that the elements
of t' do not form 'a-subgroup Qf2(. Both cases are examples of semidirect
products (Chapter 7, 4) of a group and an (operator) group. In both cases
we start out with the point group. The operator groups are either the
translations with the resulting semidirect product of the space group, or the
group cog$isting of E and R with the resulting semidirect product of the
do~ble group.
In order to establish the character table of the double group from the
original group fI, we first notice that if.the classes C i and Rei are different
their characters should be the same in those representations that qj' has in
I

Ch. 8, 3J

DOUBLE GROUPS

8.4

TABLE

Eulerian angles and values of

Ot:

P for

and

243

the elements of the cubical and hexagonal

double group

.-

Cubical

No.
1

n'

xyz

2, 2'
3, 3'
4, ~'

~i

Ttyi

'Pz

1
1

5
6
7
8
9
10
11, 11'
12, 12'
13, 13'
14, 14'
15, 15'
16, 16'
17
18
19
20
21
22
23

xly

zy~

i-

24

...

fJ
n'

V'

Element

~xz

xz5'
iyx!

t
t

-t

0
.+-"
-.....

.B(R'== -E)

C.

t/v 2
1/v'2

1/V'2
1/"\12

Cs

---t

i-

iJ-

_.*

1
1

i/-vl2

1/v 2

i/vl2

t
t
t
t

0
i{l +i)
!(1-i)
i(14-i)
!(1-;)
i(l-;)
i(l+i)
l(l-i)
t(l +i)

i!V2
(1'+i)/,l2

i~;f

-1

'9%1

xzy

j;'
0

zYx

i-

-i

*!
i-t
-i-

fliy
yzx

0
0

z.iy

0
0
0

0
-j-

-!

i:xy

Class

~iy

zxy

{J

:1f.

yxz

yx1

0
1
1
0

cp
<-

j1ix

0
1

yig

YZ$

-j-

-i
-t
-1

(1+i)/v'2

1/'\1'2
1/v'2
(1-i)!v2

-li,/2
0

~ilv2

1/v'2

=t=i/V2

i/v'2
=F (1--1)/v'2

1
1

-!
-t

*!

(Ca' = -C~)

--i/v'2

C.

1(1 +i)
!(l-i)
'-i(l +/)
-1(1-i)

C,

-!(l-t-;)

(C,,'':'-' -Cal

-t(l~i)

-i(l +0
1(I-i)

Hexagonal

Set 1: fJ = 0,
Ct = exp (bln/6)

Class

Element

Set 2: = 0,
fJ = exp,(inn/6j'
Class
" .....-----

Element

R
C~
('a

Col

4'

0
6
3, 9
2
10
8
4

11
7
5

1
I'

2,2'
3
4
3'

5'
6'

Cs
(''''6

-~-~-

--.----~

71 7'
8, 8'
9, .9'
JO, 10'
11,11'
12, 12'

#'

J, "

1, 7
5~11

0,6
2,8

4,10

,.

244

PINrrE GROUPS

[eb. 8., 3.4

common with fI and should be the opposite sign for the additional representations.
Tho intcnstiq question when the classes will double and when they will
Dot has been studied by BBtHB [1929] and 0PBaI0wsIa [1940]. Opecbowski's
major result is: if there is a rotation around an angle 2E thel) the elements
c. and llCs beton, to the same class if, and only if, there is also ano~er
rotation throuah n around an axis perpendicular to the axis of the first
rotation in the sroup.
On the basis of this result he shows that the knowledge of the character
table of the 'liDsle" &roup is sufBcient to construct the character table of the
double lfouP. Almost all character tables can be found in the literature
(BBnm (1929): hexagonal and cubic; OPEcs:OWSla [1940]: rhomboedric and
tetrahedral; ELuan [1954]: for the space groups; JABN [1938]: for the
~ groups).
In Table 8.4 the necessary maredients to construct expUcit representations
of the cubic and hexasonal groups (and their subgroups) have been indicated.
The two-dimensional representation is obtained by writing the matrix

a.*
( -,.,
II

P*)
ex

or

(txP* P)
-

Rep~tatioDl

o( hi,her dimensions can be obtained by the direct product

prodedure
exp1liDed
in Chapter S, 2.2. The resultiog representations arein
I
.
gentjfat reduci'g-" d can always be reduced with the help of tho projection
oW-tor methocl described in Chapter 3, 12.3. The mpst elaborate task is
calculation of the elements containing rotations around an axis perpendicular to the z-axis. These matrices have been tabulated by EDMONDS [19S71
and MEDEIl [1954] U < 5).

In order to study the inftuence of a mapetic field on the spin of an electron

Won. . to an ion iDaide a crystal, tho perturbation theory will have to be
extended./1be problem iI one of deaeneracy but not quite the same as was
dealt with-before. 'The difference is that all the levels dealt with are degenerated with the same de~J)efaCy while in the previous considerations only the
level under consideration was degenerated, while all others were single.
Another difference is that degeneracy is lifted only in the second order.
/

Cb. 8. 4 J

OPERATOR HAMILTONIANS

245

In Chapter 2 the perturbed energy up to the second order is Jiven by

2

ElI.pert.

==

1U
12 IH...1
E,,+AO
... +A E -E
"

(8.17)

III

where H." represents the matrix elements of the perturbation operator with
respect to the UDperturbed eigenfunctions and Ell the unperturbed eipn-

values.
We will show below that a similar formula can be given for a set of levels,
each of which have the same desenoracy r. The difference is that the symbols
H.. are now r by , matrices instead of numbers. The energy Ea...... is hence
expressed as a polynomial of matrices: the operator Hamiltonian. In I 4.3
we will see that, unless we are interested in numerical values of the constants,
perturbation theory is not necessary and the operator Hamiltonian can be
constructed on the basis of transformation properties only_
The situation, particularly in transition-element ions, is such that each
level is characterized by an orbital quantum number L, a representation
label r, and by a label that numbers the different components (irreducible
basis ~~tions) of that representation. Since the 2L+ 1 wave functions
belo., to L usually decompose into several
which will in
be
1
relatively far apart (about 10000 em.- ), the set of levels E" each be1onaiDI,toa certain
can be considered the unperturbed set in the same way as
above. 'ifhe eigenvalues E", E"" etc. were supposed to be far apart.
If th~ total spin function is introduced, we will find that the degeneracy
,factor qr every level is multiplied by 2S+ 1. The perturbation created by the
maanc~c
field will not be any different from a free iont i.e. 2S + 1 equidistant
!
leve1fJ.; unlessJhe spm-orbit coupling is introduced. This interaction will
result in the symmetry restriction being "carried over" into spin space.
It was already demonstrated how a certain orbital symmetry
aadthe
representation of the full rotation group in spin space ~. can form a direct
product space in 2.1.
The spin-oIbit interaction has no effect in the first order since
spin and
all orbit matrices have trace zero. We shall see in the next section that.tho
second order perturbation can be calculated in a way that bypasses the first
order and gives a closed expression for the energy splitting and the g-factor.

r,

aenn

r"

r,

an

4.1. VAN VLBCK

PB~TURBAnON

THEORY

We want to formulate the perturbation theory with reference to tho tranJition element ions where the crystal field effect can be assumed to be larger

246

F!NITE (JR()tJPS

[eh. s. 4

than.'tn'e spin-orbit couplingr The first step in thls problem is to diagonalize

the crystal electric field in the space spanned by the orbital part of the wave
functio~. It is at this mon1ent that th.e considerations of group theory from
the preceding section enter. l"lle sYlumetries involved make it possible to
calCulate the projection operator (compa.re Chapter 3, 2 . 3) that Will lead
to the proper linear combinations of,vavefunctions* Although this 'procedure
is straightforwar~ and does not involve any diagonalization of matrices it
ma.ynevertheless 'lead to some practical difficulties, particularly if the
syrrinietty~l' Idw'.'ln this case the chances that a certain irreducible component'
occurs more than twice are large and most of the advantage of the method
is lost since it is again necessary to diagonalize matrices of~order larger than'
two. Hence it may be profitable to separate tb.e electric field components
into a lirge contribution and a small contribution, because the large part
usually represents the main symmetry of the ion site in the crystal.
Supp'ose the wave function adapted to the main symmetry is obtained in

this way:

>+ ;:

Ii, kt ) = ILg, lvJL ) (Lu-,'.f1Ji, k c

-

L#4

IL(, .~lL> <LkfJ.li~ ki>

(8~18)

where the label i refers to the different energy eigenvalues in the electric
crystal field and the label k j = 1, ... , nj referfi to the cOlnponents of that
partiCUlar energy level. The degeneracy of tb.e level i is equal to nl" The
stubscript g refers to the ground state. rorhe ground states are known from

$und's:riiles. It could happen that the large electric field terms give rise to a
~gle ground state, in this case tIle orbital angular mome,ntnm is COl1sidered

ito be quenched, since the thermal energy is usually much smaller than; the
distan~~to:'~the nex.t higher level or group of levels. The result is that only
the lower level i is OC(:11pied and since the expectation value for the angular'
1110mentum of a single level is zero, the orbital motion does not manifest
itself anymore, or is "fr,')zen jn.H~ The other pos'itbility, namely that the
lowest electric field level is not single, occurs less often . It can be described
by a~~pseUdo angular mornentunl (ABRAGAM and PRYCE [1951]).
We wilt restrict ourselves fronl TI0,\V on to the first case of anon-degenerate
ground state.
Ire'h~ry electron ,vave functIon is rnultipUed by 2 S' ~F 1 spin "rave fllnctions,
the degeneracy of allleveis is mulb,plied by a factor 281- 1. It seenlS at fast
sight that this prohlen.l is rt~jated to the degenerate perturbation problem
m,entioned in Chapter 2~ '102 and 7.3. 11JC point is~ however, that this
problern it; slightly mort": com'P!iccJe(~ 8in':~G' the :spin. . lJrbit couplJng has the

Ch. 8, 4]

OPERATOR HAMILTONIANS

247

property that it does not lift the degeneracy in first order perturbation.
~ence ~ formalism has to be used that handles the removal of the degeneracy
~n the ~ond order. This'is conveniently done with help of th~ so-called
InteractIon representation. Transformation matrices are introduced such'
that the spin orbit coupling is transformed away in first order and a result is
obtained which expresses the Hamiltonian directly in the second order
contribution of the spin-orbit coupling. In the next section it will be seen
that the result thus obtained is actually nothing but a special case of the
tensor coupling between the orbital angular B1Qmentum operator and the
spin angular momentum operator such that the internal product is an invariant. The method, originally invented by Van Vleck (see SCHIFF [1949])
is as follows:
Let the eigenvalue to be solved be expressed as:
(8.19)
if the matrix element of these operators is taken (the electrical field
contribution is already diagonalized):
Of,

>

(j, M;LHe.+Hsoli, M s = E!O)(j, M~I~ljli, Ms)+(j, M~IHsoli, Ms).

(8.20)

Consider the following formal transformation of the state vector U into

V == (exp-S)U which leads to the transformed Hamiltonian,

H'!=- e-sHes == H.1+H.c,+[Hel' S]+[H80~ S]+t[[Helf S], S]+ .... (8.21)

r

In }the above development use has already been made of the fact that the
spinorbit coupUngis assumed
be small compared to the- large electric
o.'ld terms, an assumption which is reasonably well fu1fiJ1ed in the case of.
iransition'elements. It is not necessary to make a similar restlictionfor-the
small electric field terms since only the ground level has been taken into
ronSlderation and this was assumed to be sinafu so that the 'only contribittiOB
of the')'smaU terms will be a slight shift. Now choose the transformation
operator S in such a way--that it removes the first order terms

to

HS() + [Hcl' S] =

o.

(8.22)

If this relation is substituted in the previous equation the result is

H' == He1 +[H.c" S]+![[H.1 , S], S] == He1-![H.c" S].

(8.23)

Equation (8.22) can be solved very simply because the matrix eletnents of
1he electrical part are diagonal. Hence we have:

248

FINITB GROUPS

[Ch.8.1.

(8.24)
where E, and EJ are the diagonal elements of HelMU... If this result is
inserted ii! the transformed matrices H' the foHowing result isfounii for
the lowest level g:

= E~O)-i<gl[H..,t S]lg)

(9IH',,)

= F,0)'- ,.,
L (uIH.oII>(iIHlOlg)/(Ej-B.).

(8.25)

After introducins the explicit expression for the spin-orbit coupling

(8.26)
the final result for the so-called spin-Hamiltonian is found to be

(gIH'(S)lg) == _{2 L AlISkS,+E~O)

til

with
Aid ==

L4 (gILtll><iIL,lg)/{E,-E,).

(8.27)

The gen~al symmetry considerations imply that this linear combination of

spin operators has tO,have the same transformation properties as the electric
field potential. Hence the important question may be asked, how much of tho
exprtssion could have been constructed from the knowled.se afthe symmetry
alon,?This question will be treated in the nQXt subsection.
'J1le,expression obtained for H' as a function ofthespmqoperatormay or
ma3 not he,diaaoual with'respect to the spin quantum aumber. After dia~
D~i~~~if: aecesaatY, the second order correction to ~;oaer&Y:lovola.dae
~.the spinorbit.iIlteractionis found. We would like to r~.th.is,resultin
~terma. A spin by it$elf(frec spin} does not interact with an electrical

if

~,

:lie1d~} ~use

<

a spin:is of mapetic nature only_ For.abound. spin ~ ,s~

ment still hol~ inthe firstorl', but in the next approxmllwon the pi~o
changes. The field interacts with the orbit, that is to say (using classical
laDauase) the orbit is not entirely free to orient itself and hence the spin,
which is to ~ ce~ . extent attached to the orbit, also. partially loses its
orien~ freedom. Such an indirect interaction can, in quantum mechanics, only be obtained through at least second order perturbation since,
in the laDguage of time dependent perturbation ,9teory,enersy has to be
'borrowed' temporarily and hence we deal with. a second order process.

Ch. 8, 4]

249

OPERATOR HAMILTONIANS

We want to illustrate the spin..Hamiltonian pertUrbation treatment with

the case of the aforementioned alums. The main electric field is cubic and
there ~s ~ small electrical field added along the body diagonal of the cube,
which is taken as the axis of quantization. It turns out that in this case the
spin-Hamiltonian is already diagonal for the following reason. The spinorbit coupling operator in second order has the selection rules L1Ma := (j~
+ 1, 2 and dML == 0, 1, 2. The potential mentioned has p. = 3,0,
which gives a selection rule AML == 3, o. Hence the only contribution will
come from AML == 0, this gives AMs =: 0 and hence diagonal spin opera~ors.
This conclusion does not hold for the third order calculation, or for a
potential of lower symmetry.
Since most infonnation is obtained in experiments that are done in a
magnetic field, replace the operator H.o by

H.o+H.nr = CL S+J.(L+90 S) H

= <,S+pH) L+go/lS H.

(8.28)

where H.t represeats the contribution to the Hamiltonian due to external

mapetic fteld. This field is, in almost all experiments, of such an order of
magnitude that H.r -< H.. and hence can be considered as a small perturbation.
Assuming apin that the spin-Hamfttonianis diagonal we see from ~e
previous equation that {8 must be replaced by,S +pH and the spin-fteid
tertii must be added. For the magnetic field dependent spin-Hamiltonian the
i

resUlt is

H..+Hmf == gopS H ==

r'J A,J.CS,+p.H,XCSJ+pH

j)

r (pg'JS,H ,2Ai) Sf S + p.2Ai} H, H J)

J-

(8.29)

I.J

The dimensionk* coefficient of the first order term in H is the so-called

,.factor
g,}

= 90 a'J-{Au.

This factor can be determined experimentally~ If we have a

(8.30)
good estimate

for' it allows us to determine A.'J. Since the result is the difference bet. .n
two terms with the same order of magnitude the accuracy is usually low.
4.2. TENSOR: OPBRATOR

To illustrate how spin-Hamiltonians can be constructed from symmetry

considerations only, take the simple~ but often occurring example t of a

250

FINITE GROUPS

(eh. 8, 4

crystal in which cylindrical symmetry around a certain axis can be assumed.

If this axis is taken as the axis of spin-quantization (i.e. the z-axis of the spin
matrix isJaken along the axis of the cylinder) the spin-Hamiltonian up to
. "
second powers in S has to have the form:

H = Eo+D[3S;-S(S+ 1)]

(8.31)

where E and D are two constants. The next term would be the polynomial
or Y~ in ~ we require inversion symmetry. How high
equivalent of
a power of S is included depends on the number of levels, that is on the
total spin value, of the system we are describing. In general it can be stated that:
The number of invariant tensor operators necessary to describe a perturbation on a state of 2$' + 1 components is finite, since the total number of
independent Hermitian matrices is (2S+ 1)(S+ 1).
The generalization of the idea illustrated above is of particular importance
for the work in rare earth ions. In this case the spin orbit coupling is of such a
strength that Land S are not good quantum numbers and the electric field
will have the effect that the levels characterized by an orbital momentum J,
which has a degeneration 2J+ 1, will split into a certain number of groups
of levels. It_wQwd be impossible to calculate this by the previous perturbation
theory- since in this case the L S coupling is not a small perturbation
compared to the electric field, but the rever~e is true. However, it is possible,
,although rather tedious, to take the field as a perturbation but since it acts
only ion the orbital part of J, it is first neeessary to decompose the J.
In imany problems, particularly if the components of the multiplet 1 are
far a~art, it is sufficient to give symmetry considerations and postpone the
de~imination of the constants involved.
Although the principle for such a treatment was already indicated by
wigner arid Kramers, the first practical evaluation was made by Stevens
*~o_,~cu1ated the operator polynomials corresponding to Yf, Y~,
as
well as the multiplying factors for the ground state of each of the rare earth
ions. The first results are indicated in Table 8.S.

Y:,

y2

TABLE

~(lzl-r')

8.5

S (x1'-(3J.:+J(J+ 1)]

:E(35z4-3orzs +3r&)

== (Jr4[3SJ:-30J(J-f-l)J!+25J:-l-6J(J-+ 1)+3J2(J 1-1)1]

:E(231z4-31S"'z4-105r' zl-Sr4')
i'''''[231J!-315J(J+ 1)J:-1-735J~+ l05JS(J+ l)j;:;
-525J(J+ 1)I:+294J:-SJ8(J+ 1)3+40JZ(J+ 1)=- 6OJ(J+ 1)1.
1

A multiplet is a group of leveis!' originating from the sarne Land S value. Tbe compo-

nents are J

= L+ S, ... :L- Si.

Ch. 8, 4]

OPERATOR HAMILTONIANS

251

We will now give a short account of KOJ.UUNGA'S [19S4] treatment abo.,

how operator invariants as well as operator covariants can be constructed.
An operat6r covariant is comparable to an irreducible set ofbasis functions as
shown below. A well-known operator covariant is the operator veCtor
(S, L or 1), the three compQD.ents form an irreducible set of basis vectors
with respect to the full rotation group.
IT an operation G, corresponding to an element Of of the point group of
the crystal is taken it will induce a transformation in a certain set of basis
functions. If the basis functions are changed, the operator G will undergo
a similarity transformation

0; =

S-lG,S

Instead of us~ng an arbitrary transformation on the basis functions.one ofth~

transformations of the group itself could be used. The number of different
operators obtained by repeated transformation, cannot exceed the order of
the group h.

(j = 1 ... h).

(8.33)

Since these operators are linear and since there are only a finite number of
different operators available then each operator generated in this way from a
given G f can be written as a finite sum of operators.
m

G;(j)

= L a(j)il G ,

m < h.

(8.34)

'=1

A sond condition on m is that it cannot exceed the number -!n(n+ 1),

where
n is the dimension of the space, since this is the total number of
I
independent Hermitian matrices.
~ow ifj runs through all the elements of the group a representation with
matrices (all) is created. The operators play the role of basis functions. If
certain linear combinations are chosen the representations may be irreducible
and an "irreducible set of basis operators" or a "covariant set of basis
operators" is found.
Besides the terms in the full Hamiltonian that describe the interaction of
the electric field and the electrons (1) we have terms that describe the
coupling<betweel) two sets of degrees of freedom (2). Following Korringa
we notice that these two sets of terms have counterparts in the spin-Hamil-

tonian:
(1) Operators that are invariant under the group t th.at corresponds to the

252

FINITE GROUPS

[eb. 8, 4

(main) symmetry. of the crystal. These operators usually operate in a space

that forms an irreducible representation of the full rotation group_
This is not strictly necessary, one could imasino the use of representatioDl_
irreducible with respect to an intermediate group; i.e. a group that is a
subgroup of the fujI rotation group. This choice is sometimes important in
order to secure an unambiguous set of quantum numbers (compare ~z
PBLD and MBDm.t [1960], Chapter 4, 8).
(2) Operators in a product space of two (or three) of the following spaces:
the space of the orbital electronic wave function II., the space of the electronic spin functions It., the space of the nuclear spin Il.. - The bases used
in these spaces are irreducible with respect to one of the three groups
mentioned under (1). Take for instance the electron spin-nuclear spin couplinS in a sys~.embedded in a field of cylindrical symmetry. The three
tronspino.peratorsS+l == S%+iS" So = S"t S-1 == S.. -iSyeachformatbreedimensional, and hence irreducible, representation of the crystal symmetrygroup. (However, the first and the last are representations of the second kind.)
The .three nuclear spin operators 1+ 1 , 10 and I -1 form a three-dimensional
representation .irreducible with respect to the
rotation group. The reduction of the list set of basis operators, in the case of the subgroup of cylindrioal symmetry, is obvious and the most general coupling, invariant under
cylindrical operations, will be

mea--

fun

(8.35)

whetf1 A', B', and C' are constants, or actually functions of the radial
coor4inate.
owever, this is not real (Hermitian), hence ignoring the imaginary
part we find
(8.36)
for the coupling." Jbis result is rather obvious in this limple example, but
for more sophisticated symmetries group theory is needed in order to
construct the linear combinations of operators that transform like an
irreduciblo basis for a certain representation.
Now to, each irreducible representation of a certain dimensionality n
belonp another, such that the direct product of the bases functions generates
the unit representation, i.e. a sort of genera1ized internal product is taken as
in vector analy$ia and' an invariant is obtained. This adjoint set of basis
fUDCtions forms a space of the same dimensionality and it is also irreducible.

Ch..8, 14, 5]

XRAMEIlS' THBOIlEM

253

In vector language it is the contravariant set of components. The same holds

in operator space and by taking such an "internal product" that part of the
operator'1iamiltonian is obWned that descnDes, for example, the coupling
between the nuclear and electron spin.

s.

Kramen' 'I1leorem aacl Time Reversal

A level which is characterized by J has a degeneracy '1J + 1. In case J is a

half-integer this degeneracy is even. Kramers has discovered and proved that
an electric field will split such a level at most in a number of two-fold
degenerate levels. In the literature these leyels are referred to as Kramerspairs or Kramers-doublets and the expression Kramers-deseneracy is also
found. Only a magnetic field, which has a lower symmetry than an electric
field, as mentioned in Chapter 3, 1 is able to lift the last part of this degeneracy. The difference between an electric and a mspetic field can be illustrated classically as follows. If the time is reversed in a magnetic field the electrons will Bow in the opposite direction. The force exerted by the electrical
field _is Pl__both cases the same since the force is indepeDdent of the VC?locity.
In a magnetic field, however, the situation is different, the Lorentz force will
cbange its sign if the velocity has the opposite direction. In quantum meChanics the reversal of time is connected with complex conjugatiOn, as a siDipIe
look at the time dependence of the I-iamiltonian shows. The wave functions
1/1 *d 1/1. are both eigenfunctions of the H-o~tor belonging to the same
ene,!-gy value, provided the time dependent part is disregarded for the mom~nt. This is also true if the H-operator contains the electric:field strength. E.
It'is not true, however, if a mapetic field term is found in the Hamiltonian
since these terms are pure imaginary with the result that '" and ". do not
have to belong to the IaJM ~ipnvalue. First the proof of thi$ theorem will be
given in the same way as the original proof by Kramers and later we shall
return to the considerations about time reversal
t

5.1. XRAMEIlS' THEOQM

Kramers indicated two proofs of his theorem. The first one (1930) is the
most-.n~ and is base4. Qu. the properties of the coefficients of B and B
in the Hamiltoo.jan just mentioned. The second proof (1933) is less abstract
and has the adVasltage that it is ~ol, connected with actual calculations.
Suppose there is a single level characterized by J. The state vector is a linear

254

[Ch. 8, S

FINITE GROUPS

combination of 2J+ 1 wave functions /PM:

(8.37)
or in the Dirac-notation:

<I

(8.37a)

= <JM)(MI.

Previous considerations have shown us that the matrix elements of the electrical potential are proportional to a Clebsch-Gordan coefficient:
(J,

MlvtIJM') ex: (JM', ApIJUM)

M = M'+p,.

(8.38)

The Oebsch-Gordan coefficients have the following symmetry relation

(1M', A.pIJUM) = (_1)2J+M+M'(J

-M, A.JlIJlJ

-M').

(8.39)

The exponent of (- 1) is equal to:

2J+M+M'

= 2(J+M')+M-M'.

(8.40)

The first t~rmQf this expressi9Jl j !llways even since if J is half-integer, Mis
also half-integer. Combh.UDS (8.38) and (8.39) gives the following symmetry
relation for the matrix elements:
(-MIVI-M')

= (-l}M'-M(M'IVIM).

(8.41)

The /matrix elements of the magnetic field,

(MIPIIM)
(MI WIM 1)

= MgPBHz:

= tg"'(J + MXJ=FM + l)PB(Hx=FiHy),

obey a similar relation, that is:

<-MIW(H)I-M') = -( -1)M1-M(M'fW(H)IM)

(8.42)

as one can see n;adily from the equations above. The magt).eticfield behaves
under th~ transformation
. from M to - M' and M' to - M in "the opposite
..
way
the electric field. The secular equation for the coefficients (MI) are:

as

..

E(M'I)

==

L [(M'IVIM) + (M'IW(H)M)J(MI).
M

(8.43)

Ch. 8, 5]

255

KRAMERS' THEOREM

As a result of the properties (8.41) and (8.42) a second secular equation can be
formulated as follows:

E( -l).l+JI<i-M) =

I:
[(M'IVIM) + (M'I W( -B)IM>](-l).l+M(iM).
M

(8.M:

The whole equation was multiplied by (- l)J in order to avoid fractional

exp<?nentials.
The set of eigenfunctions of (8.43) is given '~by

la)

= IM><Mla)

and the eigenfunctions of (8.44) are given by

Ib)
If H is equal to zero,

t~e

== IM><Mlb).

(S.4Sb)

following conclusion can be' drawn:

(Mlb) = C( -l)J+M<al_M).

(8.46)

If (8.45a) and (8.45b) are eigenfunctions belonging to -the'same eigenvalue

then EIJ = E", because solutions of a homogeneous set of '-equationsare
proportional to each other. If (8.4Sa) and (8.45b) would be identical, except
fora proportionality constant, we would have:

(Mlb)

= C( _l)J+M< -Mla)*

= c(Mla).

(8.47)

Complex conjugation and substitution of M ~ for - M gives

c*( -l)J-M(Mla) = c*< -Mla)*

(8.48)

which cannot be fulfilled for a half-integer J, because' the equation

cc = (-1) 2J C*C

(8.49)

cannot be satisfied by a half-integer.

The final conclusion is that for H is equal to zero the eiaenvalues E. and E.
are equal as long as J is a half-integer. The usefulness of the aboveepnsiderations is not, however, restricted only to H = 0 but some conclusions can be
made in case there is also a magnetic field. We want to mention that the
reversal of the sign of the actual quantum number is, of course, related to
the rotation of the electrons in the opposite direction. Hence as will be seen
later, there is a close relation between the prGperties of the.,.matrix element
indicated above and time reversal. As a matter of fact itispossible,to give tbi,s
proof in such a way that only the time reversal property is used, without
going into explicit calculations of the matrix elements.

FINrrB OROUPS

[eh. 8, $

In classical mechanics a system will traverse a certain trajectory in space

aud if the~ diao, ,which serves as a measure of arc length., is reversed, the
orbit will in aenera! be the same but the direction in which it is traversed will
be the oppqsito. This statement ceases to be true in case there, are velocity
dependent forces as for instance the Lorentz-force or the Corjoli's force.
If the orbit curves to the ri&ht for forWard speeds, it will of course curve tothe left ifthe motion is reversecl.lnspection ofNewtons equation immediately tells the whole story. 1b.e acceleration term is time (reversal) invariant,
the force term is always invariant in.case it contains .only position coor",nates. In case it contains velocity components it depends on whether the function is even or odd in these components. For instance, air friction. which is.
proportiODal ~ the speed or absolute value of the velocity, Jives rise to an
even function. while the Lorentz force does not.
In case we work with an Hamiltonian formalism similar statements can
be made. The mapetic field is DOW represented by a vector potential. Tho....ult is the same, the orbit is invariant if one reverses the time, provided tho
mapetic fWd ~.inverted also.
Constderiit, the time dependont ,Schrlklinger equation without a magnetic
field, we see that complex COnjuptiOD and time reversal1eaves the operators
on the riaht and on the left side invariant. Time reversal is now postulated
equi~t to complex co!Jjuption of the Schr6dinger equation. In case a
magn~tic field is present invariance can only be expected, as illustrated
aboveJ in case time reversal is accompanied by a field reversal. This is a
reaso,1able proaedure aiDee jf the mapetic field producing device is incorpe.
rate4 in the system, the senerating curreDta would also be reversed under
~reversal and hence the field is inverted. The caae in which the spins are
tile. fielc1p~vciq apnt (a ~~t magnet) will be discussed next.
So far the Hamiltonian was considered to be without spin.
The HAmDtonian for a system with spin will be of the form.:

H -

1
~

2m

e)a +etp+pH-S.

( p+ -A
c

(S.SO)

The operation "of A -+ - A and , -+ , . will leave the first term invariant,
but fielcI,Nvenalin the second term has to be accompanied by a reversal ,in
aiaD. of all thne ~pin components. The result of these considerations istbat the
equatioDS Of quantum mechanics will be invariant under time reversal plus
complex conjugation plus reversal of the sign of the spin components. Hence

Ch. 8, S]

KRAMBRS' THEOREM

257

it is stated that time reversal is equivalent to the applica1;ion of the last two.
. The operator which reverses the spin components is

(8.51)
Kramers calls this the spin conjugated spinor. This idea can easB.y\ be
extended to arbitrary spin values.
In dealing with a II-Spin system product spaces can be easily constructed
from, and " and the corresponding operator will be a product of is,'., each
acting on one of the spins.
Having established the n-particle time reversal operator:
K := i"S~1)S~2) .. S~) C
(8.S2)
where C is the complex-conjuption operator.
It is of course necessary to demand that udoublc-time-reversaltt be equiva1eat
to the unit transformation. If this operator is applied to a Don-cieaenerate
wave function it is found that

(8.53)
henCe that the wave function would always vanish for odd values of n.
For even values of n time reversal corresponds to multiplication with a
phase factor
(8.54)
If ther~ is at least two-fold degeneracy in the case of odd n-values one may
have
(8.55)

with Ic2 t= 1 and the wave function is not equal to zero.

,'ReturniJJg to the case that H " 0 the following properties can be shown:
Applying the operator K to the representation of the mapetic moment of
the system

Jt

" x p,+goS,
=- Jl r",

(8.56)

1-1

gives, if the wave function is non degenerate:

(t/I, Jtl/l) == (Kt/I, (KA'K- 1)Ky,) =- -(y" Jtl/l).

(8.57)

Hence the expectation value or .,It is zero. The equati~1 (1.54? ~

KJtK -1 == - A' were used since the orbital part of (8.56) 18 purely UDaIl- :
nary and the spin reverses sign under this operation. For a Kramers doublet

258

we find in the same

[eb. 8, S

FINITE GROUPS

way

(t/ll' .$("'1)

(8.58}

-("'2' Jtt/l2)

hence the trace of the two by two matrix is zero and the "center of gravity"
of the two corresponding levels will not shift from the zero field value as
long as we consider the terlns proportional to H:

\ (8.59)

E = Eo+.;It H.

For an even number of particles it was seen that

th~

K-operator has the

property (8.54) and if "" = eft/! is introduced we have

Ktf;' = 1/1'

(S.S4a)

which ~plies tllat the function is real. This idea can be generalized in case
the function is degenerate.
.
FROBENlUs and SCHUR [1906] have shown that there are in principle three
different kinds of representations possible for a finite group. The first kind of
rePreSentation consists of matrices that are real, or thatare equivalent to real
matrices. The second kind consists of matrices that are complex, but complex
insucili a way that the complex conjugate matrices vlill transform in a way
equivalent to the original. The third kind of representation consists of com..
plex, m.atriCes, . but the comple;x conjugate transformations belong to an
irreducible representation which is different from the original. It is clear that
thi .cllaracters of the representations of the first and the second kind are
r~, while the characters of the representation of the third kind are imaginab.ItfthefullrotationgrouprepresentationsofintegerJvaluehavethesame
prppertyas the representations of the first kind, while the representations of
t).tesecond kind correspond to the representations which have a half-integer
{]' value. Since in the finite groups the distinction between the integer and half...
integer.values. ~f J.is lost, in. this case the distinction of the different "kinds"
of, representations has to be used.
The generalization of equation (8.54a) now is that all systems with an. ,
even number of electrons have representations of the first kind. The generalization of (8.55) is that all odd electron systems have representations of the
second kind.; The representations of the third kind are excluded since the
bilinear form
r

E=

L (1/1:, Ht/ln)

(8.60)

n=l

has to bean invariant and this excludes the possibility of such a represen-tation.

259

JAHN-..TELLER EFFECT

Ch. 8, 6]

6. Jahn-Teller Effect
6.1. INTR9PUCTION, EXAMPLES

In [1937] Jahn and Teller discovered the following interesting and useful
theorem: Unless a ~olecule is linear, stability and (orbital) degeneracy are
not possible simultaneously. The importance of this theorem is of cour~ that
certain configurations for molecules can be disqualified "a priori" and appli.
cations of this theorem are found in many places in the literature (see
bibliography).
In order to get an insight into the idea behind this theorem consider a
nuclear configuration that will have a certain symmetry which is maintained
if all distances are multiplied by a certain factor . This type of configuration
is called a similar configuration. In the case of vibrations, these similarity
vibrations are usually called breathing vibrations. Consider a configuration
with all the configurations similar to it and take the binding energy as a
function of the scale factor (or configuration coordinate). This function will
have a minimum, otherwise there is no possibility of obtaining a molecule
this way. Take the configuration coordinate corresponding to this minim1lD1
as a point of departure and study all other displaceIIJ.ent~ the'nuclei to see
whether they are stable or unstable.
Following Jahn and Teller, let us first give an example. Alineat triaiolDic
molecule can undergo vibrations in which the center atom is displaced
perpqndicu1ar to the cylindrical axis of symmetry. Sin~ the dis~lacc;Fetltd,
and ';-d are identical, we wfiI have E(d) = E( -d)$ To 'undets~hd'this
remdnber that the wave" function is characterized by' A =0, ~. 1, ~~2,' ...
(re~tively a tI, 1t, '6 etc. state) where ,1, measures th~'arilUl~t M~tiun
I
", : '
, :'
..
': "':"",}i,. ',"'"
around the z-axis, i.e. the cylindrical axis (comp~' Chapter 4,t1'4)~:The
states 1t and d, e~. are two-fold degenerate, correspo1\ding: to right~;a.nU;,16ft
circU.lar orbits. A left circular motion stays a left circtitar modon.'whethertile
molecule is bent, straight or bent the opposite way_ Thffitst cgo~'bVef'lhto
the last if the molecule is rotated J80 around 'the z":ws, h~ll,~.:thewJiy~
function rotated over -180C> ,,ill undo this and since;lhe' energy, beihg~aa
expectation value, will not depend on the phase factor we should have
E(d) == E( -d). This means that, assuming a. continuous energy curve, a
power series around the origin d = 0 will have only even terms, or that for
sm.all displacements this curve is a parabola. For A #:. o there are aetuaIJy
two parabolas,since the degeneracy \vas due to the cylindricaJ' symmetry
which js destroyed for d :1= o. (Figa 8$2.)

of

,~>'::

;',~"i,"

.>

260

FINITE GR.OUPS

o--______

[Ch.. 8, 6

:d,

~J

_ _ _ _ _ _ _ _ _o

,
\

lEa
I

I
I

,,

I
/

"

."

J"

In this cue the wave fuDctioD is of the type

w.

(8.61)

t/l(r, z, S) - tp(r, z) exp U8

8 has any value. If we consider a cWferent example, i.e. of a square

m~t ,. will aYe .pio an ,aDauW ~pc~ up U8 but 8 - 0.
J-.," __ ,que.ntU.m. DUm. Ais equal to 0, -1. 'We are oQly ia~
.~... _ ' " ~ A p,O and daJcn"" the t~opoSli9iUtic:I by their
,.OdtH.., ~ iancilia Pia. 8.3) instead of by. their ~tUm:B~_ A.
P.rca.m,;~ we Jeara api~ that' ,tho wayO fwiction is de~. abJce a'
traD.tOflDAtioD ".",,,., to ",. group 01 thU 1IIOleCllle will briD .,1. ~t9:~it2.
If d " 0 tho '~ will be lifted, if we compare E(+d) withE( -ti)"w.
,I';

find

<,

"

",:"'"

, "

, : '

,'/

""

tl!efottoWina scheme:
E 1 (+d) == E2 (-d)
E2 (+d) - E1 (-d)

(8. ,

(8.62b'

becaufe -'PoIitioD+tl" and nodeUne 1, SO over into 'PositioD--d"

and

nodeJiM 2 if ,Pis- 8.3' is rotated goo. The resultjD"eaerl1~' (indicated

in Fig. 8.3) .mudes a minimllm; except for accidental depneracies. That

Ch. 8, 6]

lAHN-TELLER. EFFBCf

261

part of the degeneracy which comes from the symmetry of the molecule will
be removed in such a way that one of the levels is lowered if the d :I: O.
Hence ~ deal with an unstable configuration. We want to stress tbatthis
example, and also the general theorem, shows that there is at least one type
of displacement which is unstable, and that this docs not exclude otlIer
displacements that are stable. As a matter of fact a stable diapJace~t has
already been postulated, the similarity displacement (or breathinavibration).
No need to say thAt one unstable displacement is aufficient to make the m0lecule go away from the initial configuration. This displacement is usually
referred to as a Jahn-Teller distortion.

V
~

DISPLACEMENT + d

............ d

,,
I
1
I

....-<~
V
OISPLACEMENT - d

NOOEUNEal,
o

- --0-------0---NOOEllNE 1

,
I

Fig: 8.3. Energy as a function of the displacement coordinate in a square molecule. The
nodelines refer to the nodes of the electronic wave function.
'6.2. NORMAL COORDINATES

In order to look into the general case it is nec;essary to makea sJisht digression into "normal coordip.ates" of vibrations in .mo~~ 'Ibil11I~
actually can also be consid~red as an appliaition of group theory, ap.d hence
is~ of interest on its own, but since this is a J)Utely classical subject it actually
falls outside the scope of this book. l TIle potential energy of a molecule under
nucle~_Jiisp1aCements

is

. 1 A Similar F9UP ~ ~blem exists in lattice vibratioaL .~ ~1DJatl~

nates are labelled by two types of4lJ.alltum numbers. one type retea:IDI totbe bJaDcIIeI
(similar to the bands in wave mechanics of solids) and the other type. rer..n..to ..
wave vector. (see PJmJm 119561.)
,

262

FINITE GROUPS

rCb. 8, 6

(8.63)
where qate the 3N coordinates of the nuclei measured from the equilibrium
positions (or any linear combination thereof) and b ,} represent either the
; apling CODIWlts in~:case the molecule is characterized by a set of masses

.coupled byweigJIdess aprhlpor the coefficients of the second term i~ the

., _ , p o w e r series around the equilibrium position in case a

more

physical picture is preferred. The generalized kinetic. energy is

T =

LIi a'i4,4J ,

(8.64)

where the a'i are. dependent on the masses. The equations ofmotion obtained,
for instance, from the Lagrangian I (the Hamiltonian is just as good, but
associated with momenta instead of the q) are

L Q'Jii} + L bijqJ = O.
J

(8.65)

We take a linear combination of q's:

(8.66)
(which, as w~ will see ,below, is again a special case of (1.3a in order to
solve the problenfoy diagonalization. Multiply (8.65) by Cj and sum. over i.
This gives by comparing coefficients

L cia,) =

hj

and

LI cfbl} ==

-AhJe

(8.67)

w~tb. the c, determined this way the original set of equations becomes:

..

(8.68)

Q+lQ == 0,

which has the solution

Q(t) == Q(O) exp icot,

(0)2=:

A).

'the":'coefBcienu' ~ c, are determined by

E(la IJ - blJ)c, ==

'.~

l'

o.

(8.69)

'
<

These equatiCclns are basically the same as (1.14) since by l1lultiplicationwith

(aU 1 J the first term. will be diasonal. Solving the 8eCl11ar pr~blem will give a
set of:eipJ;ivaJues At and 'to e&cb. a set ofcoetlicients'4:f') which.iatum doter, Ir\'le.tart *ithNewtoo"s lawF == ma, and there is Ktually no eompoDhll reason to
use the Lagranaian formalism, the kinetic energy wiD eontain only diagonal torma.

Ch. 8, 6]

lAHN..TELLER EFFECT

263

mine the coefficients h)k>. The coordinates

Q}k) ==

L h~k)qJ

(8.70)

are the normal coordinates of the problem. Each of these coordinates vibrates independent of the other, with its own frequency and amplitude. 1his
amplitude is determined by the initial condition only. This result is applicable to a molecule but also to an infinite solid and leads in the last case
to the well-known lattice waves, characterized by k, with their corresponding frequencies (J)(k).
The degeneracies in l are connected with thesymtrtetry of the molecule.
The same statements hold for solids where different k-values mayeorre&pond
to the same t. The study of these degeneracies can apjn be undertaken by
group theory since they are the result of the fact that those transformations R
in laboratory space which are allowed by the symmetry of the molecule
(lattice) will induce a transformation in the normal coordinates. The result
of a transformation will be
a) if A is non-degenerate:
RQk

== Qk 1

(8.71)

b) if l is r-fold degenerate:

,.
RQk

== l:aldQ,

(8.72)

1= 1

and/the coefficients (ale,) fonn an r-dimensional representation of the symmiiry group. This representation will be irreducible, excluding accidental
degeneracy.
Take for example a triangular molecule in a two-dimensional plane. The
number of actual degrees of freedom is 3 (i.e. 6 minus 2 translations minus
one rotation). One is the "breathing" vibration and the other two fonn a
degenerate pair~ called V2.( and vu(O) (HERZBERG, [194~], p. 84).. The
elements oft.be group (rotations of in around the center) transform one ~to
the other, or mto a linear combination of these two. This (real) twcrdimensional representation has non-symmetrical matrices. It can of course be
symmetrized, but then the matrices will no longer be real.
1 Notice the differonce with quantum mecbanka where aU cootBcicnta are --;omplox
and aU matricea unitary" The only phase factor in this problem is :1:1.

264

[Ch. 8, 6

FINITE GROUPS

6.3. GENERAL DESCRIPTION

Normal.,.coordinates are used as a new system of axes in which the displacements of the molecule are described. In order to do this they have to be
normalized, and a positive direction has to be indicated for each of them,
If the magnitu~e of the normai displacement along the r-th coordinate is
called '1r, the deformed molecule has cooi~inates,

(8.73)
where QO refers to the normal cpordinate representing the original shape of
the molecule (the corr~nding vibration is the breathing. vibration) and flo
is the scale factor and has the value corresponding to a minimum of the
energy. The Hamiltonian is a function of these t1, and close to equilibrium
(i.e. the equilibrium of the QO coordinate) it may be expanded as follows:
H

= Ho+ V

= Ho+

Lr y"(q)'1,.+ r.Ls v,.s(Q)l1r'1s

(8.74)

where V = V(Q) and hence V(q). If we want to determine the energy by

perturbatiQn calculation of a p-fold degenerate level the secular matrix
has to be solved; the elements are given by:

(n, m = 1 ... p)

(8.75)

and the
first order correction on the energy is found after diagonalization
of
!
'
this matrix

(8.76)
IThe energy is a scalar, i.e. has to transform like the unit representation.
If the matrix elements (8.75) are studied they transform according to:

r",.,xr,xr", = Lair;

(8.77)

a direct 'product of the p-diDlensional, representation of the wave function

r.;, the 'representation of the complex conjugate wave function rp.(usually
the same basis, but not in systems with an odd number of -spins) and the
r-dime~siC?~a1 representation described by (8.72). The tight-hand side of
(8,77) .
the irreducible representations contained in this product. If
at = 0 there is no "scalarn representation possible and hence Vam = o.
This happens to be the case in linear molecules. In the D 4 -case at #= 0 and
since, apart from accidental degeneracy V"m #: 0> there will be a linear term

are

Ch. 8, 6]

JAHN..TELLER EFFEcr

265

in the energy. The main part of the Jahn-Teller proof consists in showing
that for all non-linear molecules, the direct product (8.77) contains the unit
representation.
Again take the example of a two-dimensional square molecule. There are
8 - 3 = 5 vibrations possible. That is the breathing or totally symmetrical
one, plus 2 non-degenerate (called B tg and B 2g ) and one 2-fold degenerate
vibration called Eu- (Compare HERZBERG [1945], p. 92.) The wave fun'ction
can belong to one of the five representations of this group. (Compare
BETHE [1929] or HERZFELD-MEIJER [1961]) but only one is degenerate
1"4 = E. Multiplying [r4 ]2 by B 2g we find that it contains r 1 == A g (The
same holds for BIg). Hence these two modes lead to decomposition of the
molecule, or at least to a non-square distortion.
How mu~h the molecule deforms depends on the actual shape of the
potential curves in Fig. 8.3. It may, for instance, be that the curves will go
through a minimum and increase again as suggested by the dotted part.
The distance from this minimum to the origin is the final amount of distortion. This distance may range from infinite, in which case the molecule will
dissociate along this mode, to very small. In the last case the effect may not
be detectable. This means that the application of the theorem has to be
accompanied by an order of magnitude calculation. The theorem also holds
for spin-degeneracy, a statement that brings up some interesting details,
but Teller (See JAHN [1938]) estimated that the order of magnitude of the
dist~nce frOln the origins is so small that the implications are not important .

PROBLEMS
1~

1. Show that the eigeJlYa!ues of a unitary matrix have absolute value

one .
1 2.. Show the invariance of the trace by transforming the definition
(1.17) by a similuity transformation ( 3.5).
1. 3. Is the product of tWo Hermitian matrices a Hermitian matrix?
i

The same question for a unitary matrix.

1. 4. Show that every unitary matrix U in a unitary vector space can be
'Written in the fonn

= elS

where S is an Hermitiall matrix. The exponential of a matrix is

defined by its power series expansion.
1. S. Show that, if .:tn and %~ are orthogonal if they are two eigenvectors
in the unitary space a Hermitian matrix that belong to different

or

eigenvalues.

2. 1. Suppose that a particle is described by the wave function:

'" :=

f(r)

(2Y; + 2YJ+iY8),

calculate: L 2 1/1, LJI, tho PtQbability that a measurement ofL.,will

give zero, and the expectation value .of L2.
2. 2. The operator A is defined by the equation:
At/I(x)

==

l/I(x+a)

where a is a real constant. Verify that the function u,,(x)"<exp (ikx)

is a;r1 eigenfunction of A if ut(x) is a periodic function \~lith the
period a~ tc. ul(x+a) Ut{x). Is A anltermiiian 'operator?

Express the eo11l1ilutator 'of

xand A: ,in t#rms of A.Consider the

operator (A-l){aas a g<)es to zero, and give ita physical interpre-

tation .
267

268

PROBLEMS

3. 1. Determine from the -group table of the symmetry group .93 the
minimum number of eielnents necessary to generate the complete
:~roup.
3. 2. Make a group table similar to the example in Ch. III~ 2 for the
tetrahedral group. A geometrical representation of this group is
found "by taking the four points (1, -1, -1); (-1, 1" -1);
(1,1, l)and( -1, -1, 1) in three dimensional space and performing
all spacial rotations that bring these points into coincidence. Is
this group the same abstractgroup as .9'41
3. 3. The ordinary complex numbers are a special case of hypercomplex
numbers. The basis of this algebra is related to the Abelian group
of order four. The hypercomplex numbers are

, = a(l)+b(i)+c( -l)+d( -i).

Construct the regular representation.
3. 4. Show that the regular-representation matrix A given as example in
10.3 satisfies the multiplication rules:
A-1E=B,

A- 1 A=E,

A- 1 B=A, A-1C=F,

A -10 = C and A -1 F = D.
3. 5.. Obtain from the regular representation of ~5P3, as constructed in
problem 4, the two dimensional irreducible representation by
means of the projection operators. The character table is given in
12.
'3. 6~ Show that the orthogonality relations (3.16), (3.17) and (3.18) are
fulfilled for the example given in 7.4.
3. 7. It is stated in 11.2 that the matrix elements ail,
are linearly
independent, i.e. the form

a;t

LAiltaik + LAj1aj,+ .. + LA;:')a~~) = 0

it

3., ~:-

jl

,UY

is only equal to zero if all A, ;! ... are zero. Show that this is correct.
The matrix 1 is independent of the operation D. (Hint: multiply by
a:!") and sum over a.)
Calculate the characters of a regular representation and show with
the help of (3.23) ,that the regular representation cQntains every
irreducible representation as many times as the dimensionality of
that irreducible representation.

269

PROBLEMS

3. 9. The 24 operations that bring a cube to finalpositions.i.adistinguishable from the initial positions form. a group_ This group is isomorphic
with the permutation group of four objects (the body diagon.a1s,
e.g. we number the corners such that two opposite corners carry the
same label). The symmetries of a tetrahedron.imhedded in, the cube
(compare Fig. 7.1) correspond to the elements of a subgroup of the
cubical group. Establish the correspond,ing permutations. Does this.
set of permutations form a symmetric group of lower order?
1

Drawing by Steinberg;

1960 The New Yorker 1vtagazine, Inc.

Fig. 3P.l

3./ 10. A cube in the conventional sense need not to have cubical SymDletry
(compare Fig. 3P.l), conversely there are objects that do not look
like a cube, and have the cubical symmetry. To find examples,
check through. the literature: Wigner-Seitz Unitce11 for b.c.c. alld
f.c.c.lattices (compare KrrrEL (1957], p. 286). Brillouin zones and
Fermi surfaces for cubic lattices. (Compare H.AItlUSON" [1960].)
3. 11. Let any two numbers a and b be equivalent in m - n == 4K where K
is some integer. Under this equivalence any number is equivalent
to either K = 1, 2, 3, or 4. Notation

m == n (mod 4).
(This is pronounced m is equal to 11 modulo 4.) Show that the
numbers 1,2,3,4 form. an additive group by constructing the group
table of the group. How many classes does this group have1

210
3.

PR.OBLEMS
12~

(CompareCh.m7.S.)Ifthebilinearformisindicated by F-(x.I)
and the transformed version (Ax Ax) by P<C), show that

I=

L F()

(a are all elements of the set)

IJ

is invariant under the operations of the set. If a linear transformation 1 is introduced that brings .1 into the diagonal form, th~ new
set of coordinates form the basis for a unitary representation of the
set ofm.strices. Show this and apply this procedure for "unitarisation" to the example (lfthe two dimensional representation in 7.4.
J.. 13a.. Perfornt the reduction of the regular representation ofthe group sP 3
~ith the help of the Young tableaux. The three tableaux will be
labelled H (for horizontal), L (for I~shaped) and V (for vertical).
Show that PQ for the tirst and last is equal to the group itself and
that PQ for the L-tableau consists of four elements.
Show that p :..:: 6 (for H and V) and p = 3 for L.
3" 13'b. lJse the representation reSUlting from the basis t1, e2' 63 as discussed in 7.4 to calculate the ptojection operators corresponding
to the following tableaux:
1

2.

Show that projection operators, when acting on the basis functions

give rise to the following three linear combinations: el +e2+ e3;
"1 - e3; el - e2,' (Note that this new basis will result in exactly thesame representation as the example worked out in 7*4).
4. 1. Wave function for equilateral triangular molecules are constructed
by tanhg linear combinations of "atonlic orbitals"" An atomic
orbital is a wave function 'described in polar coordinates, where the
origin of the coordinate' system is the nucleus of that particular
atom. If spherical harmonics are used for the angular part, rota. .
tions of the molecule "''"in mean that the spheri6al harmonic orbital
attached to a certain corner will go over possibly into another spherical harmonic around another center (i.e. another orbital)or a iinear
combination thereof,.
1

This i'.raDJformation it not necessarily unitary.

271

PROBLEMS

Construct the irreducible linear combination of atomic orbitals

when the orbitals are s-functionso Same if they are p-functionso

5. 1. Show that the quantities X, Y, X', Y', as introduced in 2.2, have

indeed the same transfortnation rules as 11, ~'~ tjt, Le. they satisfy
(5.4) as well.
5. 2. Perfonn the calculation to obtain (5.7a) Hint: Introduce the
following quantity,

e,

(, - itT )* ( ~ -~ i11) =

and show that C is a real constant.. Next show that

Ct
(

fl.)

+, {X*) 2 + (P_.+ 22
. 2

~...

Introduce a formal angle which makes that this equation is auto..

matically satisfied and prove that this angle corresponds to OJ,.
5.. 3. Show that the coupling of 2 ,electrons gives rise to a 1S, a 3p and a
1 D-state, and determine the bases for these three representations.
There are 15 possible antisymmetrized products of two wave
functions (two by two Slater determinants). For instance:

where the numbers refer to the mrvalues of the p-electrons, the

+ signs to their ms-values and the subscripts are the electron labels.
l"'he "standa.rd orderu will be one of decreasing (or actually Don
increasing) Ins..numbers. Construct the 15 basis functions for the
representations mentioned abovc~
Hint: the ML ~ 2; Ms = 0 wave function of the representation. ! D
is unambiguously detemrincd.: (1 +' 1 -). If \\le now operate ,~dth the
oJj~rator Jlq of equatioll (5.48) \vhete J and j are equal to .L -=- 2,
\ve can create the ml = 1; lvfij ~;;::. 0 linear <.~ombination of vva""v'e
functions of the .D representatiou~ 1~11e remaining part of this linear
combination has to belong to the 3.p representation" Further operation \\~ltll Jl p and J.lq acting either on ML or Ms will create all further
basis functions. (This is the so-called Gray and Wills method,
compare CONDON and SHORTLEY [1935] . )

272

PROBLEMS

5. 4. The Wigner coefficients can be considered asa matrix which connects QJtf with qM-",.qrnc. Calculate this matrix for

6. 1. Calculate the reduced matrix elements of a spin operator between

two spin t states.
6. 2. The electric quadruple is an example of
with A, := 2, Jl = O.
Determine the coefficient in front of the reduced matrix element for
this operator.

T;

7. 1. (Compare Ch. 7, 4.) Show that the tetrahedral group or the

permutation group of four elements is the weak direct product of
the following two subgroups. One is the subgroup of permutations
of three elements keeping one fixed. The other is the subgroup of
cyclical permutations of all four elements. Show also that this is
- not a "semi-direct product. (McIntosh.)
7. 2. Determine the inverse lattice of the b.e.c.; r.c.c. and hexagonal
close packed lattices.
7. 3. (Compare Ch. VII, 9.1) Show that all group postulates hold for
the little group.
~~

1. The cubical group corresponds to the permutation group of four

objects. These four QJ)jects. are the body diagonals. The classes can
be found 'by .writing down all possible partitions of four numbers.
Construct the classes with the method of partition indicated in
Chapter 3 and calculate the character Table 8.4.
8. 2.. Show that a J = 2!- level will split into a quartet and a doublet in a
cubical field.
(

SYMBOLS

V2

with . . is associated
transforms as
functional scalar product
(time) average, also: expectation value average over probability density function
direct product of matrices or sets of matrices
addition of representation
small, so-called "virtual", v~on of x.
vector in n-dimensional space
base vector of a vector space
vector space spanned by 11 basis vectors
Kronecker symbol (equals 1 if i == k, and is zero otherwise)
Matrix consisting of elemeDts
Inverse of A that is A-1 A == I
Unit matrix I !II (6 11)
Unitary matrix used for similarity transformation
Metric tensor
Complex conjupte vector, matrix
transposed vector, niatrix
Hermitian conjupte matrix (i.e., complex conjugate and
transposed)
Operator
Adjoint operator
Domain of intearation
Fourier component or generalized Fourier component
Dirac delta function
02
a2 02
"Del-squared' - + - + -

velocity of

(m1
1m)

Bra-vector
K.et-vector

frequency

(.)

<>
x
+

bX
%

e,
illS
~fl

A = (ali)
A-I
I

$, U

(g",)
x,A~

... .....
x,A

At
A,a

A
D

p", l't
d(XI-X2)

or

ar

liaht

213

OZ2

274

SYMBOLS

= 2n:v

(J)

k
h, h

-,,;'

angular frequency
wave (number) vector; k = 271/A
. . h of Planck, h = h/2rc
\c

rnOlnentuxu vector

tJi, 4'

wave function
volume element of configuration space
~
energy eigen value, (occasionally also used for electric field
strength)
Hamilton operator
potential energy
an~r DnounentUnl vector or vector-operator
perturbing energy
perturbation parameter
set of unperturbed eigenfunctions
auxiliary parameter in nearly degenerate systems, expressing

d't
E

H
V
L,l
AW
A

n, 1, m
P,

s~ .<.~

E
~,

,9'"
.;Ie
~S!I,.

s4 3
E, A,/B
CI:
c.~

..(A)
X(i)(C)

the splitting in tenns of the strength of the interaction

radial, angular and azimuthal quantum numbers
(abstract) group element
same: unit element
group i.e., set of (abstract) elements that form a group
Permutation group of n-elements (.9' stands for symmetry)
Subgroup or invariant subgroup
Alternating group
Subgroup of //' 3 containing the three-fold rotations
Matrices respresenting the group elements E, A, B
the set of matrices representing the elements of l
character of the Inatrix A
character of class C belonging to the irreducible representation ['i
Component of the vector in class-space or "reduced char_

f-'-

acter~'t

Xi = Y hilg ~(Ci)
number of elelnents in class i

...

1:

<; or \'

C;
e(p.)

hypercolnplex number
i .. th (irreducible) representation

order of the group; i.c., the number of elements

class i of the group
idempotent element projecting the ,u-th irreducible representation

SYMBOLS

275

same (permutation group); irreducible representation corresponding to the partition {a}

matrix representation 8 above
a
dimensionality of the (irreducible) representation Jl is
equal to character of unit element X(P)(E}.
essential idempotent element corresponding to a shape
spherical harmonic
Yi
three-dimensional space group
P}3
D,
(21 + 1)-dimen.sional representation of the space group above
rotation around the x, y or z axis
unitary unimodular group in two dimensions
f1
element of this group
basis functions of tfl2' spin variables
monomials in ~ and " of the degree j
Mx,M"M: matrices representing infinitesimal rotations
Pauli matrices
Sx, S" S.
orbital angular momentwn
Lx, L y , Lf:
same for individual electrons (usually in dimensionless form)
lxi, I:
Magnetic moment around the z-axis
J/z
~or frequency
Lande g-factor
9
J
total angular momentum (in units Ii)
z-component of j
m =1%
P4.J
2j + I-dimensional representation of the rotation group
parity operator
J /
emJ ",.'!
Clebsch-Gordan coefficient (equivalent to Wigner 3j-symbol)
~:
tensor operator
<j, lnlTfj', m') matrix element of a tensor operator
<iIITllj')
reduced matrix element of a tensor operator
Rj'JII or WUlj2JJ3;j'jJl) Racah coefficient, equivalent to:

~d:z!':l
{ 13J]

Wigner 6j-symbol

t 1, t 2, t3}

a, b, c
T

(TIR)

a*, b*, c*

three primitive translations

general (allowed) translation
element of the space group consisting of a translation T and a
rotation R
reciprocal lattice vectors

276
k

uCr)
I

8.
,.
K

.I
'I,
Q,
D

II

SYMBOLS

label of the irreducible representation of the translationgrowp

periodic part of the wave function
reciprocal lattice vector
redUCIble representation of an invariant subgroup resulting from a irreducible representation
of the main group
irreducible representation of the subgroup
time reversal operator
total magnetic moment
nuclear coordinates
normal nuclear coordinates
dielectric displacement
dielectric constant
electric moment (in e.s.u.)

r.

References cited
Abrapm, A. and M.H.L. Pryce, 1951. Proc. Roy. Soc. A 230 169
,
Bethe, H. A., Ann. Phys. (~J 3 (1929) 133 or [Consultants Bureau (EDaJith Translation)
New York]
Birkhoff, 0., 1950, Hydrody1lllmiC8 (Princeton University Press, 1950; Dover Publicatioas.
New York, 1955)
-Born, M. and P. Iordan. 1930. Elemelltare QlIII1ltSII Mecltaalk (Spriaaer, BerUn)
Boucbert. L P., R. Smoluchowski and B. P. Wlpler, Phys. Rev. SO(ll~ 58 Brinkman. H. C . 1956. Applications ofSpi1tO' InWU'iants In Atomic PlqJQ (North Holland
Publishing Co.. Amsterdam)
Burc;khardt, 1. J., 1947, Die Bewegungsgruppen tIer KrUIIIIIo.,.". (BirkhIuser, Basel)
Cattan, E., 1938, ie~ons sur Ia Ii/oris dtts SpiMIUS (Hermann ot Cfe., Peril)
-Canan, E., 1894, Concernin8 tlte 8tructure 01 finite tmd COlltl1ll10118 #tl.flS/orm4t1on. grt1f1J16
(Thais N~y. 1894)
Canan, E .. , 1913, BuB. Soc. Math. de Franco 4 53
Cartan. E., 1914, Journal de Math6matiques 10 149
Cayley, A., 1854. Phil. Mag., vii (4), 40-57
Condon, E. V. and G.JI. Shortly, 1935, The TM01'Jl ofAtomic Spttctrtl (Cambriclle U.mw.
~n'

Press) - -

.-'

Dirac, P. A. M., 19S8, The Principles of Qllan/urn MecIuznlcs,4th ode (Clarenclon .....
Oxford)

Eckart, C., 1930. Revs. Mod. Ph,... Z 30.5

Bdmo_ A. R.., 19'7, Angular Momentum In QUIl1ltum M:llala (Princeton
-Press, $ow Joney)
BIliott,
1.. 1954, tPhys. Rev. " 280
_
Eyring, H~. 1. Walter aad G. Kimball, 1944, Quantum Chemiall7 (Cbapman & Halt, Ltd
Londo,.; Wiloy " Sons, Inc., New York)
Pano, Ui and G~ kacah, 19S9, 1",.dJu:ible Teuorltll Sets (Academic Presa, New York)
Froboqius, G. and I. Schur. 1906, Sitzbor. Akad. Wiss. Berlin. Pbya.Matb. 1Cl., 186
HeiDe,' V.. 1960, Group theory in QlllllftlUlt Mechanics (PorpmOIl Prell, LoDdon)
HeiUer. W ., 1957, The QlltUllum Theory of Radiation, 3rd cd. (Clarendon Press, Odord)
HatTison. W. A., 1960. Phys. Rev. 118 1190
Herman, F., 1958, R.evs. Mod. Phys. 30 1()2
Herrin& C.,. 1937&.. Ph)'s. Rev. 52 361
Honing. C:, 1937b, phyS~ R.ev. 52 365
Herzberg, lG ..~l94SfMolecuJar Spectra and Molecular StnlClwe. 1"""," IIJfd RIJIIItIII
Sputra .,1 PolJltllomic Molecules, D. Van Nostrand, Inc., New York)
Herzfeld, (:. M. and P. H. B. Meijer, 1960, Solid State Physics 12 1
Hilbert. D. and R. Courant, 1930, Metltoden der Mathematl.8cMn Physik, 2nd eel. (Berlin,

Umv..a,

Rl

. ,

Springer)

Jahn, H. A. and E. Teller, 1937, Proc. Roy. Soc. (London) A 161 220
lahn, H. A." 1938, Proc. Roy. Soc. (London) 164 117
IGttel, C., 19;57, Introduction to Solid Sl~e P/tJ1SIC3, 2nd eel. (Wiley, New York)
Korringa, I., 1954, Tecl1n. Rep. (Ohio State University) aDd Solid State Physics (to be

published)

271

278

REFERENCES CITED

Koster, G. F., 1957, Solid State Physics 5 174

Koster, G. F., 1958, Phys. Rev. 109 227
Kramers,~iI. A., 1937, Quantum Mechanics (North-Holland Publishing Co., Amsterdam)
Lee, T. D., 1960, Physics Today 13 (October 1960)
Lighthill, M. J., 1958, Introduction to Fourier analysis and generalized Functions (Cambridge
University Press)
Mayer, J. E. and M. G. Mayer, 1940, Statistical A-fechanics (John Wiley, New York)
McIntosh, H. V _, 1,958a, Symmetry adapted Functions belonging to the cry..'uallagraphlc
lattice groups, R.tA.8., report 58-3
McIntosh, H. V., 1958b, On accidental degeneracy in classical and quantum Mec/tanics,
R.I.A.S., report 58-4.
Meijer, P. H. E., 1954, Phys. Rev. 95 1443
Melvin, M. A., ,1956, Rev. Mod.. Phys. 28 18
Melvin. M. A~, 1960, Rev. Mod. Phys. 31 477
Molenaar, P. G., 1936, Eindlge Substitutie Groepen (P. NQordhotr, N.V., Groningen)
Meller, C., 1952, The Theory "I Relativity (The IJniversity Press, Oxford)
MorSe., P. M. and H. Feshbach, 1953, Methods of Theorehcol physics (McGraw-Hill

Book Co., Inc., New York)

Neumann, .1. v., 1932, Mathematische Grundlagen der Quanten Mechanik (Springer,
Berlin, 1932; Dover Publications, New York, 1943); also! MathematicoiFoundotio1l8 of
Quantum MechaniCS, Transl. from German by R. T~ Beyer, 1955 (New Jersey, Princeton
Univ~rsity Press)
Qpcc,..,.owski, W., 1940. Physica 7 552
PIliUips, P.. (;.; 1956, An IntrodUction to Crystallography (Longmans Green, London)
PltiJIips, J. C., 1956, Phys. Rev. 104 1263 (See errata Pbys. Rev. 105 (1957) J 933)
R1t:ah,;(l.,194~,-PhYS. Rev. 62438; and 63 (1943) 367
Rosenthal, J. E. and G. M. Murphy, 1936, Rev. Mod. Phys. 8 317
~, L, 1949, Quantum Mechanics (McGraw-Hill Book Co., Inc.. New York)
SC~iff, L. t 19S4, Am. J. Phys. '22 621
Seltz, P., 1934, Z . fUr Kristallogr. 88 433
Sl*ter, ~.: C., t 953, Electronic Structure of Solids, Solid State and Mol. Thet;Jry Reports
/No 4'. (M.I.T., Cambridge Mass.)
stater, J. C., .1934, Phys. Rev. 45 794
.,Speiser, A.~ 1937, Theory of Groups of Finite Order, (Dover Publications. New ~"ork)
/ 4th e4 . '~BitkhatlS4(T, '. Basel, 1956)
Sucksmith, W., 1930, Proc . Phys. Soc. 42 385
Van der Waerden, B. L., 1949, Modern Algebra (Frederick lJngar Publishing Co., New
't

"

York)

Von der Lage, F. C. and H. A. Bethe, 1947, Phys. Rev_ 71 612

Webster, A. G., 1955, Partial Differential Equations l~f J.\fathematical Physics, 2nd ed.
(Dover Publications, New York)
WeyI, 'H., '1950, The Theory of Groups and Quantum Afechanics, Translated from 2nd
revised German, 1931 ed. (Dover Publications, New York)
Weyl, H., 1922, Space-Tlme-Matler, Reprint of the 1922 ed. (Dover Publications, New
York)
Weyi, /H., 1925, Mat. Zeitschr. 23 275
Wigner, E. P . 1927, Zeitschr.
Phys. 43 624
Wigner, E. P., 1931. Gruppentheorie und Ihre .Anwendung Quf die Quantenmechanik der
Atomspektren (Vieweg & Sohn, Braunschweig)
Wigner, E. P., 1959, Group Theory and its Applicatiofl to the Quantu;71 A-fechanics 0/
.4lo1llic Spectra (.~cademk Press, Ne~1 \"'ork)

f:

SYSTEMATIC BmUOGRAPHY
1 .MATRICES AND LINEAR VECfOR SPACES
Al.BD.T, A. A., 1956, Fundamentlll Concepts 0/ Higher Algebra (University ofChicqo
Press), (Chapter 'I Groupa, 'Chapter RI< Vector Spacea a.ad Matrices)
BODBWlG, E. t 1956, Matrix Calclllus (North Holland PublisJaial Co., AmJterdam)
HAUfOS, P. B.., 1958, Flnit~DI~l Vector SptlCU (Van Nostrand. PriDcetoa)
MAllGENAU, H. and O. M. MURPHY, 1943, The Malhemotics 01 Phyric6 a1IIl Clttlmbtry
(Van Noatrand,New"Yotk), (Chaplet 10 Matricea'and Chapter IS Group TIMory)
VAN DBa W ADDIlN, B. L. t 1949, Modern A.lgebra (FftJ:derick Uupr, New York)
<

1.1. Fourier Serie,

J., 1958, Introduction to Fourier Analysis (Cambridge University Press)
WIINIR, N., 1933, The FOUl'ier Integral aM Tlain of Ita Applications (Dover Publications,
New York)
LIGHTHILL, M.

2. QUANTUM MECHANICS
",,~'

D., 19'1, Quantum Theory (Prentice Hall, NewYorlC)

Bo&N. M. and. P.J~,l~3Q. ~e (JUlllllell N~ (Spriqer, Berlin)
DIRAC, P. A. M., 1958, Quantum Mecllanics (Oxford University Press)
K.Iw.tBRs, H. A., ~1937, QUtUIIIUII MechalAks (North-HoUaacl P\JhJisbinICo., Amsterdam)
N~ J. V. t 1932, 1943, Mtlt~ GrlUldlagell de, Qlla1ltttllM4citm1k (Sprincer.
Berlin, 1932; Dover Publications, New York, 1943); Also: MathemalietdPOlllldtuiD,. of
Q~ MeclItudcs, Ta.l. from Gorman by R. T. Beyer, 1955 (princoton UDiveQiQ'
Preas, New Jersey. 1955)
SCHIfF. L., 1949, Quantum MecJco.Na (McGraw-Hill Book. Co., Inc., New York)
BOHM,

3. 'GROUPS
.BoERNE&, H., 19S?,D~stellung,nvoll GI'UP~. mit BerllC/uicltligll1J.dRlledII(I,*-tk,
. inoilir,ji"'PJ.ystk (SJ)riDger, Berlin) ( Rtpresentatlon,'oj' GroUps: :w1tll spM:IM CIJ~''''''''
for IM".,Is of Alodern Physics (transla~d froI9 German by P. G. Murphy et 01.)

(North-Holland Publishing Co., Amsterdam, 19(3)

EYRINGt H . 1. WALTER and G. E. KnmALL, 1944, Quantum Chembtr)l (Wiley & SoDa,
New York) (Chapter 10)
,.
HALL. Mit, 1959, The Tlu!ory 01 Qroups (McMillan, New York)
LaUERMAN, W., 19S1,J"trOt:blctlQn to the Tl&eo,yolFinite Groups. 2nd ed. (Methuen, London)
Lm"LBWOOD, D. E. t 19~, Tluf Theory {If Group Characters (Carendon Press, Oxford)
McINTOSH, H. Vet f9S7, Group Theory (RIAS, Inc. Baltimore, Technical report, No S7)
MUltNAOHAN, F. D., 1938, The Theory of Group Rep,esentations (Johns Hopkins Press,
Baltimore)
WEYL, H., 1947. Tlte C/08sica/ Groups (Princeton University Press)
SPEISER, A., 1956, Theory of Groups of Fillite Order 4th ed. (.8irkha.user~ Basel)

279

280

SYSTBMATIC BIBUOGItAPHY

3.1. Tableaux

LUmAtJ, L. D. and B. M. LIPsmTz, 1958. QlIIIIItlIm MItt.utIa (Perp.mon Preas, London)

Ch.9

.. ~

MOl.'INAAJl, P. G., 1930, Bid,. SlIbstitutlt! OrHpe" (Noordboff. GroDinaen)

R1J'I1II1t1'OD, D. a, 1948, 811b8t1tJ111onai AnGly.u (Edlnburah Umv.dty Pna, BcUnburah)
YAMANOUcm. T. t 1937. Proc. PlIJI. Math. ,Soc. JapanS,. 436 (On the conatructioD of
unitary irreducible representations of the symmetric poup)
4. APPLICAnONS OP GR.OUP THEOR.Y TO PHYSICS IN GBNB1lAL

ALTANN, S. L, 1961. Group Theory in Volume 2 of Quantum TheoIY. (0.. A. Batea,

editor) (Acactomlc Pna, New York)
ALTMANl!I. S. L. t in preparatioD. GrotIP Tlleory in p.,.. tIIId A./JplWPIl7,w (Oxford
Uni..,., PrIll)
BIfACMVANTUM, s. and T. VBNKA1'MAYUDU, 19SI,TIICo", ofGrtJ.,.lUItl". ~,pIkatIo" 10
Physical hr1lM- (Waltair, ,1IMlia. ADdhra Uaiwraity)
Bcv.aT, c., 1930, R.evs. Mod. PJvs. 2 305
HBIMB. V 1960. Group T1acor7ln QIMlIItlIm M~C6 (Peqamoa ~ Lo.csoa)
ILuo4DIdSII, M., 1962, Oro", 77reory antllts appllctltio1l8 to pA7IfIIjltobl.in8(AddiaoD
Wesley, ....din. MuL)
HIGMAN, ,a, 19$5.
O,oup-tIIIIoretlc .1Id.MII'" M.".. (aa.actoa Pr-.0aI0Id)
KOS"l'Bllt O. F., 1956, Nota on Group ThSO'7 (Solid-state and Molecular TbeoJ)'TtMDkaJ
Rop No.8. Cambridp, Masa.)
LoMONT, ~ S.,. 1959, ApplicatlDu 0/ FInite GroIqJ8 (AcadelDicPrIlJ., .NewYort)
LyABAUKB, G. YA., 1960, TM .4J1pilctltlon 0/ GI'OIIJJ TIwory In PAy8lcs (Perpmon Pna,

"',plil

London)
MELVIN, M.

A. aDd C. M. HDZRL'D. MW" .A.l"Jwa and Modem Pltylic8 (in prepara-

tion)

MBtVIN, M. A., 1'56, Rev. Mod. Phya. 28 18

VAJtDBR WAIIItDIN, B. L.,1931,m. 011l/l~nt"JNtI8cM M.thode I" der QlUllflfttlleclltmlk
($prlnpr. BerIJn)
V~'f"'YVtJDvJ T., 1953, Application o/Oro", TINory to Pllydcal Pro/Mms (New York
JJnivenity)
W'BY~ H., 1950. 'I'M TMory of 0,.,. tmdQ""'''''' Msclllmla(Translated from. 2nd
,/ revised Gamao. 1931 eel.) (Dover Publications, New York)
. WIGNBa, B .. P., 1931, Gruppe,.,Aeorle IlIId ihre Allwendlmg auf die Qlltlllte1lPMdumlk . ,

~~~f!~""'.(V~ " SoJqa,.Bra~

It. P., 1959, Oro,. TMoI'yantl 118 A.ppllctztIDn to 1M QIIllI'IhUn MecIuud.cI 0/
Atomic S,.CIm .(Academic .~. New York)
WIOND.. E. P., 1'27J Zeitachr. f. Phys. 43 Q4.
WI~_

4.1. Application to Waoe GuIde Junetioru

AULD, B. A. 1952, Appllcatloa of Gl'ol/ll TIwIJ,y ",

tlte Study of Symmetrlc4l WIlY.

t
GlUM J~t101l8 (Stanford Uaivenity. Microwave Laboratory lleport DO. 137)
PJ.'NNBNJIO..o, A. E~, 19S2. Philips Re8earcb Iteports 7 131. 169 and 270

s.

R.OTATIONS IN SPACB

CoHBH, A., 1931, An IntnJductton to tM 1M Thttory of O,..Pal'ameter GNIIIl" (D. C. Hoath

Co N~w York)

SYSTEMATIC BIBLIOGRAPHY

281

E. V. and O. H. SHOR.TLY, 1935. Th8 Theory of Atomic Spectra (C8mbr.lc.tp

University Press)
EDMONDS. A. R.., 1957, Angular Momentum in Quantum Mechanics (princeton Uaivenity

CoNDON,

Press, 'Nflw Jersey)

0/ AngultV AI""..",,,,,,
(Addison-Wesloy. Cambridge, Mass.)
ROSE, M. B., 1951, Elementary Theory of Angular Momentum (WHey & Sons, New York)

FaNBBR.G, E. and G. E. PAU. 1953, Notes on the quantum theory

S.I. Spinors
BlUNDtAN, H. c.., 19S6, ApplicatJolfS of SpilfOl' Invariant' in Atomic Phyric6 (North-Holland
Publishing Co., Amsterdam)
E., 1938, Le~ons sur la thsorle des SpinelUs (Hermann & Cio, Paris)

CARTAN"

6. CLEBSCH-GORDAN AND RACAH COEFFICIENTS

EDMONDS, A. R.., 1957, Angular Momentum in Quantum Mechania (Princeton, University
Preas, Now Jersey)
FANO, U. and o. RACAH, 19S9, Irreducible Te1l8orlal lets (Academic Preas, New York)
ROSE, M. R t 1957, Elementary TMO'Y of Angular Momentum (Wiley & Sons, New York)

6.1. Tables
L. C" 195~ Tables 0/ R.caIa eoeJjkients (Oak Ridge Natienal Laboratel'J,;
Physica Div., ORNL-lSOl, suppl. 1, Febr. 1952)
8JJI>lblIUdlNt L. C, J. M. BUIT and M. B. ROE, 1952, Revs. Mod. Phys. 24249 (T'heIe
tables aive W(abcd; eI) for e :i 2 and abcd/ arbitrary)
CoNDON, E. V. and G. H. SHORTLY, 1935, The Theory 01 Atomic Spectra (Cambtidae
University Press)
BDM9NDS. A. R", 1957, Angular Momentum in Quantum Mechanics (Princeton Uaivenity
~, New Jersey)
ROTlNBEllG. M., R. BIVINS, N. MImlOPOLII and J. K.. WOOI'BN, 1960, The 3-} aDd f,.,)
I~bols (Wiley & SoDS, New York)
BlBDENJIA.llN,

7.1. SPACE OROUPS

BA.Rms, R.. B.t R. R.. BttATI'AIN and F. SElTZ, 1935. Phys. Rev. 41 582
BaLL, D., 1954, Rev. MOd. PhyS. 26 311
.BOU<XABlt.Tt L P., R.. SMOLUCROWSICI and S .. P. WXGNBR., 1936, P1;1ys. Rev.. 50 58
BRILLOUIN, L., 1953, Wave Propagation in Periodic Structure8 (Dover Pubuc.tioas.
New York)
,
BRAVAIS, A., 18S0, On the systems!ormed by poilUS Nlg1llarly dlstri6ated Oil 4Jl_~O!, 11)41
space (Ens- transl. 1949 by tho Crystallography Society of America, Ho\lltoJl, TexaS>
BUllCKHAllDT, J. J . , 1947, Die Bewegungsgruppen der Kr"'tallog'aplt~ (Blrkhl\JS6r, Buef)
DAVYDOV, A. S., 1951, TMory of Molecular ExcilolU (Translated by M. Kasha and
M. Oppenheimer, 1962) (McGraw-Hill, New York)
ELLIOTT, R. J., 19S4, Phys. Rev. 9Ci 280
JOHNSTON, D. F., 1960, Rep. Progress Physics 23 66
10NESt H." 1960, The Theory 0/ Brillouin ZOllftS and Electronic States In Crystau (North..
Holland Publishing Co., Amsterdam)
HERMAN, F~!l J958,. Rev. Mod. Phys. 30 102
lIERRJNG, (~." !937~ Phys. Rev. 51 365

282

SYSTEMATIC BIBLIOGRAPHY

c., 1942, Franklin Institute 233 S25.

Example of representation of space group

for which no point has the symmetry of the pointgroup
K1TrBL, C., 1~S7, Introduction to Solid-state Physics, 2nd cd. (Wiley & Sons, New York)
KOSTBR., O. P., 1957, Solid State Physics 5 174
OvERHAUD&, A. W., 1955, Pbys. Rev. 101 1702
.
PHILUPS, F. C., 1946, An Introduction to Crystallography (Lonamans, London)
SBlTZ, F., 1934, Z. f. Kristallographie, 88 433
SElTZ, F., 1934, Z. f. Kristallographie 90 289
SEITZ, F., 1935, Z. f. Kristallographie 91 336
SEITS, F., 1936, Z. f. Kristallographie 94 100
SElTZ, F., 1936, Annals of Math. 37 17 (On the ~uction of space groups)
VON DB }.,AGE, F. C. and H. A. BETHE, 1947, Phys. Rev. 71 612
ZAMORZAEV, A. M., 1958, Soviet Physics Crystallography 2 10
ZACHARIASEN, W. H., 1945, Theory 01 X-ray D{ffraction in Crystals, (Wiley~ New York)
(section 11.6)
HOllING,

7.2. Tables
Inlernaliolllli Tables for XRay Crystallography 1952 (Eng. Kynoch Pross, Birmingham)

8.1. Finite Groups, Crystal Field Theory

Banm, H. A., 1929, Ann. Phys. [5] 3 133 Translated by The Consultants Bureau,
New Yor~
1laANY, B. and K. W. H. STEVBNS,.19S3 Rep. Progr. in Phys.16 108, continued in: BOWENS
K. D. and J. OWEN, 1955, Rep. Progr. Phys. 18 304; and ORTON J., 1959, Rep. Progr.
Phys. 22 204FiCIC, E. and G. loos, 1957, Handbuch der Physik 28 205
OIUPiITH, J. S., 1961, Theory o/transilion Melal ions (Cambridge University Press)
HBIl~D, C. M. and P. H. B. MEDER, 1961 Solid-state Physics 12 1
~AU, L. D. and E. M. LIPsHITZ, 1958, Quantum Mechanics (pergamon Press, London)
~ 12
McCLUB.E, 1959, Solid State Physics 9 400
~w, W., 1960, Paramagnetic Resonance in Sollds (Academic Press, New York)
MiOPPl1T, W. and C. J. BALLHAUSBN, 1956, Ann. Rev. Phys. Chem. 7 107
OPECHOWSJa, W., 1940, Physiea 7 SS2
ORGEL, :L. E., 1961, An Introduction to Transition Metal Chemistry (Methuen..Wiley
&. Sons)
PaATmut~ 1. L., 1961, Atomic Energy Levels in Crystals (N.B.S. Monograph No. 19, U.S.
Department of Conunorce, National Bureau of Standards, Washington, D.C.)
STBVBNI.K. W. H., 1952, Proc. l'hys. Soc. 65 209
TAVGER, B. A. and V. M. ZArrsEv, 1956, J.E.T.P. 3 430

8.2. Finite Groups, Molecules

EYRlNG,

cH., J. WALTER and O. KIMBALL, 1944, Quantum Chemistry (Wiley & Sons.

New York)
W., 1928, Z. f. Physik 47 835
HEITLER, W. and G. ReMER!! 1931. Z. f. Physik 68 12
MULLIKEN. R. S., 1933, PhY:3, ReTll A~3 279

BEITLER,

SYSTEMA'TIC BIBLIOGRAPHY

283

ROOTHAAN, C. C. J., 1951, Rev. Mod. Phys. 23 69

ROSENTHAL, J. E. and G. M. MURPHY, 1936, Rev. Mod. Phys.8 317
TIsZA, L., 1933, Z. f. Physik 81 48
WlgNER, ~, 1930, On the characteristic elastic ibralions of symmetrical systems, GOttinpl'
Nacbrlchlen 133

8.3. Time Reversal

KuIN, M. J., 1952, Amar. J. Phys. 20 65
HSlUUNG, C., 1937, Phys. Rev. 52 361
WIGNJUl, E., t 932, Math. Phys. KI. 546 Ober die Operation de, Zeltumkehr in de,

Quantell1Mcha"i1c, Nacbr. Akad. der Wiss. GOttingen

8.4. John-Teller Effect

BJ...BANBY, B., K.. D. BowERS and M. H. L. PRYCE, 1955, Proc. R.oy. Soc. (LondonlA 228 166
CLINTON, W. L. and B. RICE, 1959, 1. Chem. Phys. 30 542
DUNITZ, 1. D. and L. B. ORGBL, 1957, J. Chem. Phys. of Solids 3 20
JAHN; H. A., 1938, ProCe Roy. SOc. (London) 164 117
IAHN, H. A. and E. TELLER.. 1937, Proc. Roy. Soc. (London) A 161 220
MoPFI1T, W. and W. THOMSON, 1957, Phys. Rev. 108 1251
OPllt, U. and M. H. L. PRYCE, 1956, Proc. Roy. Soc. (London) 238 425; LoNOUBTHlOGJNS, H. C., U. OPl1t and M. H. L. PItYCE,19S8~ Prac. Roy. Soc. (London) 2441
VAN VLECK, J. H., 1939, J. Chern. Phys. 7 72
WOJTOWICZ, P. J., 1959, Phys. Rev.. 116 32

8.5. Molecular Vibrations

.HJmzao.o, G., 1945, Moiec,!iar Spectra and Molecular StructllTe. Infrared and RaM
Sp,eclra 0/ Po/yalomic Mokc"/~B (D. Van Nostrand, Inc., New York)
MA11mau, .

.J

1945, Spectres de vibration et symm~/r;e .s. molecules el des crute,,1IX

(llermann , Paris)
MA~ F . 1961, Gruppenth~orie der Eigeruchw;ngungen von PlUtktaysteme (Springer,
Berlin)

MVLuICEN, R.. S., 1955, J. Chern. Phys. 23 1997

J. C. DEems and P. C. CRoss, 19S5, Molecular Vibrations (McGraw-Hin.
New York)

WILSON, E. B.,

INDEX
Affine. 1
Analos
-

Buler, 191

Aqularmomontum. 34, 142, 259

-acldition rules of, 153
- quantum number. 142
- operators, 184
Axil (axes)
-

COOl))leto. &JIteID of ortholOnal axes, 94

-fixed.40
-

main, 11, SO
of' quantization, 249
lee also Hoisenberg, Scbr6dinaer

BiUnear form, .. - form

Bindin,
- covalont. 208
- ionic, 208

mot.Uie. -208
Bloch theorem, 217
Bra. He unit, vectors
Bfavail lattic:ca. .. 223
Brillouin zone(s). 54, 213. 219

CoIl 1
- coklventional unit. 206
- w;ilt 213

cu.,8cter(s), 86

~ ).,f ,opat&,

98

-:1 of .the~matrices, 234

- of a representation, 8S
- primitive, 86
- tables, 9S, 97, 98
- thoory of, 99
Clua (c1uaes), 67, 98. 101
- function of, 8-'
- Dumber of, 87

of permutations, 11

- Racah, 193
- structure, 94, 95. 98
- Wiper, 181
Combination
- linear. of atomic. orbitals, 209
Commutative operations, 62
Commutator, 139, 140
Commute, 14
Component8. lee spectral
Conditions
- integrability, 139
- periociicooundary, 211
Confiauration
- space, 34, 35. 110
- unstable, 261
Congruency. 69
Conjugated, transformations
Conjugation, 221
- complex, 2S6
Conventional unit cell, 206
Convergence, 21
Coordinates
- homogeneous, 133
- normal, 261
Correspoft<ience
- one-to-one, 61
Coset, 65, 222
- left, 65
Couplinl
- between levels, 53
- RusseUSaundon, 161
- spin-orbit, 190, 246
Cross over, 54
Cycle (cycles), 70
- notation with, 100
- order of, 70
Darstellung, 74
Decomposition, 76
Degeneiacy, 23. 235
- accidental, 224, 264
_ .. essential, 55, 119, 223
- exchange, 59, 120
--- quasi, 49, 5S
Determinant
- secular, 239

- sum of a, 94, 95
Qaaical waves, lee waves
Oebach.;Qordan
- coefticiont(s), 181, 190
- formul~ 152
- series, 181
Coeftlcient(s), 8ee Fowier
- Qebsch-Gordan. 181, 190

284

285

INDEX
DlasonaUzation, 264
Dirac delta tunction, 24
Divisor
- normal.. 67
Doublet(s)
- Kramers, 2S3, 257
Bffect(s)
- lYlo-maanetic, IS7
- Paschen-Back, 172
- Stark, 167
- Zeeman, 156, 168, 183
Eigonfunctiona, 21, 32
- symmetry adapted, 233
Eipnvalues, 21, 32
- dqone.rate, 1J3
- spectrum of, 22
Eiaenvectora. 21, 175
Bloment(s)
- complex or, 65
- diagonal, 105
- idempotent, 97, 98
- of tho space group, 202
- reduced matrix, 192
- screw, . 202
- unit, 203
Energy splittina, 245
Equation
- Scb.rOdinaer, 33, 110
- secular, 51
~t, 8ft also transformation, " 74
Euclidean, 7
Euler, 1M anal.

.-

square iDtearablo, 17

wave, 291f, 209

Golden-rulo, 47
Group
- Abelian, 62, 121, 167, 204, 207
-

abstract, 241

additive, 62
additive submitted to a system of 0perators, 63
ackUtivo'with IIlwtipIlcaton, 63
alaebra. 88
continuous, 138
crystalliDe point, 198
cubical, 234
cyclic, 64
double, 241
factor, 68, 222
full rotatio~ 198
homomorphic, 72
permutation, 100; 110
point. 199, 228
postulates, 61
rotation(I), 110, 133, 142
space, 88, 92, 206
symmetric. symmetry, 60, lOS
table, 63

Pactor
- Land6, 6-, 149, 171, 249
- phase, 38
Field
- electric, 254
- mapetic, 254
Form
- bilinoar, 8, 21, 258
- box, 99, 115
- diaaonal, 11
- Hermitian. 80
- quadratic, 9
Fourier
- coefDcients, 18
- expansion, 41
- series, 24
Function(s)
- anaular wave. 240

Bloch, 218
free electron wave, 218
fundamental basis, 17
of a claaa, 86
ortho,onal, 92

three-cIimeDsional poin~ 229

unitary UDear, 145
lee also little aroup

Hamiltonian
- spin-, 248
Harmomc(a)
- spherical, 184, 190, 239
Hoisenberl
- axes, 42

- matrices, 43
- representation, 42
Hermitian. 9, 40
- fol'lD, 80
- matrix, 11
- operator, 21. 22, 37, 40
Hilbert apace, 41
- rays of, sell ray
Holohedra1 isomorphism, 72
Homomorphic

286

INDEX

- groups, 72
- matrices, 97
Hund9 s rules, 246
Idempotent (s), 102
- elements, 97
- essential, 103
- induced.. by, 105
Identity, 60, 61
Intensity, 30
Interaction(s)
- strong, 166
- weak, 166
Invariant(s), 112
- operator, 251
Inverso(a),inW1'lioa. 60, 62, 64, 164
Irreducible representations, see lOPresen..
tations
- system of, 83
Isomorphic, 72
- connection, 74
- matrices, 97
Isomorphism, 83, 84
- holohedral, 72
- merohedral, 72
Ket, see also unit, vectors
- projected, 10
Kra11leIS doublets, 253, 257
Land6, see factor
Laporte, rule of, 165
Latti~(s)
- lJravais, 223
--,Inverse or reciprocal, 213, 218'
utw, associative, 62, 204
Legendre, see polynomials
Length, 6
- of a vector, 28, 80
Linearly independent, 1
Little group
- of the first kind, 222
- of the se.cond kind, 221
Mappings, 75, 82, 88
Matrjx (~trices)
- adjo"int, 8
- diagonal, 99
- Heisenberg, 43
-. !-fermitian.. 1 J, 144
.- hcilllomorphic, 97
~ inverse, 4

isomorphic, 97
mapping, 2
mechanics, 15
Pauli, 144
projection, 64
rectangular, 82
- secular, 50, 264
- step-wise, 14, 85
- sum-of-a-class, 9S
- transformation, see 1lI80 transforma...
tion, 1
- unit, 2
- unitary, 11, 20
Metric, 6
Moment
- magnetic, 257
Monomials, 173, 178
MUltiplet(s), 161, 250
Multiplication rule, 3
Multiplicators, see group, additive
Non-holonomic, 140
Norm, 6, 16
- of a vector, 28
Notation
- with cycles, 100
Numbers
- hypercomplex, 88, 103, lOS, 106
- rational, 62
Observable. 32, 38
- physical, 40
Octahedron, 201
Operations
- from the right to the left, 71
- the P.Q-, 162
Operator(s), 3, 61
- Hermitian, see Hermitian
- irreducible tensor, 184
- linear, 19, 3.7
- nuclear spin, 252
- projection.. 94, 97, 99, 104, 237
- time reversal, 257
- see also angular momentum
Orbit, 221
Orbital motion
- "frozen inu, 246
Ctrders 62
- of a cycle, 70
Orthogonal, see vectors
Orthogonality relations, 89, 91
Orthonormal, 17

287

INDEX
Orthononnality
- of character systems, 100

Parameter kroup

unitary

three-, 131

Parity, 164
Parseval, formula of, 19
Partitio numerorum, 101
Paschen-Back etrect, 172
Pauli principle, 106
Permutation (permutations), 59, 221
- class of, 71
-

cyclical, 101
even or odd, 71
group, 100

- product of two, 60
Perturbation theory, 47
Polynomials
- independent, 18S
- orthogonal, 18S
- Legendre, 185
Primitiv~ see character
Principle, complementarity, 31
Probability, 39
Product, 203
- antisymmetrical, 189
- direct, 151
- of two groups, 206
- scalar, 7, 16, 28
- semidirect, 207, 242
- weak direct, 207
Projection, 88
- even odd, 102
- operators, 97
- stereographic, 132
Quadratic form, 9
Quantity, physical, 40
Quantum mechanical waves, see waves
Racah coefficient, 193
Ray(s), of the function space, 38
Reduction, comp1ete, 76
Reflection (s), 124, 203
Relation(s), orthogonality, 89, 91
- commutation, 186
- compatibility, 225
Representation, 74
-- double valued, 137
- equivalent, 74, 85
- faithful, 74, 137
- identity, 74

irreducible, 7S,

77~

89,

91~

96. 98. 114,

126, 152, 173, 174, 190, 208, 220, 221,

264

multiple occuning irreducible, 117'

of rotation group, 131
of the group t', 113
unitary, 80

alternati11l, 89

characters of, 8S, 98

conjugate, 221
interaction. 247
irreducible reprosontation project-

ing out ot a roduciblo, 98

multiple 0CCU1"Iing, 187

non-equivalent irreducible, 87
of the first kind, 2S8
of the second kind, 252, 258
of the third kind, 2S8
product, 193
reducible, 234
- - regular, 88, 89, 96, 104
Rotations, 203
- infinitesimal, 137, 174
- plane, 122
Rule(s)
- golden-, 47
- Hund's 246
- triangular, 240
Russell-Saunders coupling, 161

Scalar product, see product

SchrOdinger equation, 33, ,110
- axes, 40, 42
Schur's Lemma, 82, go, 115
Secular equation, 14, SI
- problem, 262
Seitz, see Wigner
Shape, 101
Sommerfeld, 130
Space(s)
- bra and ket, 9
- class, 93, 94
-

configuration, 136

function, 15
group, 92

product, 159

- representation, 74
- spinor, 136
- unitary, 7
Spectral components, 30
Spectrum, see eigenvalues
Spmor(s), 136, 158, 241

288

INDEX

-:- spiD colPlated. 257

Sp,inor,,~ 6H 'spaCe
Spin apaCe" J58. 236

Spur. J5. :~'.

Star, 222 ..

Stark effect, 167

State, 32
- stationary, 33. 45
- of the system, 3 f~ 38
Subaroup, 64, 119, 204, 218, 220
- Abelian? 147, lOS': '
~,,~d~ of, 66 ,
- invada.nt, 67, 68, lOS, 206, 242

- normal, 220,' ~
Subspaee", 14" ,111
-

factor, 76

irreducible, 170

Substitutions
- linear, 140
Sum. or sq~J 23
Symbol
- 3-J, 181
Symmotry __
128

SyatcIn

complete system of orthoJODllues, 94

term, 119

Ta~les, charactoF, 95, 97

Teqaor. 136, l~
.
- ,Of the ~nd rank.; l.51
_;"
-/lymmetrlC or tho SOCODd rank, 2

Theorem

Blocb. .. 217

.- Wianer, 114, 128, 142

~,. Wlp.or-Eckhart, 168
Time avorap, 45
Time reversal. 256
- .6 operator
Trace,

l'

Tnu-.ladon(s)
- primitive)" 210.

putC, 2ll
TraDJ~ inverse,

Trl.DlP6iitionll 70
Transformation
-

C&ll()Dical, S

corUuaa~ 67

oquivalent, 67

from the teft, 65

induced, 111

Uniqueness, theorem, 81

Unit
- bra, 10
- element, 60
- ket, 10
Unitary space, 7
- and unimodular group, 131, 152
- matrix, 11. 20
- representations, 80
- transformation, 8
- see also parametersroup

in~iant,'f 75-

- spherical,

infinitesimal. 35, 137, 138

mattie, 1
of axes. 20
of the space onto itself, 61
orthoaonal, 111
similarity, S, 102, 20S, 221, 251
unitary, 8

vectors, 212
- axial, 164- basis. 1, ISO
- bra-unit, 9
- in tho sroup space, 88
- tet-unit, 9
- null-, 82, 135
- of the two-dimlDlional space, 136
-

orthogonal, 92
orthogonal basis, 28
orthoaonal unit. 7
primitive translation, 202
reduced, 222
reduced wave, 218
unit, 1

wave~

29

Wave(s)
- "cl#sSica1. 29

fu.1lctiOD,

38

- monochromatic, 31
.- packet, 30

._- quantum mec~l, 30

_ow see also vector
2

Wigncr

.- ..Bckha.rt tbeorern., 168

- theorem, 114, 128, 142
Wigner-Seitz, 213
Youns ts.bleau, 100
Zeeman effect, 156, 168. 183