1.

1 Angular momentum operators and states 1

Angular momentum in Quantum Mechanics
Detailed understanding of the structure and interaction of atoms and other three dimensional
objects relies on the quantum mechanical description of angular momentum. In this section
we summarize some essential parts of the quantum theory of angular momentum. The
description is brief and is not intended to be a substitute for a quantum mechanics textbook
which should be consulted for proofs of the results given below
1
1.1 Angular momentum operators and states
We denote the angular momentum operator by

J which has Cartesian components
2

J =

J
x
e
x
+

J
y
e
y
+

J
z
e
z
. Any operator satisfying the commutation relations [

J
i
,

J
j
] = i
ijk

J
k
is an angular momentum. Equivalently any operator satisfying

J

J = i

J is an angular
momentum. A complete set of commuting observables (CSCO) for states carrying angular
momentum is provided by the operators

J
2
,

J
z
. The eigenstates of these operators are labeled
[j, m) where j 0 is the angular momentum and m is the magnetic quantum number which
gives the projection of

J on the quantization axis which we will take to be e
z
. The possible
values of these quantities are j 0, integer or half-integer values only being allowed and
j m j, with successive values of m separated by one. For example for j = 1 we have
m = 1, 0, 1 and for j = 3/2, m = 3/2, 1/2, 1/2, 3/2.
The states are orthogonal, j

, m

[j, m) =
jj

mm
and are complete

j
m=j
[j, m)j, m[ =

I, where the identity operator acts on a subspace with the given value of j. The eigenvalue
relations are

J
2
[j, m) = j(j + 1)
2
[j, m) (1.1a)

J
z
[j, m) = m[j, m). (1.1b)
We characterize the magnitude of the angular momentum by
_

J
2
) =
_
j(j + 1), which is
greater than the value of the angular momentum j.
It is convenient to introduce the raising and lowering operators
3

=
1

2
_

J
x
i

J
y
_
. (1.2)
The square of the angular momentum can be written in terms of these operators as

J
2
=

J
+

J

J
+
+

J
2
z
. These operators raise or lower the value of m according to

[j, m) =
1

2
_
j(j + 1) m(m1)[j, m1). (1.3)
1
Good references for angular momentum theory include M. Weissbluth, Atoms and molecules, student
edition, (Academic, New York, 1978), D. A. Varshalovich, A. N. Moskalev, and V. K. Khersonskii, Quantum
theory of angular momentum, World Scientic, Singapore, 1989). A word of caution: many dierent
conventions are in use regarding minus signs and where to put factors of 2, etc. These notes do not coincide
completely with the notation of Basdevant & Dalibard.
2
Since hats denote quantum operators we use e
x
etc., instead of the more customary x to denote unit
vectors.
3
These denitions are slightly dierent than those in Basdevant & Dalibard in order to be compatible
with the denitions of spherical basis vectors used in Sec. 1.3.
February 4, 2008 M. Saman
2 7 Atomic physics
We also have the commutation relations [

J
2
,

J

] = 0, [

J
z
,

J

] =

J

, and [

J
+
,

J

] =

J
z
.
In the case of integer values of j = l the angular momentum states can be written in a
coordinate representation in terms of the spherical harmonics Y
l,m
(, ) = , [l, m). The
eigenvalue relations then take the form

L
2
Y
l,m
(, ) = l(l + 1)
2
Y
l,m
(, ) (1.4a)

L
z
Y
l,m
(, ) = mY
l,m
(, ) (1.4b)
where

L
2
=
2
_
1
sin

sin

+
1
sin
2

2
_
(1.5a)

L
z
= i

. (1.5b)
The three dimensional Laplacian is related to

L
2
by

2
=
1
r

2
r
2
r
1
r
2

L
2
.
angular
1.1.1 Addition of angular momenta
In order to describe the states of composite systems which are combinations of subsystems
with angular momenta

J
1
,

J
2
, .. it is necessary to understand how the individual momenta
combine to give a total angular momentum

J. For example we could have

J
1
=

L and

J
2
= s
corresponding to orbital and spin degrees of freedom of a particle. Alternatively we could
have

J
1
= s
1
and

J
2
= s
2
corresponding to spin degrees of freedom of two dierent particles.
Two angular momenta

J
1
,

J
2
acting in separate Hilbert spaces c
1
, c
2
can be combined to
give a coupled angular momentum

J =

J
1

I
2
+

I
1

J
2
=

J
1
+

J
2
= (

J
1x
+

J
2x
)e
x
+ (

J
1y
+

J
2y
)e
y
+ (

J
1z
+

J
2z
)e
z
=

J
x
e
x
+

J
y
e
y
+

J
z
e
z
. (1.6)
We assume that the angular momenta

J
1
,

J
2
act in dierent spaces and thus commute with
each other. It is important to understand that Eq. (1.6) applies to operators and does not
imply that the value of the coupled angular momentum j is always equal to j
1
+ j
2
. As we
will see one of the results of angular momentum coupling theory is that j can take on a range
of possible values for xed j
1
, j
2
.
Some insight into states of coupled angular momenta can be obtained with the vector
model of Fig. 1.1. In an uncoupled basis j
1
, m
1
and j
2
, m
2
are good quantum numbers, i.e.
they are constant in time and describe the states of the separate subsystems. The sum of
the angular momentum projections m = m
1
+m
2
is also a good quantum number, although
1.1 Angular momentum operators and states 3
j
1
j
2
m
2
m
1
uncoupled
basis
j
2
m
m
1
coupled
basis
j
j
1
m
2
e
z
e
z
Figure 1.1: Vector model of angular momentum coupling showing uncoupled states (left)
and coupled states (right).
the vector sum [J
1
+J
2
[ is not constant as the two vectors may precess about e
z
at dierent
rates. Interactions between J
1
, J
2
cause them to be coupled and precess about the composite
vector J. In the coupled basis the magnitude of J and its projection along e
z
are constant
so j, m are good quantum numbers. However, in the coupled basis, m
1
, m
2
are time varying
and are no longer good quantum numbers.
To understand rigorously the eect of angular momentum coupling consider rst an
uncoupled basis corresponding to the observables

J
1
,

J
2
. A CSCO is formed by the operators
_

J
2
1
,

J
1z
,

J
2
2
,

J
2z
_
. The states are labelled [j
1
, m
1
; j
2
, m
2
) and the eigenvalue relations are

J
2
1
[j
1
, m
1
; j
2
; m
2
) = j
1
(j
1
+ 1)
2
[j
1
, m
1
; j
2
, m
2
) (1.7a)

J
1z
[j
1
, m
1
; j
2
; m
2
) = m
1
[j
1
, m
1
; j
2
, m
2
) (1.7b)

J
2
2
[j
1
, m
1
; j
2
; m
2
) = j
2
(j
2
+ 1)
2
[j
1
, m
1
; j
2
, m
2
) (1.7c)

J
2z
[j
1
, m
1
; j
2
; m
2
) = m
2
[j
1
, m
1
; j
2
, m
2
). (1.7d)
There are (2j
1
+ 1)(2j
2
+ 1) states in this uncoupled basis.
We can couple the eigenstates to form states [j
1
, j
2
; j, m). In the coupled basis a CSCO
is given by the operators
_

J
2
1
,

J
2
2
,

J
2
,

J
z
_
. The eigenvalue relations in the coupled basis are

J
2
[j
1
, j
2
; j, m) = j(j + 1)
2
[j
1
, j
2
; j, m) (1.8a)

J
z
[j
1
, j
2
; j, m) = m[j
1
, j
2
; j, m) (1.8b)

J
2
1
[j
1
, j
2
; j, m) = j
1
(j
1
+ 1)
2
[j
1
, j
2
; j, m) (1.8c)

J
2
2
[j
1
, j
2
; j, m) = j
2
(j
2
+ 1)
2
[j
1
, j
2
; j, m). (1.8d)
In the coupled basis j takes on the possible values [j
1
j
2
[, [j
1
j
2
[ + 1, ...j
1
+j
2
and j
m j. The total number of states is unchanged since

j
1
+j
2
j=|j
1
j
2
|
2j + 1 = (2j
1
+ 1)(2j
2
+ 1).
An intuitive explanation for the range of allowed values of j follows from vector additions
in the triangle construction shown in Fig. 1.2. Since any two of the angular momenta can
February 4, 2008 M. Saman
4 7 Atomic physics
j
1
j
2
j
Figure 1.2: Triangle condition for addition of angular momenta.
be combined to give the third we require that
j j
1
+ j
2
(1.9a)
j
1
j
2
+ j (1.9b)
j
2
j
1
+ j. (1.9c)
The last two conditions can be written as j j
1
j
2
and j j
2
j
1
which can be combined
to give j [j
1
j
2
[. This together with the rst condition gives [j
1
j
2
[ j j
1
+ j
2
.
The transformation from the uncoupled to the coupled bases is expressed through the
Clebsch-Gordan coecients. They can be written as
[j
1
, j
2
; j, m) =

I[j
1
, j
2
; j, m)
=
_

m
1
,m
2
[j
1
, m
1
; j
2
, m
2
)j
1
, m
1
; j
2
, m
2
[
_
[j
1
, j
2
; j, m)
=

m
1
,m
2
j
1
, m
1
; j
2
, m
2
[j
1
, j
2
; j, m)[j
1
, m
1
; j
2
, m
2
)
=

m
1
,m
2
C
jm
j
1
m
1
j
2
m
2
[j
1
, m
1
; j
2
, m
2
), (1.10)
where
C
jm
j
1
m
1
j
2
m
2
j
1
, m
1
; j
2
, m
2
[j
1
, j
2
; j, m) (1.11)
are the Clebsch-Gordan coecients. These coecients vanish unless
[j
1
j
2
[ j j
1
+ j
2
(1.12a)
m
1
+ m
2
= m. (1.12b)
The rst of these conditions is the triangle condition, and the second condition expresses
the conservation of the angular momentum projection along the quantization axis. The
Clebsch-Gordan coecients can be dened with dierent phase conventions. We will use the
convention that
C
j
1
+j
2
j
1
+j
2
j
1
j
1
j
2
j
2
= j
1
j
1
; j
2
j
2
[j
1
j
2
; j
1
+ j
2
j
1
+ j
2
) = 1.
which results in all the Clebsch-Gordan coecients being real. Some additional properties
and particular values of Clebsch-Gordan coecients are given in Sec. 1.6.
1.2 Scalar and vector operators 5
There are many dierent notations in use for the Clebsch-Gordan coecients. Some
authors prefer to use the Wigner 3j symbols
_
j
1
j
2
j
m
1
m
2
m
_
which have a higher symmetry.
The 3j symbols are related to the Clebsch-Gordan coecients by
_
j
1
j
2
j
m
1
m
2
m
_
=
(1)
j+m+2j
1

2j + 1
C
jm
j
1
m
1
j
2
m
2
. (1.13)
1.2 Scalar and vector operators
Quantum mechanical operators can be grouped according to their properties under rotation.
The simplest case is that of a scalar operator which is invariant under rotations. This is
expressed mathematically by the statement that a scalar operator

O commutes with all
components of the angular momentum

J, i.e. [

O,

J] = 0.
To see that this implies that expectation values are independent of rotation of the operator
consider the matrix elements , j, m[

O[, j

, m

). Here , are additional quantum numbers

which do not depend on the orientation of the states. There are a total of (2j + 1)(2j

+ 1)
matrix elements. We want to show that if j ,= j

or m ,= m

the matrix elements vanish, and

when j = j

and m = m

all 2j + 1 matrix elements are equal.

To prove this consider the commutator [

O,

J
z
] = 0. Thus
0 = , j, m[[

O,

J
z
][, j

, m

) = (m

m), j, m[

O[, j

, m

).
So if m

,= m the matrix element must vanish and if m

= m we dene the quantity O

m
=
, j, m[

O[, j

, m).
We then consider the commutators [

J

,

O] = 0. evaluation of matrix elements of these
commutators leads to the equalities
_
j(j + 1) m(m+ 1)O
m
=
_
j

(j

+ 1) m(m + 1)O
m+1
(1.14a)
_
j(j + 1) m(m + 1)O
m+1
=
_
j

(j

+ 1) m(m + 1)O
m
. (1.14b)
If j ,= j

there is a contradiction unless O

m
= 0 for all m and if j = j

all the O
m
are equal.
Thus for any scalar operator

O we have that
, j, m[

O[, j

, m

) = O
m

jj

mm
.
This is a convenient result since it is enough to calculate a single matrix element and the
value of m can be chosen to simplify the calculation as much as possible. Examples of scalar
operators are r
2
, p
2
,

J
2
.
One might think that any scalar quantity is a scalar operator, but this is not the case.
Consider the Zeeman Hamiltonian

H = B.

H is a scalar, but if it were a scalar operator
levels with dierent values of m would have the same Zeeman shift when a magnetic eld
is applied. This disagrees with experiment. The explanation is that a scalar operator is
invariant under rotations of the operator with respect to the quantization axis. When the
atom is rotated the direction of the dipole moment changes, but the magnetic eld stays
xed which breaks the spherical symmetry.
February 4, 2008 M. Saman
6 7 Atomic physics
The next type of operator to consider is a vector operator which we dene by requiring
that expectation values transform under rotation in the same way as classical vectors do. In
classical physics a vector V = V
x
e
x
+V
y
e
y
+V
z
e
z
transforms to V

= RV = V

x
e
x
+V

y
e
y
+V

z
e
z
where R is a 3 3 orthogonal rotation matrix. In terms of Cartesian components this can
be written as V

i
=

j
R
ij
V
j
.
In quantum mechanics we require correspondingly that the expectation value of each
component of

V transforms as
[

V[)
i
= [

V
i
[) [

R
ij

V
j
[) =

R
ij
[

V
j
[). (1.15)
A quantum mechanical state [) transforms as
[)

D
R
[) and [ [

R
where

D
R
is a rotation operator
4
. Thus
[

V
i
[) [

V
i

D
R
[) (1.16)
and the requirement that the right hand sides of (1.15,1.16) are equal for arbitrary [) results
in

V
i

D
R
=

R
ij

V
j
. (1.17)
This must be true for an arbitrary rotation angle and thus must also apply to an innitesimal
rotation by an angle about the e
z
axis. Then

D
R
= 1 i

J
z

and
R =
_
_
1 0
1 0
0 0 1
_
_
.
Keeping only linear terms in (1.17) leads to
[

V
x
,

J
z
] = i

V
y
, [

V
y
,

J
z
] = i

V
x
, [

V
z
,

J
z
] = 0.
If we repeat the calculation for rotation about the other Cartesian axes we can write the
results as a single commutation relation
[

V
i
,

J
j
] = i
ijk

V
k
. (1.18)
Any operator which satises Eq. (1.18) thus has expectation values which transform under
innitesimal rotations in the same way as ordinary vectors do. Since nite rotations can
be described as a succession of innitesimal rotations the same relationship must hold. We
therefore take Eq. (1.18) as the dening property of a vector operator. Examples of vector
operators are the position r, momentum p, and orbital angular momentum

L of a point
particle.
4
The operator for rotation of a state about axis e
j
by an angle is

D
R
= e

Jj /
, see HW set 5 from
PH448. The case worked out there was for orbital angular momentum but the result generalizes to an
arbitrary angular momentum

J. An arbitrary rotation in three dimensions can be described by successive
application of rotation operators about diferent axes.
1.3 Spherical coordinates 7
1.3 Spherical coordinates
Why do we care whether or not an operator is a vector operator? The usefulness of this
identication appears when we wish to calculate matrix elements of vector operators, which
are greatly simplied by use of the Wigner-Eckart theorem. However, in order to apply this
theorem in a convenient form it is necessary to rst work in a spherical basis instead of the
Cartesian coordinate system we have been using so far.
An arbitrary vector A can be written in terms of Cartesian unit vectors as
A = A
x
e
x
+ A
y
e
y
+ A
z
e
z
.
Alternatively we can use spherical basis vectors e
+
, e
0
, e

which are dened by

e
0
= e
z
, e
1
= e
+
=
1

2
(e
x
+ ie
y
), e
1
= e

=
1

2
(e
x
ie
y
)
e

0
= e
z
, e

1
= e

+
=
1

2
(e
x
ie
y
), e

1
= e

=
1

2
(e
x
+ ie
y
)
We see that e

p
= (1)
p
e
p
and
e

p
e
q
= (1)
p
e
p
e
q
=
pq
.
The inverse transformations are
e
x
=
1

2
(e
1
e
1
), e
y
=
i

2
(e
1
+e
1
), e
z
= e
0
.
The q
th
component of a vector A is by denition A
q
= A e
q
. We wish to write A such
that this denition holds in the spherical basis, and therefore write A =

q
A
q
e
q
, where the
basis vectors e
q
are to be determined. We therefore require
A e
q
=

A
q
e
q
e
q
= A
q
.
This is true provided e
q
e
q
=
q

q
which implies e
q
= e

q
. Thus in the spherical basis we
have
A =

q=1,0,1
A
q
e

q
=

q=1,0,1
(1)
q
A
q
e
q
,
where
A
0
= A
z
, A
1
= A
+
=
1

2
(A
x
+ iA
y
), A
1
= A

=
1

2
(A
x
iA
y
). (1.19)
February 4, 2008 M. Saman
8 7 Atomic physics
The dot product of two vectors is
A B =

j=x,y,z
A
j
B
j
=
_

p=1,0,1
A
p
e

p
_

_

q=1,0,1
B
q
e

q
_
=

p,q=1,0,1
A
p
B
q
e

p
e

q
=

p,q=1,0,1
A
p
B
q
e

p
(1)
q
e
q
=

p,q=1,0,1
A
p
B
q
(1)
q

p,q
=

p=1,0,1
A
p
B
p
(1)
p
. (1.20)
We can express any vector A in terms of the spherical polar angles , as
A
0
(, ) = [A[ cos = [A[
_
4
3
Y
10
(, )
A
1
(, ) = [A[
e

sin

2
= [A[
_
4
3
Y
11
(, )
A
1
(, ) = [A[
e

sin

2
= [A[
_
4
3
Y
11
(, ). (1.21)
Thus
A(, ) =

q=1,0,1
A
q
(, )e

q
= [A[
_
4
3

q
Y
1q
(, )e

q
. (1.22)
Note that the raising and lowering operators for the angular momentum dened in Eq.
(1.2) are just the components of the angular momentum in a spherical basis since

J =

J
x
e
x
+

J
y
e
y
+

J
z
e
z
=

J
+
e

e
+
+

J
0
e
0
with

J

=
1

2
_

J
x
i

J
y
_
.
Finally we note that relation (1.18) which denes a vector operator can be written in a
spherical basis as
[

J
0
,

V
q
] = q

V
q
, (1.23a)
_

,

V
q
_
=
1

2
_
2 q(q 1)

V
q1
, (1.23b)
where q = 1, 0, 1.
1.4 Spherical tensors 9
1.4 Spherical tensors
Tensors can be thought of as higher dimensional generalizations of scalars and vectors. In
classical physics tensors transform according to
T
ijk...
T

ijk...
=

t
... R
ir
R
js
R
kt
... T
rst...
. (1.24)
The number of indices of the tensor T denes its rank. A scalar is a rank 0 tensor, and a
vector is a rank 1 tensor. In quantum mechanics we will mostly be only interested in tensors
of rank 0,1, or 2. A Cartesian tensor of rank 2 can be formed from two vectors U, V giving
what is called a dyadic:
T
ij
= U
i
V
j
. (1.25)
This rank 2 tensor has 33 = 9 components since i = x, y, z and j = x, y, z and it transforms
according to Eq. (1.24) where two rotation matrices appear.
It turns out that it is not convenient to work with Cartesian tensors because they are
reducible. That is to say a Cartesian tensor can be decomposed into objects which each
transform like tensors of dierent ranks. Consider the rank 2 Cartesian tensor dened above.
We can equivalently write it as
T
ij
=
_
U V
3

ij
_
+
_
U
i
V
j
U
j
V
i
2
_
+
_
U
i
V
j
+ U
j
V
i
2

U V
3

ij
_
. (1.26)
It can be readily checked that the components T
ij
dened by Eqs. (1.25) or (1.26) are the
same. However the three terms in parentheses which appear in Eq. (1.26) have very dierent
transformation properties. They are irreducible Cartesian tensors of ranks 0, 1, and 2. We
can write
T
ij
= T
(0)
ij
+ T
(1)
ij
+ T
(2)
ij
(1.27)
where
T
(0)
ij
=
U V
3

ij
(1.28a)
T
(1)
ij
=
U
i
V
j
U
j
V
i
2
(1.28b)
T
(2)
ij
=
U
i
V
j
+ U
j
V
i
2

U V
3

ij
(1.28c)
The rst term T
(0)
ij
is a scalar which is invariant under rotation. The second term T
(1)
ij
is
an antisymmetric tensor with zero trace (sum of the diagonal elements) which can be written
as the vector product
ijk
(U V)
k
/2 and transforms as a vector. The last term T
(2)
ij
is a
symmetric rank 2 tensor with zero trace which transforms as a rank 2 tensor.
The number of independent components of these three terms is 1 (for the scalar), 3
(for the vector), and 5 (for the tensor). The total number of independent components is
unchanged by the above decomposition since
3 3 = 9 = 1 + 3 + 5.
February 4, 2008 M. Saman
10 7 Atomic physics
Note that the numbers on the right hand side of the above equality correspond to the
number of possible states for objects with angular momenta 0, 1, and 2. Indeed by writing
T
ij
in the form of Eq. (1.27) we have decomposed it into irreducible tensors of rank 0, 1, 2,
which are objects that transform like angular momentum states (or spherical harmonics)
with l = 0, 1, 2.
The decomposition (1.27) is the irreducible representation of the Cartesian tensor T
ij
. We
write spherical tensors of rank 0, 1, 2 as T
0
, T
1
, and T
2
. It can be shown that the elements
T
,q
of the spherical tensors of rank are related to the Cartesian components of the rank 2
tensor T
ij
by
T
0,0
=
1
3

i
T
ii
T
1,0
= T
(1)
xy
T
1,1
=
1

2
_
T
(1)
yz
iT
(1)
zx
_
T
2,0
= T
(2)
zz
T
2,1
=
_
2
3
_
T
(2)
zx
iT
(2)
zy
_
T
2,2
=
_
1
6
_
T
(2)
xx
T
(2)
yy
2iT
(2)
xy

.
These relations are specic to the decomposition of a rank 2 tensor. Higher rank Carte-
sian tensors are more complicated to work with since their irreducible representation is not
unique. We dene a spherical tensor of rank as an operator which satises the commutation
relations
_

J
0
,

T
,q
_
= q

T
,q
, (1.29a)
_

,

T
,q
_
=
1

2
_
( + 1) q(q 1)

T
,q1
, (1.29b)
where q . A vector is a rst rank tensor and putting = 1 in this denition
recovers the commutation relations for vector operators given in Eqs. (1.23).
1.5 Matrix elements of spherical tensors and the
Wigner-Eckart theorem
After these preliminaries we come to an important result which enables us to calculate
matrix elements using the Wigner-Eckart theorem. Before stating the general form of this
theorem lets consider the example of calculating matrix elements of the position operator
between angular momentum states. This task is important for example when calculating
dipole matrix elements and Rabi frequencies for optically induced transitions between atomic
states.
1.5 Matrix elements of spherical tensors and the Wigner-Eckart theorem 11
The matrix element of the position operator between angular momentum states is
n

[r[nlm) = n

[r
_
4
3

q
Y
1q
e

q
[nlm)
=
_
4
3

q
e

q
n

[rY
1q
[nlm)
= n

[r[nl)
_
4
3

q
e

q
l

[Y
1q
[lm).
Here n, n

are radial or other quantum numbers specifying degrees of freedom which do not
depend on the angular coordinates. To nd the angular matrix element we use the identity
Y
l
1
m
1
(, )Y
l
2
m
2
(, ) =

lm
_
(2l
1
+ 1)(2l
2
+ 1)
4(2l + 1)
_
1/2
C
l0
l
1
0l
2
0
C
lm
l
1
m
1
l
2
m
2
Y
lm
(, )
where l runs over all values satisfying the triangle inequality ([l
1
l
2
[ l l
1
+ l
2
) and for
each value of l, l m l. Thus
l

[Y
1q
[lm) =
_
d Y

l

m
Y
1q
Y
lm
=
_
d Y

l

_
3(2l + 1)
4(2l

+ 1)
_
1/2
C
l

0
l010
C
l

lm1q
Y
l

_
3(2l + 1)
4(2l

+ 1)
_
1/2
C
l

0
l010
C
l

lm1q
_
d Y

l

m
Y
l

=
_
3(2l + 1)
4(2l

+ 1)
_
1/2
C
l

0
l010
C
l

lm1q
, (1.30)
and we get
n

[r[nlm) = n

[r[nl)
_
2l + 1
2l

+ 1

q
C
l

0
l010
C
l

lm1q
e

q
(1.31)
which is known as the Gaunt formula. Using Eq. (1.42) we can simplify this further to
n

[r[nlm) = (1)
l

l1
2
n

[r[nl)
_
Max(l, l

)
2l

+ 1

q
C
l

lm1q
e

q
. (1.32)
The matrix element of an individual component is nonzero only when l

= l 1 and
m

= m + q. We note that the matrix elements are the product of a factor n

[r[nl) which
is independent of m, m

, q and a geometrical factor which does depend on m, m

, q. Thus,
the dependence on orientation of the matrix elements is always the same, for any radial
dependence of the wavefunctions.
February 4, 2008 M. Saman
12 7 Atomic physics
1.5.1 Wigner-Eckart theorem
The Gaunt formula gives a complete description for the matrix elements of the position
operator between angular momentum states [lm). The result can be generalized to matrix
elements of arbitrary spherical tensors using the Wigner-Eckart theorem. It states that
j

T
q
[jm) =
j

[[

[[j)

2j

+ 1
(1)
2
C
j

jmq
(1.33)
Here

T
q
is the q
th
component of a spherical tensor operator

T

of rank with 2 + 1 com-

ponents q . The tensor components satisfy the angular momentum commutation
relations given in Eqs. (1.29). The symbol j

[[

[[j) is known as a reduced matrix

element which is independent of m, m

. For completeness we note that the Wigner-Eckart

theorem can also be written in terms of 3j symbols as
j

T
q
[jm) = (1)
j

[[

[[j)
_
j

j
m

q m
_
. (1.34)
Using Eq. (1.33) and the properties of the Clebsch-Gordan coecients we can immedi-
ately write down selection rules for transition matrix elements of spherical tensor operators.
For scalar operators ( = 0, q = 0)
j

= j (1.35a)
m

= m, (1.35b)
for vector operators ( = 1, q = 1, 0, 1)
j

= j 1, j, j + 1 (1.36a)
j + j

1 (= ) (1.36b)
m

= m1, m, m+ 1, (1.36c)
and for rank 2 tensor operators
j

= j 2, j 1, j, j + 1, j + 2 (1.37a)
j + j

2 (= ) (1.37b)
m

= m2, m1, m, m + 1, m + 2. (1.37c)

The supplementary conditions (1.36b,1.37b) are due to the triangle inequalities Eqs. (1.9).
The selection rules for the matrix elements of spherical tensor operators have a very simple
interpretation. The matrix element j

T
q
[jm) can only be nonzero if the angular
momentum states [jm) and [q) can be combined to give a state [j

). In other words the

rules for matrix elements are just like the rules for combining angular momenta, i.e. the
spherical tensor of rank is like an angular momentum of value . This close analogy was
pointed out above in connection with Eq. (1.27).
The Wigner-Eckart theorem states that matrix elements can be divided into the product
of a reduced matrix element that has no orientation dependence and an angular term that
1.5 Matrix elements of spherical tensors and the Wigner-Eckart theorem 13
is independent of the norm of the operator. The reduced matrix element can be calculated
most easily by considering particular values of m, m

, q. For example
n

[[

[[nj) = (1)
2
_
2j

+ 1
n

0[

T
0
[nj0)
C
j

0
j00
, (1.38)
provided the denominator does not vanish. Note that the reduced matrix element is not a
usual Dirac braket since in general n

[[

[[nj) ,= nj[[

[[n

.
To make these results more explicit let us consider an example. We wish to calculate
the matrix element of a spherical harmonic Y
1q
between orbital angular momentum states
[l, m), [l

, m

). We need the reduced matrix element which from Eq. (1.38) is

5
l

[[

Y
1
[[l) =

2l

+ 1
l

0[

Y
10
[l0)
C
l

0
l010
.
We then use (1.30) to get
l

0[

Y
10
[l0) =
_
3
4
_
2l + 1
2l

+ 1
_
C
l

0
l010
_
2
so we can write
l

[[

Y
1
[[l) =
_
3
4

2l + 1C
l

0
l010
.
Therefore
l

Y
1q
[lm) =
_
3
4

2l + 1

2l

+ 1
C
l

0
l010
C
l

lm1q
which agrees with Eq. (1.31).
1.5.2 Lande projection theorem
An important special case of the Wigner-Eckart theorem can be used to calculate matrix
elements of vector operators in terms of matrix elements of

J. Consider a vector operator

A.
The Wigner-Eckart theorem implies that jm
2
[

A[jm
1
) = Cjm
2
[

J[jm
1
) where C is a
constant and

J is the angular momentum operator. To evaluate the constant use
jm
2
[

J[jm
2
) =

m
1
jm
2
[

A[jm
1
) jm
1
[

J[jm
2
)
= C

m
1
jm
2
[

J[jm
1
) jm
1
[

J[jm
2
)
= Cjm
2
[

J
2
[jm
2
)
= Cj(j + 1)
2
.
5
We suppress the quantum number n in the braket since it labels a radial dependence and Y
1q
has no
dependence on r.
February 4, 2008 M. Saman
14 7 Atomic physics
Thus
jm
2
[

A[jm
1
) =
jm
2
[

J[jm
2
)
j(j + 1)
2
jm
2
[

J[jm
1
). (1.39)
The result (1.39) is often referred to as the Lande projection theorem. Since the matrix
elements of vector operators in a given subspace corresponding to a particular value of j
are all proportional to each other the matrix elements of an operator

A can be calculated
by projecting

A onto

J, multiplying by the matrix elements of

J, and normalizing by the
expectation value of

J
2
which is

J
2
) = j(j + 1)
2
.
This result can also be understood intuitively using the vector model of angular momen-
tum coupling discussed above. Since j is a good quantum number in the coupled basis the
expectation value of a vector A is given by the projection of A on J times the expectation
value of J, i.e A) (A J)J) A J)J). We then have to normalize by [J[
2
, and since
this is quantum mechanics we replace [J[
2
) by j(j + 1)
2
. Thus

A) =

J
j(j+1)
2

J) which is
Eq. (1.39).
1.5.3 Matrix elements of coupled angular momenta
In multielectron problems as well as when dealing with ne structure and hyperne structure
manifolds it is necessary to calculate matrix elements between states of coupled angular
momenta.
When

J =

L+s coupled states [nls; jm) are linear combinations of states in the uncoupled
basis [nlm
l
; sm
s
). Applying the Wigner-Eckart theorem we have
n

T
q
[nlsjm) =
n

[[

[[nlsj)

2j

+ 1
(1)
2
C
j

jmq
(1.40)
When the tensor operator

T

commutes with s, the matrix element is only nonzero when

s = s

and we can use the result

n

sj

[[

[[nlsj) =
ss
(1)
j+l

+s+
_
(2j + 1)(2j

+ 1)
_
l s j
j

_
n

[[

[[nl). (1.41)
Here the symbol in curly braces is a 6j symbol which describes the coupling of three angular
momenta. This result is true for any uncoupled angular momenta, i.e. we can replace

L by

J
1
and s by

J
2
in the above expression.
The 6j symbol can be written as a sum over Clebsch-Gordan coecients
_
j
1
j
2
j
3
j
4
j
5
j
6
_
=
(1)
j
1
+j
2
+j
4
+j
5
_
(2j
3
+ 1)(2j
6
+ 1)

C
j
5
m
5
j
3
m
3
j
4
m
4
C
j
3
m
3
j
1
m
1
j
2
m
2
C
j
5
m
5
j
1
m
1
j
6
m
6
C
j
6
m
6
j
2
m
2
j
4
m
4
.
The sum is to be taken over all possible values of m
1
, m
2
, m
3
, m
4
, m
6
while m
5
is held xed.
In practice it is most convenient to look up values needed in a table.
1.6 Values of Clebsch-Gordan coecients and symmetry properties 15
1.6 Values of Clebsch-Gordan coecients and symme-
try properties
There exist a number of explicit expressions for the Clebsch-Gordan coecents. An expres-
sion due to Wigner is
C
c
ab
=
,+
(abc)
_
(c + )!(c )!(2c + 1)
(a + )!(a )!(b + )!(b )!
_
1/2

z
(1)
b++z
(c + b + z)!(a + z)!
z!(c a + b z)!(c + z)!(a b + z)!
where
(abc) =

(a + b c)!(a b + c)!(a + b + c)!

(a + b + c + 1)!
and in the summation z assumes all integer values for which the factorial arguments are
nonnegative. The Clebsch-Gordan coecients satisfy the following unitarity relations

m
1
m
2
C
jm
j
1
m
1
j
2
m
2
C
j

j
1
m
1
j
2
m
2
=
jj

mm

jm
C
jm
j
1
m
1
j
2
m
2
C
jm
j
1
m

1
j
2
m

2
=
m
1
m

m
2
m

2
.
There are a large number of symmetry relations involving permutations of indices.Some
of them are:
C
c
ab
= (1)
a+bc
C
c
ba
= (1)
a
_
2c + 1
2b + 1
C
b
ac
= (1)
a
_
2c + 1
2b + 1
C
b
ca
= (1)
b+
_
2c + 1
2a + 1
C
a
cb
= (1)
b+
_
2c + 1
2a + 1
C
a
bc
C
c
ab
= (1)
a+bc
C
c
ab
When one of the momenta is zero:
C
00
ab
= (1)
a

ab

2a + 1
C
c
a00
=
ac

When the third momentum is the maximum possible:

C
a+b+
ab
=
_
(2a)!(2b)!(a + b + + )!(a + b )!
(2a + 2b)!(a + )!(a )!(b + )!(b )!
_
1/2
When all the m

s are zero:
C
a+b0
a0b0
=
(a + b)!
a!b!
_
(2a)!(2b)!
(2a + 2b)!
_
1/2
C
ab0
a0b0
= (1)
b
a!
b!(a b)!
_
(2b)!(2a 2b + 1)!
(2a + 1)!
_
1/2
February 4, 2008 M. Saman
16 7 Atomic physics
In particular when b = 1
C
j10
j010
=

j
max
2j + 1
(1.42)
where j
max
is the larger of j, j 1.
The coecients for the cases when one of the momenta is 1/2 or 1 are often needed in
atomic calculations. The b = 1/2 cases are:
C
a+1/21/2
a1/21/2
=
_
a + 1
2a + 1
C
a1/21/2
a1/21/2
=
_
a
2a + 1
The b = 1 cases are:
C
a+1
a10
=

(a + + 1)(a + 1)
(2a + 1)(a + 1)
, C
a+11
a11
=

(a + 1)(a + 2)
2(2a + 1)(a + 1)
C
a
a10
=

_
a(a + 1)
, C
a1
a11
=

(a + 1)(a )
2a(a + 1)
C
a1
a10
=

(a + )(a )
a(2a + 1)
, C
a11
a11
=

(a 1)(a )
2a(2a + 1)