Chapter 1

Angular Momentum

Understanding the quantum mechanics of angular momentum is fundamental in

theoretical studies of atomic structure and atomic transitions. Atomic energy
levels are classified according to angular momentum and selection rules for ra-
diative transitions between levels are governed by angular-momentum addition
rules. Therefore, in this first chapter, we review angular-momentum commu-
tation relations, angular-momentum eigenstates, and the rules for combining
two angular-momentum eigenstates to find a third. We make use of angular-
momentum diagrams as useful mnemonic aids in practical atomic structure cal-
culations. A more detailed version of much of the material in this chapter can
be found in Edmonds (1974).

1.1 Orbital Angular Momentum - Spherical

Harmonics
Classically, the angular momentum of a particle is the cross product of its po-
sition vector r = (x, y, z) and its momentum vector p = (px , py , pz ):

L = r p.

The quantum mechanical orbital angular momentum operator is defined in the

same way with p replaced by the momentum operator p ih. Thus, the
Cartesian components of L are

h h
Lx = i y z z y , Ly = i z x x z , Lz = hi x y

y x . (1.1)

With the aid of the commutation relations between p and r:

[px , x] = ih, [py , y] = ih, [pz , z] = ih, (1.2)

1
2 CHAPTER 1. ANGULAR MOMENTUM

one easily establishes the following commutation relations for the Cartesian
components of the quantum mechanical angular momentum operator:
Lx Ly Ly Lx = ihLz , Ly Lz Lz Ly = ihLx , Lz Lx Lx Lz = ihLy .
(1.3)
Since the components of L do not commute with each other, it is not possible to
find simultaneous eigenstates of any two of these three operators. The operator
L2 = L2x + L2y + L2z , however, commutes with each component of L. It is, there-
fore, possible to find a simultaneous eigenstate of L2 and any one component of
L. It is conventional to seek eigenstates of L2 and Lz .

1.1.1 Quantum Mechanics of Angular Momentum

Many of the important quantum mechanical properties of the angular momen-
tum operator are consequences of the commutation relations (1.3) alone. To
study these properties, we introduce three abstract operators Jx , Jy , and Jz
satisfying the commutation relations,
Jx Jy Jy Jx = iJz , Jy Jz Jz Jy = iJx , Jz Jx Jx Jz = iJy . (1.4)
The unit of angular momentum in Eq.(1.4) is chosen to be h, so the factor of
h on the right-hand side of Eq.(1.3) does not appear in Eq.(1.4). The sum of
the squares of the three operators J 2 = Jx2 + Jy2 + Jz2 can be shown to commute
with each of the three components. In particular,
[J 2 , Jz ] = 0 . (1.5)
The operators J+ = Jx + iJy and J = Jx iJy also commute with the angular
momentum squared:
[J 2 , J ] = 0 . (1.6)
Moreover, J+ and J satisfy the following commutation relations with Jz :
[Jz , J ] = J . (1.7)
One can express J 2 in terms of J+ , J and Jz through the relations
J2 = J+ J + Jz2 Jz , (1.8)
2
J = J J+ + Jz2 + Jz . (1.9)
We introduce simultaneous eigenstates |, mi of the two commuting opera-
tors J 2 and Jz :
J 2 |, mi = |, mi , (1.10)
Jz |, mi = m |, mi , (1.11)
and we note that the states J |, mi are also eigenstates of J with eigenvalue
2

. Moreover, with the aid of Eq.(1.7), one can establish that J+ |, mi and
J |, mi are eigenstates of Jz with eigenvalues m 1, respectively:
Jz J+ |, mi = (m + 1) J+ |, mi, (1.12)
Jz J |, mi = (m 1) J |, mi. (1.13)
1.1. ORBITAL ANGULAR MOMENTUM - SPHERICAL HARMONICS 3

Since J+ raises the eigenvalue m by one unit, and J lowers it by one unit,
these operators are referred to as raising and lowering operators, respectively.
Furthermore, since Jx2 + Jy2 is a positive definite hermitian operator, it follows
that
m2 .
By repeated application of J to eigenstates of Jz , one can obtain states of ar-
bitrarily small eigenvalue m, violating this bound, unless for some state |, m1 i,
J |, m1 i = 0.
Similarly, repeated application of J+ leads to arbitrarily large values of m, unless
for some state |, m2 i
J+ |, m2 i = 0.
Since m2 is bounded, we infer the existence of the two states |, m1 i and |, m2 i.
Starting from the state |, m1 i and applying the operator J+ repeatedly, one
must eventually reach the state |, m2 i; otherwise the value of m would increase
indefinitely. It follows that
m2 m1 = k, (1.14)
where k 0 is the number of times that J+ must be applied to the state |, m1 i
in order to reach the state |, m2 i. One finds from Eqs.(1.8,1.9) that
|, m1 i = (m21 m1 )|, m1 i,
|, m2 i = (m22 + m2 )|, m2 i,
leading to the identities
= m21 m1 = m22 + m2 , (1.15)
which can be rewritten
(m2 m1 + 1)(m2 + m1 ) = 0. (1.16)
Since the first term on the left of Eq.(1.16) is positive definite, it follows that
m1 = m2 . The upper bound m2 can be rewritten in terms of the integer k in
Eq.(1.14) as
m2 = k/2 = j.
The value of j is either integer or half integer, depending on whether k is even
or odd:
1 3
j = 0, , 1, , .
2 2
It follows from Eq.(1.15) that the eigenvalue of J 2 is
= j(j + 1). (1.17)
The number of possible m eigenvalues for a given value of j is k + 1 = 2j + 1.
The possible values of m are
m = j, j 1, j 2, , j.
4 CHAPTER 1. ANGULAR MOMENTUM

Figure 1.1: Transformation from rectangular to spherical coordinates.

Since J = J+ , it follows that

J+ |, mi = |, m + 1i, J |, m + 1i = |, mi.

Evaluating the expectation of J 2 = J J+ + Jz2 + Jz in the state |, mi, one finds

||2 = j(j + 1) m(m + 1).

Choosing the phase of to be real and positive, leads to the relations

p
J+ |, mi = (j + m + 1)(j m) |, m + 1i, (1.18)
p
J |, mi = (j m + 1)(j + m) |, m 1i. (1.19)

1.1.2 Spherical Coordinates - Spherical Harmonics

Let us apply the general results derived in Section 1.1.1 to the orbital angular
momentum operator L. For this purpose, it is most convenient to transform
Eqs.(1.1) to spherical coordinates (Fig. 1.1):

x = r sin cos , y = r sin sin , z = r cos ,

p
r = x2 + y 2 + z 2 , = arccos z/r, = arctan y/x.

In spherical coordinates, the components of L are

Lx = ih sin + cos cot , (1.20)



Ly = ih cos + sin cot , (1.21)


Lz = ih , (1.22)

1.1. ORBITAL ANGULAR MOMENTUM - SPHERICAL HARMONICS 5

and the square of the angular momentum is

1 1 2
L2 = h2 sin + . (1.23)
sin sin2 2
Combining the equations for Lx and Ly , we obtain the following expressions for
the orbital angular momentum raising and lowering operators:


L = hei + i cot . (1.24)

The simultaneous eigenfunctions of L2 and Lz are called spherical harmonics.
They are designated by Ylm (, ). We decompose Ylm (, ) into a product of a
function of and a function of :
Ylm (, ) = l,m ()m () .
The eigenvalue equation Lz Yl,m (, ) = hmYl,m (, ) leads to the equation
dm ()
i = mm () , (1.25)
d
for m (). The single valued solution to this equation, normalized so that
Z 2
|m ()|2 d = 1 , (1.26)
0

is
1
m () = eim , (1.27)
2
where m is an integer. The eigenvalue equation L2 Yl,m (, ) = h2 l(l +
1)Yl,m (, ) leads to the differential equation

1 d d m2
sin + l(l + 1) l,m () = 0 , (1.28)
sin d d sin2
for the function l,m (). The orbital angular momentum quantum number l
must be an integer since m is an integer.
One can generate solutions to Eq.(1.28) by recurrence, starting with the
solution for m = l and stepping forward in m using the raising operator L+ ,
or starting with the solution for m = l and stepping backward using the lowering
operator L . The function l,l () satisfies the differential equation

d
L l,l ()l () = hl+1 () + l cot l,l () = 0 ,
d

which can be easily solved to give l,l () = c sinl , where c is an arbitrary

constant. Normalizing this solution so that
Z
|l,l ()|2 sin d = 1, (1.29)
0
6 CHAPTER 1. ANGULAR MOMENTUM

one obtains r
1 (2l + 1)!
l,l () = l sinl . (1.30)
2 l! 2
l+m
Applying L+ to Yl,l (, ), leads to the result
s
(1)l+m (2l + 1)(l m)! m dl+m
l,m () = sin sin2l . (1.31)
2l l! 2(l + m)! d cos l+m

For m = 0, this equation reduces to

r
(1)l 2l + 1 dl
l,0 () = l sin2l . (1.32)
2 l! 2 d cos l
This equation may be conveniently written in terms of Legendre polynomials
Pl (cos ) as r
2l + 1
l,0 () = Pl (cos ) . (1.33)
2
Here the Legendre polynomial Pl (x) is defined by Rodrigues formula

1 dl 2
Pl (x) = (x 1)l . (1.34)
2l l! dxl
For m = l, Eq.(1.31) gives
r
(1)l (2l + 1)!
l,l () = l sinl . (1.35)
2 l! 2
Starting with this equation and stepping backward l m times leads to an
alternate expression for l,m ():
s
(1)l (2l + 1)(l + m)! dlm
l,m () = l sinm sin2l . (1.36)
2 l! 2(l m)! d cos lm

Comparing Eq.(1.36) with Eq.(1.31), one finds

l,m () = (1)m l,m () . (1.37)

We can restrict our attention to l,m () with m 0 and use (1.37) to obtain
l,m () for m < 0. For positive values of m, Eq.(1.31) can be written
s
(2l + 1)(l m)! m
l,m () = (1)m Pl (cos ) , (1.38)
2(l + m)!

where Plm (x) is an associated Legendre functions of the first kind, given in
Abramowitz and Stegun (1964, chap. 8), with a different sign convention, defined
by
dm
Plm (x) = (1 x2 )m/2 m Pl (x) . (1.39)
dx
1.2. SPIN ANGULAR MOMENTUM 7

The general orthonormality relations hl, m|l0 , m0 i = ll0 mm0 for angular mo-
mentum eigenstates takes the specific form
Z Z 2

sin dd Yl,m (, )Yl0 ,m0 (, ) = ll0 mm0 , (1.40)
0 0

for spherical harmonics. Comparing Eq.(1.31) and Eq.(1.36) leads to the relation

Yl,m (, ) = (1)m Yl,m

(, ) . (1.41)

The first few spherical harmonics are:

q
1
Y00 = 4
q q
Y10 = 3
4 cos Y1,1 = 8
3
sin ei
q q
Y20 = 5
16 (3 cos2 1) Y2,1 = 8
15
sin cos ei
q
Y2,2 = 3215
sin2 e2i
q q
7
Y30 = 16 cos (5 cos2 3) Y3,1 = 64
21
sin (5 cos2 1) ei
q q
105
Y3,2 = 32 cos sin2 e2i Y3,3 = 64
35
sin3 e3i

1.2 Spin Angular Momentum

1.2.1 Spin 1/2 and Spinors
The internal angular momentum of a particle in quantum mechanics is called
spin angular momentum and designated by S. Cartesian components of S satisfy
angular momentum commutation rules (1.4). The eigenvalue of S 2 is h2 s(s + 1)
and the 2s + 1 eigenvalues of Sz are hm with m = s, s + 1, , s. Let us
consider the case s = 1/2 which describes the spin of the electron. We designate
the eigenstates of S 2 and Sz by two-component vectors , = 1/2:

1 0
1/2 = , 1/2 = . (1.42)
0 1

These two-component spin eigenfunctions are called spinors. The spinors

satisfy the orthonormality relations

= . (1.43)

The eigenvalue equations for S 2 and Sz are

S 2 = 34 h2 , Sz = h .
8 CHAPTER 1. ANGULAR MOMENTUM

We represent the operators S 2 and Sz as 2 2 matrices acting in the space

spanned by :

1 0 1 0
S 2 = 34 h2 , Sz = 12 h .
0 1 0 1

One can use Eqs.(1.18,1.19) to work out the elements of the matrices represent-
ing the spin raising and lowering operators S :

0 1 0 0
S+ = h , S = h .
0 0 1 0

These matrices can be combined to give matrices representing Sx = (S+ +S )/2

and Sy = (S+ S )/2i. The matrices representing the components of S are
commonly written in terms of the Pauli matrices = (x , y , z ), which are
given by

0 1 0 i 1 0
x = , y = , z = , (1.44)
1 0 i 0 0 1

through the relation

1
S= h . (1.45)
2
The Pauli matrices are both hermitian and unitary. Therefore,

x2 = I, y2 = I, z2 = I, (1.46)

where I is the 2 2 identity matrix. Moreover, the Pauli matrices anticommute:

y x = x y , z y = y z , x z = z x . (1.47)

The Pauli matrices also satisfy commutation relations that follow from the gen-
eral angular momentum commutation relations (1.4):

x y y x = 2iz , y z z y = 2ix , z x x z = 2iy . (1.48)

The anticommutation relations (1.47) and commutation relations (1.48) can be

combined to give

x y = iz , y z = ix , z x = iy . (1.49)

From the above equations for the Pauli matrices, one can show

a b = a b + i [a b], (1.50)

for any two vectors a and b.

In subsequent studies we will require simultaneous eigenfunctions of L2 , Lz ,
S 2 , and Sz . These eigenfunctions are given by Ylm (, ) .
1.2. SPIN ANGULAR MOMENTUM 9

1.2.2 Infinitesimal Rotations of Vector Fields

Let us consider a rotation about the z axis by a small angle . Under such a
rotation, the components of a vector r = (x, y, z) are transformed to

x0 = x + y,
y0 = x + y,
z0 = z,

neglecting terms of second and higher order in . The difference (x, y, z) =

(x0 , y 0 , z 0 ) (x, y, z) between the values of a scalar function evaluated in
the rotated and unrotated coordinate systems is (to lowest order in ),


(x, y, z) = x y (x, y, z) = i Lz (x, y, z).
y x

The operator Lz , in the sense of this equation, generates an infinitesimal rotation

about the z axis. Similarly, Lx and Ly generate infinitesimal rotations about the
x and y axes. Generally, an infinitesimal rotation about an axis in the direction
n is generated by L n.
Now, let us consider how a vector function

A(x, y, z) = [Ax (x, y, z), Ay (x, y, z), Az (x, y, z)]

transforms under an infinitesimal rotation. The vector A is attached to a point

in the coordinate system; it rotates with the coordinate axes on transforming
from a coordinate system (x, y, z) to a coordinate system (x0 , y 0 , z 0 ). An in-
finitesimal rotation about the z axis induces the following changes in the
components of A:

Ax = Ax (x0 , y 0 , z 0 ) Ay (x0 , y 0 , z 0 ) Ax (x, y, z)

= i [Lz Ax (x, y, z) iAy (x, y, z)] ,
Ay = Ay (x0 , y 0 , z 0 ) + Ax (x0 , y 0 , z 0 ) Ay (x, y, z)
= i [Lz Ay (x, y, z) + iAy (x, y, z)] ,
Az = Az (x0 , y 0 , z 0 ) Az (x, y, z)
= i Lz Az (x, y, z) .

Let us introduce the 3 3 matrix sz defined by

0 i 0
sz = i 0 0 .
0 0 0

With the aid of this matrix, one can rewrite the equations for A in the form
A(x, y, z) = i Jz A(x, y, z), where Jz = Lz + sz . If we define angular
momentum to be the generator of infinitesimal rotations, then the z component
10 CHAPTER 1. ANGULAR MOMENTUM

of the angular momentum of a vector field is Jz = Lz +sz . Infinitesimal rotations

about the x and y axes are generated by Jx = Lx + sx and Jz = Ly + sy , where

0 0 0 0 0 i
sx = 0 0 i , sy = 0 0 0 .
0 i 0 i 0 0

The matrices s = (sx , sy , sz ) are referred to as the spin matrices. In the following
paragraphs, we show that these matrices are associated with angular momentum
quantum number s = 1.

1.2.3 Spin 1 and Vectors

The eigenstates of S 2 and Sz for particles with spin s = 1 are represented by
three-component vectors , with = 1, 0, 1. The counterparts of the three
Pauli matrices for s = 1 are the 3 3 matrices s = (sx , sy , sz ) introduced in
the previous section. The corresponding spin angular momentum operator is
S = hs where

0 0 0 0 0 i 0 i 0
sx = 0 0 i , sy = 0 0 0 , sz = i 0 0 .
0 i 0 i 0 0 0 0 0
(1.51)
The matrix s2 = s2x + s2y + s2z is

2 0 0
s2 = 0 2 0 . (1.52)
0 0 2
The three matrices sx , sy , and sz satisfy the commutation relations
sx sy sy sx = isz , sy sz sz sy = isx , sz sx sx sz = isy . (1.53)

It follows that S = hs satisfies the angular momentum commutation relations

(1.4).
Eigenfunctions of S 2 and Sz satisfy the matrix equations s2 = 2 and
sz = . The first of these equations is satisfied by an arbitrary three-
component vector. Solutions to the second are found by solving the correspond-
ing 3 3 eigenvalue problem,

0 i 0 a a
i 0 0 b = b . (1.54)
0 0 0 c c
The three eigenvalues of this equation are = 1, 0, 1 and the associated eigen-
vectors are

1 0 1
1 = 12 i , 0 = 0 , 1 = 12 i . (1.55)
0 1 0
1.3. CLEBSCH-GORDAN COEFFICIENTS 11

The phases of the three eigenvectors

are chosen in accordance with Eq.(1.18),
which may be rewritten s+ = 2+1 . The vectors are called spherical
basis vectors. They satisfy the orthogonality relations

= .

It is, of course, possible to expand an arbitrary three-component vector v =

(vx , vy , vz ) in terms of spherical basis vectors:

X
1
v = v , where
=1

v = v .

Using these relations, one may show, for example, that the unit vector r ex-
pressed in the spherical basis is
r
4 X
1
r = Y (, ) . (1.56)
3 =1 1,

1.3 Clebsch-Gordan Coefficients

One common problem encountered in atomic physics calculations is finding
eigenstates of the sum of two angular momenta in terms of products of the
individual angular momentum eigenstates. For example, as mentioned in sec-
tion (1.2.1), the products Yl,m (, ) are eigenstates of L2 , and Lz , as well as
S 2 , and Sz . The question addressed in this section is how to combine product
states such as these to find eigenstates of J 2 and Jz , where J = L + S.
Generally, let us suppose that we have two commuting angular momentum
vectors J1 and J2 . Let |j1 , m1 i be an eigenstate of J12 and J1z with eigenvalues
(in units of h) j1 (j1 + 1), and m1 , respectively. Similarly, let |j2 , m2 i be an
eigenstate of J22 and J2z with eigenvalues j2 (j2 +1) and m2 . We set J = J1 + J2
and attempt to construct eigenstates of J 2 and Jz as linear combinations of the
product states |j1 , m1 i |j2 , m2 i:
X
|j, mi = C(j1 , j2 , j; m1 , m2 , m)|j1 , m1 i|j2 , m2 i . (1.57)
m1 ,m2

The expansion coefficients C(j1 , j2 , j; m1 , m2 , m), called Clebsch-Gordan coeffi-

cients, are discussed in many standard quantum mechanics textbooks (for ex-
ample, Messiah, 1961, chap. 10). One sometimes encounters notation such as
hj1 , m1 , j2 , m2 |j, mi for the Clebsch-Gordan coefficient C(j1 , j2 , j; m1 , m2 , m).
Since Jz = J1z + J2z , it follows from Eq.(1.57) that
X
m|j, mi = (m1 + m2 )C(j1 , j2 , j; m1 , m2 , m)|j1 , m1 i |j2 , m2 i . (1.58)
m1 ,m2
12 CHAPTER 1. ANGULAR MOMENTUM

Since the states |j1 , m1 i |j2 , m2 i are linearly independent, one concludes from
Eq.(1.58) that
(m1 + m2 m)C(j1 , j2 , j; m1 , m2 , m) = 0 . (1.59)
It follows that the only nonvanishing Clebsch-Gordan coefficients are those for
which m1 +m2 = m. The sum in Eq.(1.57) can be expressed, therefore, as a sum
over m2 only, the value of m1 being determined by m1 = mm2 . Consequently,
we rewrite Eq.(1.57) as
X
|j, mi = C(j1 , j2 , j; m m2 , m2 , m)|j1 , m m2 i |j2 , m2 i . (1.60)
m2

If we demand that all of the states in Eq.(1.60) be normalized, then it follows

from the relation
hj 0 , m0 |j, mi = j 0 j m0 m ,
that
X
C(j1 , j2 , j 0 ; m0 m02 , m02 , m0 )C(j1 , j2 , j; m m2 , m2 , m)
m02 ,m2

hj1 , m0 m02 |j1 , m m2 i hj2 , m02 |j2 , m2 i = j 0 j m0 m .

From this equation, one obtains the orthogonality relation:
X
C(j1 , j2 , j 0 ; m1 , m2 , m0 ) C(j1 , j2 , j; m1 , m2 , m) = j 0 j m0 m . (1.61)
m1 ,m2

One can make use of this equation to invert Eq.(1.60). Indeed, one finds
X
|j1 , m m2 i|j2 , m2 i = C(j1 , j2 , j; m m2 , m2 , m)|j, mi . (1.62)
j

From Eq.(1.62), a second orthogonality condition can be deduced:

X
C(j1 , j2 , j; m01 , m02 , m) C(j1 , j2 , j; m1 , m2 , m) = m01 m1 m02 m2 . (1.63)
j,m

The state of largest m is the extended state |j1 , j1 i |j2 , j2 i. With the aid
of the decomposition, J 2 = J12 + J22 + 2J1z J2z + J1+ J2 + J1 J2+ , one may
establish that this state is an eigenstate of J 2 with eigenvalue j = j1 + j2 ; it
is also, obviously, an eigenstate of Jz with eigenvalue m = j1 + j2 . The state
J |j1 , j1 i |j2 , j2 i is also an eigenstate of J 2 with eigenvalue j = j1 + j2 . It is
an eigenstate of Jz but with eigenvalue m = j1 + j2 1. The corresponding
normalized eigenstate is
s
j1
|j1 + j2 , j1 + j2 1i = |j1 , j1 1i|j2 , j2 i
j1 + j2
s
j2
+ |j1 , j1 i|j2 , j2 1i . (1.64)
j1 + j2
1.3. CLEBSCH-GORDAN COEFFICIENTS 13

Table 1.1: C(l, 1/2, j; m ms , ms , m)

ms = 1/2 ms = 1/2
q q
l+m+1/2 lm+1/2
j = l + 1/2 2l+1 2l+1
q q
j = l 1/2 lm+1/2
2l+1
l+m+1/2
2l+1

By repeated application of J to the state |j1 , j1 i|j2 , j2 i, one generates, in this

way, each of the 2j + 1 eigenstates of Jz with eigenvalues m = j1 + j2 , j1 + j2
1, , j1 j2 . The state
s
j2
|j1 + j2 1, j1 + j2 1i = |j1 , j1 1i|j2 , j2 i
j1 + j2
s
j1
+ |j1 , j1 i|j2 , j2 1i , (1.65)
j1 + j2

is an eigenstate of Jz with eigenvalue j1 +j2 1, constructed to be orthogonal to

(1.64). One easily establishes that this state is an eigenstate of J 2 corresponding
to eigenvalue j = j1 + j2 1. By repeated application of J to this state, one
generates the 2j + 1 eigenstates of Jz corresponding to j = j1 + j2 1. We
continue this procedure by constructing the state orthogonal to the two states
|j1 +j2 , j1 +j2 2i and |j1 +j2 1, j1 +j2 2i, and then applying J successively
to generate all possible m states for j = j1 + j2 2. Continuing in this way, we
construct states with j = j1 + j2 , j1 + j2 1, j1 + j2 2, , jmin . The algorithm
terminates when we have exhausted all of the (2j1 + 1)(2j2 + 1) possible linearly
independent states that can be made up from products of |j1 , m1 i and |j2 , m2 i.
The limiting value jmin is determined by the relation

jX
1 +j2

(2j + 1) = (j1 + j2 + 2)(j1 + j2 ) jmin

2
+ 1 = (2j1 + 1)(2j2 + 1) , (1.66)
j=jmin

which leads to the jmin = |j1 j2 |. The possible eigenvalues of J 2 are, therefore,
given by j(j + 1), with j = j1 + j2 , j1 + j2 1, , |j1 j2 |.
Values of the Clebsch-Gordan coefficients can be determined from the con-
struction described above; however, it is often easier to proceed in a slightly
different way. Let us illustrate the alternative for the case J = L + S, with
s = 1/2. In this case, the possible values j are j = l + 1/2 and j = l 1/2.
Eigenstates of J 2 and Jz constructed by the Clebsch-Gordan expansion are also
eigenstates of
= 2L S = 2Lz Sz + L+ S + L S+ .
14 CHAPTER 1. ANGULAR MOMENTUM

Table 1.2: C(l, 1, j; m ms , ms , m)

ms = 1 ms = 0 ms = 1
q q q
(l+m)(l+m+1) (lm+1)(l+m+1) (lm)(lm+1)
j =l+1 (2l+1)(2l+2) (2l+1)(l+1) (2l+1)(2l+2)
q q
j=l (l+m)(lm+1)
2l(l+1)
m (lm)(l+m+1)
2l(l+1)
l(l+1)
q q q
(lm)(lm+1) (lm)(l+m) (l+m+1)(l+m)
j =l1 2l(2l+1) l(2l+1) 2l(2l+1)

The eigenvalues of are = j(j + 1) l(l + 1) 3/4. Thus for j = l + 1/2, we

find = l; for j = l 1/2, we find = l 1. The eigenvalue equation for ,

|j, mi = |j, mi

may be rewritten as a set of two homogeneous equations in two unknowns:

x = C(l, 1/2, j; m 1/2, 1/2, m) and y = C(l, 1/2, j; m + 1/2, 1/2, m):
p
x = (m 1/2) x + (l m + 1/2)(l + m + 1/2) y
p
y = (l m + 1/2)(l + m + 1/2) x (m + 1/2) y .

The solutions to this equation are:

q
l+m+1/2
for = l,
qlm+1/2
y/x = (1.67)
lm+1/2 for = l 1.
l+m+1/2

We normalize these solutions so that x2 + y 2 = 1. The ambiguity in phase is re-

solved by the requirement that y > 0. The resulting Clebsch-Gordan coefficients
are listed in Table 1.1.
This same technique can be applied in the general case. One chooses j1 and
j2 so that j2 < j1 . The eigenvalue equation for reduces to a set of 2j2 + 1
equations for 2j2 + 1 unknowns xk , the Clebsch-Gordan coefficients for fixed j
and m expressed in terms of m2 = j2 + 1 k. The 2j2 + 1 eigenvalues of can
be determined from the 2j2 + 1 possible values of j by = j(j + 1) j1 (j1 +
1)
Pj2 (j2 + 1). One solves the resulting equations, normalizes the solutions
to k x2k = 1 and settles the phase ambiguity by requiring that the Clebsch-
Gordan coefficient for m2 = j2 is positive; e.g., x2j2 +1 > 0. As a second
example of this method, we give in Table 1.2 the Clebsch-Gordan coefficients
for J = L + S, with s = 1.
A general formula for the Clebsch-Gordan coefficients is given in Wigner
(1931). Another equivalent, but more convenient one, was obtained later by
1.3. CLEBSCH-GORDAN COEFFICIENTS 15

Racah (1942):
q
(j1 +j2 j)!(j+j1 j2 )!(j+j2 j1 )!(2j+1)
C(j1 , j2 , j; m1 , m2 , m) = m1 +m2 ,m (j+j1 +j2 +1)!

P (1)k (j1 +m1 )!(j1 m1 )!(j2 +m2 )!(j2 m2 )!(j+m)!(jm)!
k k!(j1 +j2 jk)!(j1 m1 k)!(j2 +m2 k)!(jj2 +m1 +k)!(jj1 m2 +k)! .

With the aid of this formula, the following symmetry relations between Clebsch-
Gordan coefficients (see Rose, 1957, chap. 3) may be established:
C(j1 , j2 , j; m1 , m2 , m) = (1)j1 +j2 j C(j1 , j2 , j; m1 , m2 , m) , (1.69)
j1 +j2 j
C(j2 , j1 , j; m2 , m1 , m) = (1) C(j1 , j2 , j; m1 , m2 , m) , (1.70)
C(j1 , j, j2 ; m1 , m, m2 ) =
s
j1 m1 2j2 + 1
(1) C(j1 , j2 , j; m1 , m2 , m) . (1.71)
2j + 1

Expressions for other permutations of the arguments can be inferred from these
basic three. As an application of these symmetry relations, we combine the
easily derived equation
C(j1 , 0, j; m1 , 0, m) = j1 j m1 m , (1.72)
with Eq.(1.71) to give
(1)j1 m1
C(j1 , j, 0; m1 , m, 0) = j j m1 m . (1.73)
2j + 1 1
Several other useful formulas may also be derived directly from Eq. (1.68):
C(j1 , j2 , j1 + j2 ; m1 , m2 , m1 + m2 ) =
s
(2j1 )!(2j2 )!(j1 + j2 + m1 + m2 )!(j1 + j2 m1 m2 )!
, (1.74)
(2j1 + 2j2 )!(j1 m1 )!(j1 + m1 )!(j2 m2 )!(j2 + m2 )!
C(j1 , j2 , j; j1 , m j1 , m) =
s
(2j + 1)(2j1 )!(j2 j1 + j)!(j1 + j2 m)!(j + m)!
. (1.75)
(j1 + j2 j)!(j1 j2 + j)!(j1 + j2 + j + 1)!(j2 j1 + m)!(j m)!

1.3.1 Three-j symbols

The symmetry relations between the Clebsch-Gordan coefficients are made more
transparent by introducing the Wigner three-j symbols defined by:

j1 j2 j3 (1)j1 j2 m3
= C(j1 , j2 , j3 ; m1 , m2 , m3 ) . (1.76)
m1 m2 m3 2j3 + 1
The three-j symbol vanishes unless
m 1 + m2 + m3 = 0 . (1.77)
16 CHAPTER 1. ANGULAR MOMENTUM

The three-j symbols have a high degree of symmetry under interchange of

columns; they are symmetric under even permutations of the indices (1, 2, 3):

j3 j1 j2 j2 j3 j1 j1 j2 j3
= = , (1.78)
m3 m1 m2 m2 m3 m1 m1 m2 m3
and they change by a phase under odd permutations of (1, 2, 3), e.g.:

j2 j1 j3 j1 j2 j3
= (1)j1 +j2 +j3 . (1.79)
m2 m1 m3 m1 m2 m3
On changing the sign of m1 , m2 and m3 , the three-j symbols transform according
to

j1 j2 j3 j1 j2 j3
= (1)j1 +j2 +j3 . (1.80)
m1 m2 m3 m1 m2 m3
The orthogonality relation (1.61) may be rewritten in terms of three-j sym-
bols as
X j1 j2 j30

j1 j2 j3

1
0 = j 0 j m0 m , (1.81)
m1 m2 m3
m1 ,m2
m1 m2 m3 2j3 + 1 3 3 3 3

and the orthogonality relation (1.63) can be rewritten

X
j1 j2 j3 j1 j2 j3
(2j3 + 1) = m1 m01 m2 m02 . (1.82)
m1 m2 m3 m01 m02 m3
j3 ,m3

We refer to these equations as orthogonality relations for three-j symbols.

The following specific results for three-j symbols are easily obtained from
Eqs. (1.73-1.75) of the previous subsection:

j j 0 (1)jm
= j j m1 m , (1.83)
m m 0 2j + 1 1

j1 j2 j1 + j2
= (1)j1 j2 +m1 +m2
m1 m2 m1 m2
s
(2j1 )!(2j2 )!(j1 + j2 + m1 + m2 )!(j1 + j2 m1 m2 )!
, (1.84)
(2j + 1 + 2j2 + 1)!(j1 m1 )!(j1 + m1 )!(j2 m2 )!(j2 + m2 )!

j1 j2 j3
= (1)j2 +j3 +m3
m1 j1 m3 m3
s
(2j1 )!(j2 j1 + j3 )!(j1 + j2 + m3 )!(j3 m3 )!
.
(j1 + j2 + j3 + 1)!(j1 j2 + j3 )!(j1 + j2 j3 )!(j2 j1 m3 )!(j3 + m3 )!
(1.85)

From the symmetry relation (1.80), it follows that

j1 j2 j3
= 0,
0 0 0
1.3. CLEBSCH-GORDAN COEFFICIENTS 17

unless J = j1 + j2 + j3 is even. In that case, we may write

s
j1 j2 j3 J/2 (J 2j1 )!(J 2j2 )!(J 2j3 )!
= (1)
0 0 0 (J + 1)!
(J/2)!
. (1.86)
(J/2 j1 )!(J/2 j2 )!(J/2 j3 )!

Two maple programs, based on Eq. (1.68), to evaluate Clebsch-Gordan coeffi-

cients (cgc.map) and three-j symbols (threej.map), are provided as part of
the course material.

1.3.2 Irreducible Tensor Operators

A family of 2k + 1 operators Tqk , with q = k, k + 1, , k, satisfying the
commutation relations
[Jz , Tqk ] = qTqk , (1.87)
p
[J , Tqk ] = (k q + 1)(k q) Tq1
k
, (1.88)
with the angular momentum operators Jz and J = Jx iJy , are called ir-
reducible tensor operators of rank k. The spherical harmonics Ylm (, ) are,
according to this definition, irreducible tensor operators of rank l. The opera-
tors J defined by

2 (Jx + iJy ), = +1,
1

J = Jz , = 0, (1.89)
1
2
(Jx iJy ), = 1,

are also irreducible tensor operators; in this case of rank 1.

Matrix elements of irreducible tensor operators between angular momentum
states are evaluated using the Wigner-Eckart theorem (Wigner, 1931; Eckart,
1930):

j1 k j2
hj1 , m1 |Tqk |j2 , m2 i = (1)j1 m1 hj1 ||T k ||j2 i . (1.90)
m1 q m2

In this equation, the quantity hj1 ||T k ||j2 i, called the reduced matrix element of
the tensor operator T k , is independent of the magnetic quantum numbers m1 ,
m2 and q.
To prove the Wigner-Eckart theorem, we note that the matrix elements
hj1 m1 |Tqk |j2 m2 i satisfies the recurrence relations
p
(j1 m1 + 1)(j1 m1 ) hj1 m1 1|Tqk |j2 m2 i =
p
(j2 m2 + 1)(j2 m2 ) hj1 m1 |Tqk |j2 m2 1i
p k
+ (k q + 1)(k q) hj1 m1 |Tq1 |j2 m2 i . (1.91)
18 CHAPTER 1. ANGULAR MOMENTUM

They are, therefore, proportional to the Clebsch-Gordan coefficients

C(j2 , k, j1 ; m2 , q, m1 ), which satisfy precisely the same recurrence relations.
Since

p j1 m1 j1 k j2
C(j2 , k, j1 ; m2 , q, m1 ) = 2j1 + 1 (1) , (1.92)
m1 q m2

the proportionality in Eq.(1.90) is established.

As a first application of the Wigner-Eckart theorem, consider the matrix
element of the irreducible tensor operator J :

j1 m1 j1 1 j2
hj1 , m1 |J |j2 , m2 i = (1) hj1 ||J||j2 i . (1.93)
m1 m2

The reduced matrix element hj1 ||J||j2 i can be determined by evaluating both
sides of Eq.(1.93) in the special case = 0. We find
p
hj1 ||J||j2 i = j1 (j1 + 1)(2j1 + 1) j1 j2 , (1.94)

where we have made use of the fact that

j1 1 j1 m1
= (1)j1 m1 p . (1.95)
m1 0 m1 j1 (j1 + 1)(2j1 + 1)
As a second application, we consider matrix elements of the irreducible tensor
operator r
k 4
Cq = Ykq (, ) ,
2k + 1
between orbital angular momentum eigenstates:

k l1 m1 l1 k l2
hl1 m1 |Cq |l2 m2 i = (1) hl1 ||C k ||l2 i . (1.96)
m1 q m2

The left-hand side of Eq.(1.96) is (up to a factor) the integral of three spherical
harmonics. It follows that
r
X 2k + 1
Ykq ()Yl2 m2 () =
4
l1

l1 m1 l1 k l2
(1) hl1 ||C k ||l2 i Yl1 m1 () , (1.97)
m1 q m2
where we use the symbol to designate the angles and . With the aid of
the orthogonality relation (1.81) for the three-j symbols, we invert Eq.(1.97) to
find
X l1 k l2

Ykq () Yl2 m2 () =
m1 q m2
m2 q
r
2k + 1 (1)l1 m1
hl1 ||C k ||l2 iYl1 m1 () . (1.98)
4 2l1 + 1
1.4. GRAPHICAL REPRESENTATION - BASIC RULES 19

Evaluating both sides of this equation at = 0, we obtain

p l1 k l2
hl1 ||C k ||l2 i = (1)l1 (2l1 + 1)(2l2 + 1) . (1.99)
0 0 0

1.4 Graphical Representation - Basic rules

In subsequent chapters we will be required to carry out sums of products of
three-j symbols over magnetic quantum numbers mj . Such sums can be for-
mulated in terms of a set of graphical rules, that allow one to carry out the
required calculations efficiently. There are several ways of introducing graph-
ical rules for angular momentum summations (Judd, 1963; Jucys et al., 1964;
Varshalovich et al., 1988). Here, we follow those introduced by Lindgren and
Morrison (1985).
The basic graphical element is a line segment labeled at each end by a pair
of angular momentum indices jm. The segment with j1 m1 at one end and j2 m2
at the other end is the graphical representation of j1 j2 m1 m2 ; thus
j1 m1 j2 m2
= j1 j2 m1 m2 . (1.100)

A directed line segment, which is depicted by attaching an arrow to a line

segment, is a second important graphical element. An arrow pointing from
j1 m1 to j2 m2 represents the identity:
j1 m1 j2 m2 j2 m2
- = j1 m1 = (1)j2 m2 j1 j2 m1 m2 . (1.101)

Reversing the direction of the arrow leads to

j1 m1
j2 m2 = (1)j2 +m2 j1 j2 m1 m2 . (1.102)

Connecting together two line segments at ends carrying identical values of

jm is the graphical representation of a sum over the magnetic quantum number
m. Therefore,
X j1 m1 j2 m2 j2 m2 j3 m3 j m j3 m3
= j3 j2 1 1 . (1.103)
m2

It follows that two arrows directed in the same direction give an overall phase,
j1 m1
j2 m2 =
j1 m1
-
- j2 m2
= (1)2j2 j1 j2 m1 m2 , (1.104)

and that two arrows pointing in opposite directions cancel,

j1 m1 j2 m2 j1 m1
-
= - j2 m2 = j1 j2 m1 m2 . (1.105)

Another important graphical element is the three-j symbol, which is repre-

sented as j3 m3

j1 j2 j3 j2 m2
=+ . (1.106)
m1 m2 m3
j1 m1
20 CHAPTER 1. ANGULAR MOMENTUM

The + sign designates that the lines associated with j1 m1 , j2 m2 , and j3 m3 are
oriented in such a way that a counter-clockwise rotation leads from j1 m1 to
j2 m2 to j3 m3 . We use a sign to designate that a clockwise rotation leads
from j1 m1 to j2 m2 to j3 m3 . Thus, we can rewrite Eq.(1.106) as
j1 m1

j1 j2 j3 j2 m2
= . (1.107)
m1 m2 m3
j3 m3

The symmetry relation of Eq.(1.78) is represented by the graphical identity:

j3 m3 j2 m2 j1 m1

+ j2 m2 = + j1 m1 = + j3 m3 . (1.108)
j1 m1 j3 m3 j2 m2

The symmetry relation (1.79) leads to the graphical relation:

j3 m3 j3 m3

j2 m2 = (1)j1 +j2 +j3 + j2 m2 . (1.109)

j1 m1 j1 m1

One can attach directed lines and three-j symbols to form combinations such
as j1 m1

+
6 j3 m3 = (1) j1 m1 j1 j2 j3
. (1.110)
m1 m2 m3
j2 m2

Using this, the Wigner-Eckart theorem can be written

j1 m1

hj1 , m1 |Tqk |j2 , m2 i = 6 kq hj1 ||T k ||j2 i . (1.111)

j2 m2

Furthermore, with this convention, we can write

j1 m1
?
p
C(j1 , j2 , j3 ; m1 , m2 , m3 ) = 2j3 + 1 j3 m3 . (1.112)
?
j2 m2

Factors of 2j + 1 are represented by thickening part of the corresponding line
segment. Thus, we have the following representation for a Clebsch-Gordan
coefficient: j1 m1
?
C(j1 , j2 , j3 ; m1 , m2 , m3 ) = j3 m3 . (1.113)
?
j2 m2
1.5. SPINOR AND VECTOR SPHERICAL HARMONICS 21

The orthogonality relation for three-j symbols (1.81) can be written in graph-
ical terms as
j1 m1 j1 m1
j1
X
j30 m03 j3 m3
def j30 m03 j3 m3 1
+ =
= j j 0 m m0 .
m1 m2
+ 2j3 + 1 3 3 3 3
j2 m2 j2 m2 j2
(1.114)
Another very useful graphical identity is
j2 m2
s
6J 2j3 + 1
+
?= j1 j2 m1 m2 J0 (1.115)
2j1 + 1
j1 m1 j3

1.5 Spinor and Vector Spherical Harmonics

1.5.1 Spherical Spinors
We combine spherical harmonics, which are eigenstates of L2 and Lz , and
spinors, which are eigenstates of S 2 and Sz to form eigenstates of J 2 and Jz ,
referred to as spherical spinors. Spherical spinors are denoted by jlm (, ) and
are defined by the equation
X
jlm (, ) = C(l, 1/2, j; m , , m)Yl,m (, ) . (1.116)

From Table 1.1, we obtain the following explicit formulas for spherical spinors
having the two possible values, j = l 1/2:
q
l+m+1/2
Yl,m1/2 (, )
l+1/2,l,m (, ) = q 2l+1 , (1.117)
lm+1/2
2l+1 Y l,m+1/2 (, )
q
lm+1/2 Yl,m1/2 (, )
l1/2,l,m (, ) = q 2l+1 . (1.118)
l+m+1/2
2l+1 Y l,m+1/2 (, )

Spherical spinors are eigenfunctions of L and, therefore, of the operator

K = 1 L.

The eigenvalue equation for K is

Kjlm (, ) = jlm (, ) , (1.119)

where the (integer) eigenvalues are = l 1 for j = l + 1/2, and = l for

j = l 1/2. These values can be summarized as = (j + 1/2) for j = l 1/2.
The value of determines both j and l. Consequently, the more compact
notation, m jlm can be used.
22 CHAPTER 1. ANGULAR MOMENTUM

Spherical spinors satisfy the orthogonality relations

Z Z 2
sin d d 0 m0 (, )m (, ) = 0 m0 m . (1.120)
0 0

The parity operator P maps r r. In spherical coordinates, the operator

P transforms + and . Under a parity transformation,

P Ylm (, ) = Ylm ( , + ) = (1)l Ylm (, ) . (1.121)

It follows that the spherical spinors are eigenfunctions of P having eigenvalues

p = (1)l . The two spinors m (, ) and m (, ), corresponding to the
same value of j, have values of l differing by one unit and, therefore, have
opposite parity.
It is interesting to examine the behavior of spherical spinors under the op-
erator r, where r = r/r. This operator satisfies the identity

r r = 1 , (1.122)

which follows from the commutation relations for the Pauli matrices. Further-
more, the operator r commutes with J and, therefore, leaves the value of j
unchanged. The parity operation changes the sign of r. Since the value of j
remains unchanged, and since the sign of r changes under the parity transfor-
mation, it follows that

r m (, ) = am (, ) , (1.123)

where a is a constant. Evaluating both sides of Eq.(1.123) in a coordinate

system where = 0, one easily establishes a = 1. Therefore,

r m (, ) = m (, ) . (1.124)

Now, let us consider the operator p. Using Eq.(1.122), it follows that

[r p]
p = r r p = ir ir p . (1.125)
r

In deriving this equation, we have made use of the identity in Eq.(1.50).

From Eq.(1.125), it follows that

df +1
pf (r)m (, ) = i + f m (, ) . (1.126)
dr r

This identities (1.124) and (1.126) are important in the reduction of the central-
field Dirac equation to radial form.
1.5. SPINOR AND VECTOR SPHERICAL HARMONICS 23

1.5.2 Vector Spherical Harmonics

Following the procedure used to construct spherical spinors, one combines spher-
ical harmonics with spherical basis vectors to form vector spherical harmonics
YJLM (, ):
X
YJLM (, ) = C(L, 1, J; M , , M )YLM (, ) . (1.127)

The vector spherical harmonics are eigenfunctions of J 2 and Jz . The eigenvalues

of J 2 are J(J + 1), where J is an integer. For J > 0, there are three correspond-
ing values of L: L = J 1 and L = J. For J = 0, the only possible values of
L are L = 0 and L = 1. Explicit forms for the vector spherical harmonics can
be constructed with the aid of Table 1.2. Vector spherical harmonics satisfy the
orthogonality relations
Z 2 Z
d sin d YJ0 L0 M 0 (, ) YJLM (, ) = J 0 J L0 L M 0 M . (1.128)
0 0

Vector functions, such as the electromagnetic vector potential, can be ex-

panded in terms of vector spherical harmonics. As an example of such an
expansion, let us consider
X
r Ylm (, ) = aJLM YJLM (, ) . (1.129)
JLM

With the aid of the orthogonality relation, this equation can be inverted to give
Z 2 Z

aJLM = d sin d YJLM r Ylm (, ).
0 0

This equation can be rewritten with the aid of (1.56) as

X
aJLM = C(L, 1, J; M , , M ) hl, m|C1 |L, M i . (1.130)

Using the known expression for the matrix element of the C1 tensor operator
from Eqs.(1.96,1.99), one obtains
r
2L + 1
aJLM = C(L, 1, l; 0, 0, 0) Jl M m (1.131)
2l + 1
r r !
l l+1
= Ll1 Ll+1 Jl M m . (1.132)
2l + 1 2l + 1

Therefore, one may write

r r
J J +1
r YJM (, ) = YJJ1M (, ) YJJ+1M (, ) . (1.133)
2J + 1 2J + 1
24 CHAPTER 1. ANGULAR MOMENTUM

This vector is in the direction r and is, therefore, referred to as a longitudinal

vector spherical harmonic. Following the notation of Akhiezer and Berestetsky,
(1)
we introduce YJM (, ) = r YJM (, ). The vector YJJM (, ) is orthogonal to
(1)
YJM (, ), and is,therefore, transverse. The combination
r r
J +1 J
YJJ1M (, ) + YJJ+1M (, ).
2J + 1 2J + 1
(1)
is also orthogonal to YJM (, ) and gives a second transverse spherical vector.
It is easily shown that the three vector spherical harmonics
r r
(1) J J +1
YJM (, ) = YJJ1M (, ) YJJ+1M (, ) (1.134)
2J + 1 2J + 1
(0)
YJM (, ) = YJJM (, ) (1.135)
r r
(1) J +1 J
YJM (, ) = YJJ1M (, ) + YJJ+1M (, ) (1.136)
2J + 1 2J + 1
satisfy the orthonormality relation:
Z
() (0 )
d YJM () YJ 0 M 0 () = JJ 0 M M 0 0 . (1.137)

The following three relations may be also be proven without difficulty:

(1)
YJM (, ) = r YJM (, ), (1.138)
(0) 1
YJM (, ) = p L YJM (, ), (1.139)
J(J + 1)
(1) r
YJM (, ) = p YJM (, ) . (1.140)
J(J + 1)

The first of these is just a definition; we leave the proof of the other two as
exercises.
Bibliography

M. Abramowitz and I. A. Stegun, editors. Handbook of Mathematical Functions.

Applied Mathematics Series 55. U. S. Government Printing Office, Washing-
ton D. C., 1964.

C. Eckart. The application of group theory to the quantum dynamics of

monatomic systems. Rev. Mod. Phys., 2:305380, 1930.

A. R. Edmonds. Angular Momentum in Quantum Mechanics. Princeton Uni-

versity Press, Princeton, New Jersey, 1974.

A. Jucys, Y. Levinson, and V. Vanagas. Mathematical Apparatus of the Theory

of Angular Momentum. Israel Program for Scientific Translations, Jerusalem,
1964.

B. R. Judd. Operator Techniques in Atomic Spectroscopy. McGraw-Hill, New

York, 1963.

I. Lindgren and J. Morrison. Atomic Many-Body Theory. Springer-Verlag,

Berlin, 2nd edition, 1985.

A. Messiah. Quantum Mechanics II. North Holland, Amsterdam, 1961.

G. Racah. Theory of complex spectra. ii. Phys. Rev., 62:438462, 1942.

M. E. Rose. Elementary Theory of Angular Momentum. Wiley, New York, 1957.

D. A. Varshalovich, A.N. Moskalev, and V.K. Khersonski. Quantum Theory of

Angular Momentum. World Scientific, Singapore, 1988.

E. P. Wigner. Gruppentheorie. Friedrich Vieweg und Sohn, Braunschweig, 1931.

25