Supplement #1 to Lecture #27

Angular Momentum Eigenvalues (from lecture notes by Professor Dud-

ley Herschbach)
Consider any Hermitian operator J∼ whose components satisfy the following
commutation rules
[Jx , Jy ] = i}Jz

and the cyclic permutations thereof. Equivalently, the rules may be written
as
J
∼
×J ∼
= i}J∼

or as
X
[J` , Jm ] = i} ε`mn Jn
n

where

ε`mn = +1 if `, m, n are in cyclic order

= −1 if `, m, n are in anti-cyclic order
= 0 if any two of `, m, n are the same.

Seek to find eigenvalues λ for J 2 and µ for Jz such that

J 2 |λµi = λ |λµi
Jz |λµi = µ |λµi .

Since J 2 and Jz are Hermitian, λ and µ are real, and |λµi are the simultaneous
eigenvectors which render J 2 and Jz simultaneously diagonal.
First show λ ≥ µ2
Proof: hλµ |J 2 − Jz2 | λµi = λ − µ2
But
J 2 − Jz2 = Jx2 + Jy2 + �2
z −�
J� 2
J�
z

1
Lecture #27 Supplement #1 Page 27S(1)-2

 2  X
λµ Jx  λµ = hλµ |Jx | λ0 µ0 i hλ0 µ0 |Jx | λµi
λ 0 µ0
| {z }

 † 0 0
λµ Jx  λ µ and Jx† = Jx
X 2
= |hλµ |Jx | λ0 µ0 i| → 0 and similarly for Jy2 term.
λ0 µ 0

So

 2  
λµ Jx + Jy2  λµ = λ − µ2 ≥ 0 Q. E. D.
Since µ2 ≥ 0 this also implies λ ≥ 0.
It is convenient to use the non-Hermitian operators

J± = Jx ± iJy Note J+† = J− , J−† = J+ .

These satisfy

[Jz , J± ] = ±}J± since [Jz , Jx ± iJy ] = i}Jy ± i(−i}Jx )

= }(Jx ± iJy ) = }J± .

Apply this to |λµi and find

(Jz J± − J± Jz ) |λµi = ±}J± |λµi

or

Jz (J± |λµi) = (J± Jz ± }J± ) |λµi

= (µ ± })(J± |λµi) since Jz |λµi = µ |λµi .

Thus J± |λµi is an eigenvector of Jz with eigenvalue µ ± }. Hence J+ “raises”

the eigenvalue of µ to µ + } and J− “lowers” the eigenvalue of µ to µ − }.
Now note
[J 2 , J± ] = 0
since J 2 commutes with its components Jx and Jy . Thus

J 2 (J± |λµi) = J± J 2 |λµi = λ(J± |λµi).

| {z }
λ |λµi

2
Lecture #27 Supplement #1 Page 27S(1)-3

Thus J± |λµi remains an eigenvector of J 2 with the same eigenvalue λ as

|λµi.
By repeated application of J+ we can get eigenvectors with Jz eigenvalues of
µ + }, µ + 2}, . . . but the same eigenvalue λ of J 2 . Since µ2 ≥ λ, for a given
λ there must be some highest value of µ, call it µh , such that J+ |λµh i = 0
rather than generating a new eigenvector of still higher Jz –eigenvalue. Simi-
larly, repeated application of J− gives µ − }, µ − 2}, . . . but would eventually
violate µ2 ≤ λ unless there is some lowest value of µ, call it µ` , such that
J− |λµ` i = 0.
Now we use these conditions to show µn = −µ` . Consider applying J− to
J+ |λµh i = 0. Note the identity:
J− J+ = (Jx − iJy )(Jx + iJy )
= Jx2 + Jy2 + i[Jx , Jy ]
= J 2 − Jz2 − }Jz .
Thus
J− J+ |λµi = (λ − µ2h − }µh ) |λµh i = 0.
Taking the matrix element with hλµh | we find
λ − µ2h − }µh = 0.
Similarly,
J+ J− |λµ` i = (J 2 − Jz2 + }Jz ) |λµ` i
leads to
λ − µ2` + }µ` = 0.
Hence
λ = µh (µh + }) = µ` (µ` − })
| {z }

Two solutions: µh = −µ`

or µh = µ` − } but this second solution
must be rejected since µh was assumed to
be larger than µ` .

3
Lecture #27 Supplement #1 Page 27S(1)-4

Now we can conclude also that µh = µ` + n} where n is some integer. This

follows since, if we start from |λµ` i and apply J+ repeatedly, we obtain the
sequence of eigenvectors:

|λµ` i, J+ |λµ` i, J 2 |λµ` i, . . . J n |λµ` i = |λµn i

| {z } | {z } | + {z } | + {z }
µ` µ` +} µ` +2} µ` +n}=µ}

Thus
µh = −µ` = µ` + n}

or
n n
µ` = − }, µh = + }
2 2
where n = 0, 1, 2, . . . is some integer (related to the value of λ).
For convenience, we write

µ = m}, m = −j, −j + 1, · · · + j

where j = n2 , with j = 0, 21 , 1, 23 , 2, . . .
Then eigenvalues of Jz are −j}, (−j + 1)}, . . . j}
| {z }
2j+1 different values
Eigenvalues of J 2 are given by

λ = µh (µh + }) = µ` (µ` − }) = j}(j} + }) = −j}(−j} − })

λ = j(j + 1)}2

Also, it is convenient to label the eigenvectors by j, m rather than λ, µ, so

J 2 |jmi = j(j + 1)}2 |jmi

jz |jmi = m} |jmi .

Comments
We derived the above eigenvalues using only the commutation property and
the Hermitian property. We find that both integer and half-integer values of
j and m are allowed.

4
Lecture #27 Supplement #1 Page 27S(1)-5

Actually, we have solved a much more general problem than that posed by
the orbital angular momentum of a particle. Thus, for several particles in
the same central force field, the total angular momentum,
X
(n)
L
∼
,
n

also satisfies these relations, even if the particles interact with each other.
Spin angular momenta likewise satisfy these relations.
For orbital angular momentum, L ∼
= q × p must require, in addition, that the
∼ ∼
system returns to its original state under a rotation by 2π . Such a rotation
takes p → p and q → q so q × p → q × p and hence the eigenvectors of L2
∼ ∼ ∼ ∼ ∼ ∼ ∼ ∼
and Lz must be unchanged:

e−i2πJz /} |jmi = e−i2πm |jmi

e−i2πm = +1 if m is integer and hence integer eigenvalues are acceptable for

L2 , Lz . Half-integer values give e−i2πm = −1 and hence are not acceptable
for orbital angular momentum.
Half-integers do apply for spin angular momenta, which are not constructed
from any q ×p and thus can take on both integer and half-integer eigenvalues.
∼ ∼
This illustrates the power of operator derivation. A more general case would
not have been included if we had used wave mechanical methods and repre-
sentations by differential operators.
We have shown that, for a fixed j value,

J+ |jmi = am |j, m + 1i and J− |jmi = bm |j, m − 1i ,

where am and bm are constants, possibly complex numbers. The proportion-

ality constants are simply related to each other, since
⎛ ⎞∗
* +∗ Z
am = hj, m + 1|J+ |jmi = jm |J+† | j, m + 1 = ⎝ ψjm ∗
J− ψj,m+1 dτ ⎠
⎜ ⎟
|{z} | {z }
J− bm+1 ψj,m

= b∗m+1

5
Lecture #27 Supplement #1 Page 27S(1)-6

Now, to evaluate am , consider the identity

J− J+ = J 2 − Jz2 − }Jz .

Apply this to |jmi, then you have

J− J+ |jmi = am J− |j, m + 1i = am bm+1 |jmi = |am |2 |jmi

J 2 − Jz2 − }Jz |jmi = (j(j + 1) − m2 − m)}2 |jmi .

Hence

am = [j(j + 1) − m(m + 1)]1/2 }eiφm

| {z }
(j − m)(j + m + 1) ← another common way of writing it,

where eiφm is an arbitrary phase factor. The usual convention is to take

φ = 0; this fixes the relative phases of the vectors |jmi having different
values of m but the same j.
The only non-vanishing matrix elements of J+ and J− are:

hj, m + 1|J+ |jmi = hj, m|J− |j, m + 1i = [j(j + 1) − m(m + 1)]1/2 }

6
always the lower times the
higher of the two m–values in
Or you can write this alternatively as the matrix element

hj 0 , m0 |J+ |jmi = [j(j + 1) − m(m + 1)]1/2 }δj 0 ,j δm0 ,m+1

hj 0 , m0 |J− |jmi = [j(j + 1) − m(m − 1)]1/2 }δj 0 ,j δm0 ,m−1

6
Lecture #27 Supplement #1 Page 27S(1)-7

List of non-zero elements:

jm|J 2 |jm = j(j + 1)}2 “add the bigger m to j and


subtract the smaller”
hjm|Jz |jmi = m}
z ? }| {
hj, m ± 1|J± |jmi = [j(j + 1) − m(m ± 1)]1/2 } = [(j ± m + 1)(j m)]1/2 }
1 1
Jx = (J+ + J− ), Jy = (J+ − J− )
2 2i
1
hj, m ± 1|Jx |jmi = [j(j + 1) − m(m ± 1)]1/2 }
2
1
hj, m ± 1|Jy |jmi = ± [j(j + 1) − m(m ± 1)]1/2 }
2i

We can summarize elements of Jx , Jy , Jz by:


jm|J|jm
∼
ˆ
= zm}
1
= (x̂ ± iŷ) [j(j + 1) − m(m ± 1)]1/2 }.


j, m ± 1|J|jm
∼ 2
Comment
Thus we have found all matrix elements of J with eigenvectors |jmi of J 2 , Jz .
These eigenvectors and their properties are important, since any time we have
a system of particles isolated in free space, their total angular momentum
J 2 , Jz commutes with the total Hamiltonian, no matter what kind of forces
hold the system together (central or not). That is, the total angular momen-
tum of an isolated system is a constant of the motion in quantum mechanics,
just as in classical mechanics.
Hence it is important to be able to take matrix elements of other operators
in the angular momentum states which characterize an isolated system.

Examples
j = 0 : J+ = (0) J− = (0) Jz = (0) J 2 = (0)

7
Lecture #27 Supplement #1 Page 27S(1)-8

    1  3 
1 0 1 0 0 0 2 0
j = : J+ = J− = Jz = 2 1 J = 4 3
2 0 0 1 0 0 −2 0 4
j = 1: ⎛ √ ⎞ ⎛ ⎞
0 2 √0 1 0 0
J + = ⎝0 0 2⎠ Jz = ⎝0 0 0 ⎠
0 0 0 0 0 −1
⎛ ⎞ ⎛ ⎞
√0 0 0 2 0 0
J− = ⎝ 2 √0 0⎠ J 2 = ⎝0 2 0⎠
0 2 0 0 0 2
j = 32 :
⎛ √ ⎞ ⎛3 ⎞
0 3 0 0 2
0 0 0
⎜0 0 2 √0 ⎟ ⎜0 1 0 0 ⎟
J+ = ⎜ ⎟ Jz = ⎜ 2 ⎟
⎝0 0 0 3⎠ ⎝0 0 − 1 0 ⎠
2
0 0 0 0 0 0 0 − 32
⎛ ⎞ ⎛ 15 ⎞
√0 0 0 0 4
0 0
15
0
⎜ 3 0 0 0⎟ 2 ⎜ 0 4 0
⎟ ⎜ 0⎟
J− = ⎜
⎝ 0 2 ⎠ J = ⎝ 0 0 15
⎟
√0 0 4
0⎠
15
0 0 3 0 0 0 0 4

8
Lecture #27 Supplement #1 Page 27S(1)-9

MOMENTA AS DISPLACEMENT OPERATORS:

Geometrical Meaning of Commutation Rules
Linear Momentum
Let |x1 i be an eigenvector of the position operator X with eigenvalue x1 , i.e.

X |x1 i = x1 |x1 i .

Consider the new state vector defined by e−iapx /} |x1 i; we might ask whether
it is also an eigenvector of X. To find out, evaluate

X e−iapx /} |x1 i = e−iapx /} (X |x1 i) + X, e−iapx /} |x1 i

�
 
| {z } | {z }
x1 |x1 i
6
..
..
..
..
Now ..
..
..
..
..
..
...
...
d � −iap ..
X, e−iapx /} = i} /}
  x .....


e ...
.
dpx .
...
....
.....
.......
..........
.
= ae−iap /} ...
..
..
..
..
..
..
.
.........
....x
..

Thus

X e−iapx /} |x1 i = e−iapx /} (X + a) |x1 i = e−iapx /} (x1 + a) |x1 i

�

= (x1 + a) e−iapx /} |x1 i .

�

Hence e−iapx /} |x1 i is indeed an eigenvector of X with eigenvalue x1 +a instead

of x1 . The unitary operator e−iapx /} formed from the linear momentum oper-
ator px acts as a displacement operator for x position coordinates. Similarly,
py generates displacements of the y coordinate and pz of the z coordinate.
It is a geometrical fact that linear displacements of a point commute. For
example:

9
Lecture #27 Supplement #1 Page 27S(1)-10

xt2 y2
y 6 first move along x
b
x1 y1 t - t by a, then along y
6
a
|x2 y2 i = e−ibpy /} e−iapx /} |x1 y1 i ⇒ - x by b

t a - tx2 y2
y 6
bt reverse the
=e −iapx /} −ibpy /}
e |x1 y1 i ⇒ x1 y1 - x order

The same result is obtained by applying displacements in either order. This

agrees with [px , py ] = 0 (and {px , py } = 0).

Angular momentum
Lx , Ly , Lz operators generate angular displacements or rotations; e.g.,

e−iφLx /}

gives a rotation by angle φ about the x–axis, etc. However, geometrical

rotations about different axes do not commute. For example, z
6
consider a state representing a particle on the z–axis, |z0 i. Now.............................................t
.
...
...
..
−i π2 Ly /} −i π2 Lx /}
e| {z } e |z0 i ⇒ particle on −y axis ?
t -y
| {z }
rotation by π/2 rotation by π/2

about y–axis about x–axis x z

But 6
..
..
..
..
....t
...... .....
....
...
...
...
π π ..
e|−i {z
L /}
e−i 2 Ly /} |z0 i ⇒ particle on +x axis
...
2 x ...
... -y
...
R t
..
} | {z }
rotation by π/2 rotation by π/2

about x–axis about y–axis x

The results of these two rotations taken in opposite order differ by a rotation
about the z–axis. Thus, because the rotations about different axes don’t
commute, we must expect the angular momentum operators, which generate
these rotations, not to commute with each other. Indeed,

[Lx , Ly ] = i}Lz

10
Lecture #27 Supplement #1 Page 27S(1)-11

corresponds to the above example, in which the commutator of rotations

about the x and y axes depends on a z–axis rotation.
Rotational Transformation Properties and Selection Rules
The various observables of a dynamical system can be classified according
to their transformation properties under rotations. This is of great value in
determining the matrix elements of the corresponding operators and, in par-
ticular, leads to selection rules which limit the number of non–zero matrxix
elements.
iφ·J /}
Under action of the rotation operator U = e ∼ ∼ an operator O is trans-
formed according to
0
O
∼
=U
∼∼∼
OU † .
A scalar operator S
∼
is one which is invariant to this transformation (e.g., the
Hamiltonian of an isolated system). Hence, for a scalar operator

U SU † = S
∼∼∼ ∼

or
U
∼∼
S − SU
∼∼
=0 or [U , S ] = 0.
∼ ∼

Thus, a scalar commutes with every rotation operator. Consider, in partic-

ular, an infinitesimal rotation dφ, for which
∼

i
U
∼
= 1 + dφ · J∼ .
} ∼
Since the direction of dφ is arbitrary, S
∼
must commute with each component
∼
of J∼ , or [S,
∼ ∼
J ] = 0. As shown below, this property leads to the selection rules

Δj = 0, Δm = 0

for the non-zero matrix elements of a scalar operator.

A vector operator V∼ is one with three components, Vx , Vy , Vz which transform
under rotations like the coordinates of a point. For an infinitesimal rotation,
   
0 i i
V∼ = 1 + dφ · J∼ V∼ 1 − dφ · J .
} ∼ } ∼ ∼

11
Lecture #27 Supplement #1 Page 27S(1)-12

r 0 is obtained from ∼
Now note that if the position vector ∼ r by rotation through
a small angle dφ about an axis in the direction of the vector dφ, we have, to
∼
first order in dφ,
r0
∼
r + dφ × ∼
=∼ r dφ .........: .......
 .
∼ ∼
 ..
....
 ..
.
...

  - ........ dφ
XXX
XXr
∼X
.
...
...
.
and so XX ...................
0 XX ....
r z
V∼ 0 = V∼ + dφ × V∼ . ∼
∼
Hence, if terms in (dφ)2 are neglected, we obtain
i h   i
dφ × V∼ = dφ · J V − V∼ dφ · J∼ .
∼ } ∼ ∼ ∼ ∼

Since dφ is arbitrary, this relation gives the commutator of V∼ with any com-
∼
ponent of J∼ . Thus, if dφ = εẑ is a rotation about the z–axis, we find
∼

i 

z×V
ε(ˆ ∼
) = ε
Jz V∼ − V∼ Jz
}
or
 
Jz , V∼ = −i}(ˆ
z × V∼ ) or [Jz , Vx ] = −i}(−Vy ) = i}Vy
[Jz , Vy ] = −i}(−Vx ) = −i}Vx
[Jz , Vz ] = 0

etc.
In this way we obtain a set of nine commutation rules:

[Jx , Vx ] = 0 [Jy , Vx ] = −i}Vz [Jz , Vx ] = −i}Vy

[Jx , Vy ] = i}Vz [Jy , Vy ] = 0 [Jz , Vy ] = −i}Vy
[Jx , Vz ] = −i}Vy [Jy , Vz ] = i}Vx [Jz , Vz ] = 0

The selection rules for non-zero matrix elements of a vector operator, i.e. an
r , linear momentum
operator which satisfies the above rules (e.g., position ∼
p, the angular momentum J∼ itself) are shown below to be given by
∼

Δj = 0 and ± 1 for all components of V∼

12
Lecture #27 Supplement #1 Page 27S(1)-13

with

Δm = 0 for Vz
Δm = ±1 for V± = Vx ± iVy .

Scalar Operators, S1
Defined by [S , J ] = 0, for all three components of J∼ . Corollary is [S, J 2 ] = 0
∼ ∼
and [S, Jz ] = 0. If we take the matrix elements, we have

j 0 m0 |[S, J 2 ]|jm = 0



= j 0 m0 |SJ 2 − J 2 S|jm



= }2 hj 0 m0 |Sj(j + 1) − j 0 (j 0 + 1)S|jmi
= }2 [j(j + 1) − j 0 (j 0 + 1)] hj 0 m0 |S|jmi .

Also,

hj 0 m0 |[S, Jz ]|jmi = 0
= hj 0 m0 |SJz − Jz S|jmi
= } hj 0 m0 |Sm − m0 S|jmi
= }(m − m0 ) hj 0 m0 |S|jmi .

Therefore, hj 0 m0 |S|jmi must vanish unless j = j 0 and m0 = m0 . “Selection

rules” for non-zero elements are: Δj = 0 and Δm = 0.
Let sjm ≡ hjm|S|jmi denote the non-vanishing element. Since this is the
only non-zero matrix element, |jmi is an eigenvector of S, i.e. S |jmi = sjm |jmi.
Now we can show that the eigenvalues of the scalar operator S don’t depend
on m. Since S commutes with J± = Jx ± iJy , we have

S (J+ |jmi) = J+ S |jmi = sjm (J+ |jmi) .

But J+ |jmi is proportional to |j, m + 1i and still has some eigenvalue sjm
of S. We could continue this with J+2 → m + 2, . . . and with J− → m − 1,
1
These notes were prepared by Professor Dudley Herschbach of Harvard University

13
Lecture #27 Supplement #1 Page 27S(1)-14

J−2 → m−2, etc., and would get the same eigenvalue sjm of S for all m states
of a given j. Hence we would obtain
hj 0 m0 |S|jmi = hjkSkji δjj 0 δmm0
where hjkSkji is called a reduced matrix element, a number that does not
depend on m.
The above equation only describes the properties of S which are associated
with its scalar character. In general, the states of the system will depend
upon other quantum numbers in addition to j and m. If these are denoted
collectively by α, the scalar operator need not be diagonal in α, so the general
statement becomes
hα0 j 0 m0 |S|αjmi = hα0 jkSkαji δjj 0 δmm0
for
[S, J]
∼
= 0.
Vector Operators, V
Definition: A vector operator V∼ with respect to the angular momentum J∼
is any set of three operators Vx , Vy , Vz that satisfy the following com-
mutation rules:
X
[Ji , Vj ] = i} εijk Vk εijk = 1, ijk cyclic
k
= −1, ijk anti-cyclic
= 0, any two subscripts the same
This is shorthand for
[Jx , Vx ] = 0 [Jy , Vx ] = −i}Vz [Jz , Vx ] = i}Vy
[Jx , Vy ] = i}Vz [Jy , Vy ] = 0 [Jz , Vy ] = −i}Vx
[Jx , Vz ] = −i}Vy [Jy , Vz ] = i}Vx [Jz , Vz ] = 0.
It is convenient to use
1 1
V± = Vx ± iVy Vx = (V+ + V− ); Vy = (V+ − V− ).
2 2i

14
Lecture #27 Supplement #1 Page 27S(1)-15

Selection Rules for m

Consider the commutators involving Jz , take matrix elements of the commu-
tators:

a) [Jz , Vz ] = 0

hj 0 m0 |Jz Vz − Vz Jz |jmi = hj 0 m0 |m0 }Vz − Vz m}|jmi = }(m0 −m) hj 0 m0 |Vz |jmi

Thus hj 0 m0 |Vz |jmi = 0 unless m0 = m, Δm = 0

b) [Jz , V+ ] = [Jz , Vx + iVy ] = i}Vy + i(−i}Vx ) = }V+

hj 0 m0 |Jz V+ − V+ Jz |jmi = } hj 0 m0 |V+ |jmi.

or

}(m0 − m − 1) hj 0 m0 |V+ |jmi = 0

hj 0 m0 |V+ |jmi = 0 unless m0 − m = +1, Δm = +1

c) Similarly, [Jz , V− ] = −}V− and

hj 0 m0 |V− |jmi = 0 unless m0 − m = −1, Δm = −1

Selection Rules for j

To find the selection rules for j, we want to examine commutators of V∼ with
J∼ 2 . For this, some vector identities are useful. First we show
(1) J∼ × V∼ + V∼ × J∼ = 2i}V∼ .
This relation is another way to define a vector operator. It states that,
because of the non-commuting algebra of quantum mechanics, J∼ × V∼ = 6
−V∼ × J∼ as would hold for ordinary vectors.

15
Lecture #27 Supplement #1 Page 27S(1)-16

0.0.1 Proof:

re-label via j ↔ k
X
(J∼ × V∼ + V∼ × J ) =
∼ i
(εijk Jj Vk + εijk..... Vj66
Jk ) .....
.
......
..................... ......................
.......6
6 .............
j,k
X Then use εijk = −εijk
= εijk (Jj Vk − Vk Jj )
j,k
X X using the definition
= εijk [Jj , Vk ] = i} εijk εjk` V`
of a vector operator
j,k jk`

Note εijk εjk` = εijk ε`jk as a cyclic permutation of subscripts leaves εijk un-
changed.
Then
X factor 2 appears because both odd-odd
εijk ε`jk = 2δi` and even-even permutations give a
j,k contribution
So
X
(J∼ × V∼ + V∼ × J∼ )i = 2i} δi` V` = 2i}Vi Q.E.D.
`

Now we show
(2) [J∼ 2 , V∼ ] = i}(V∼ × J∼ − J∼ × V∼ ).

Proof:

X
[J 2 , Vj ] = [Ji2 , Vj ]
i
X
= {Ji [Ji , Vj ] + [Ji , Vj ]Ji }
i
⎧ ⎫
X⎨ ⎬
2
[J , Vj ] = i} Ji εijk Vk + εijk Vk Ji
⎩ |{z} |{z} ⎭
i,k

= i}(−J
∼
× V∼ + V∼ × J )
∼ j
Q.E.D.

16
 replace by replace by
−εjik εjki
Lecture #27 Supplement #1 Page 27S(1)-17

It is convenient to define the operator

1
K
∼
≡ (V∼ × J∼ − J∼ × V∼ ).
2
This is Hermitian (since V∼ and J∼ are) and is a vector operator if V∼ is.
Then Equation (2) states
[J∼ 2 , V∼ ] = 2i}K
∼
. (1)

However, we can’t yet use this commutator to get selection rules on V∼ , since
the matrix elements of the commutator K ∼
would seem to bear no simple
relation to those of V∼ . We will find that selection rules can be obtained from
an identity involving the double commutator,
(3) [J∼ 2 , [J 2
∼ ∼
, V ]] = 2}2 {J∼ 2 V∼ − 2(J∼ · V∼ )J∼ + V∼ J∼ 2 }.
This can be proven by examining further the properties of K ∼
.

[J 2 , [J∼ 2 , V∼ ]] = 2i}[J∼ 2 , K
∼
].

Since K
∼
is a vector operator, we have from (2) that

[J∼ 2 , K]
∼
= i}(K
∼
× J∼ − J∼ × K
∼
).

Also, from Equation (1) we have

J∼ × K
∼
+K
∼
× J∼ = 2i}K
∼
.

Hence

J∼ × K
∼
−K
∼
× J∼ = J∼ × K
∼
− (2i}K − J
∼
×K
∼
)
= 2J∼ × K
∼
− 2i}K
∼
.

Also, from equation (1)

1
K
∼
≡ (V∼ × J∼ − J∼ × V∼ ) = V∼ × J
∼
− i}V∼ .
2

17
Lecture #27 Supplement #1 Page 27S(1)-18

Thus,

J
∼
×K ∼
= J × (V∼ × J) ∼
− i}(J∼
× V∼ )
X
(J
∼
×K ) =
∼ i
εijk Jj (V∼ × J ) −i}(J∼ × V∼ )i
∼ k
j,k
| {z }
X
εijk Jj ε....k`m V` Jm .......shift to ε`mk = εk`m since cyclic
...... ....
.............................
.... ...... permutation of subscripts leaves ε
jk`m ........ 6 .......
................................

unchanged

X X
|(δi` δjm − δim δj` )Jj V` Jm = (Jj Vi Jj − Jj Vj Ji )
j`m j
X
= (Jj Jj Vi − Jj [Jj , Vi ] − Jj Vj Ji )
j
X
= J 2 Vi − Jj i}εji` V` −(J∼ − V∼ )Ji
j`
replace by −εij` ,
6
| {z } a non-cyclic
X permutation
+ i} Jj εij` V` = i}(J∼ × V∼ )i .
j`

So we find

2  
(J∼ × K ) =J
∼ i ∼
Vi − (J · V )Ji + 
∼ ∼
(J×V)i − 
i} (J×V)i
i}

or
J∼ × K
∼
= J∼ 2 V∼ − (J · V )J .
∼ ∼ ∼

Now we can use these results to simplify the double commutator,

[J∼ 2 , [J,
∼ ∼
V ]] = 2i}[J
∼
2
, K]
∼
= (2i})(i})(−1)(2J
∼
×K
∼
− 2i}K)
∼

= 2}2 (2J∼ × K
∼
− 2i}K
∼
)
= 2}2 {2J 2 V∼ − 2(J∼ · V∼ )J∼ − [J∼ 2 , V∼ ]
| {z }

18
Lecture #27 Supplement #1 Page 27S(1)-19

J 2 V∼ − V∼ J 2

and finally,

[J∼ , [J∼ 2 , V∼ ]] = 2}2 {J 2 V∼ − 2(J∼ · V∼ )J∼ + V∼ J 2 } Q.E.D.

Now we can obtain selection rules by taking matrix elements of this relation.
Consider two cases:

Case I: Elements diagonal in j: Wigner-Eckart Theorem

0 2
jm |[J , A]|jm = hjm0 |j(j + 1)A − Aj(j + 1)|jmi = 0


for any operator A. Thus,

0 2 2
jm |[J , [J , V ]]|jm = 0 = 2}2 jm0 |J 2 V∼ − 2(J∼ · V∼ )J∼ + V∼ J 2 |jm



or

}2 j(j + 1) jm0 |V∼ |jm − jm0 |(J



· V
∼ ∼ ∼
)J |jm =0
| {z }
X

jm0 |(J∼ · V∼ |j 00 m00 j 00 m00 |J∼ |jm .



=
j 00 m00

The operator (J∼ · V∼ ) is a scalar with respect to J∼ and therefore diagonal in

both m and j, so that j 00 = j and m00 = m0 , and its matrix elements are
independent of m. Hence we find



0  j|J∼ · V∼ |j
0 
jm |V∼ |jm = 2 jm |V∼ |jm
} j(j + 1)

This is the Wigner-Eckart theorem for a vector operator.

19
Lecture #27 Supplement #1 Page 27S(1)-20

Powell & Craseman, page 371



0  j |J · V |j
0
∼ ∼

jm |V∼ |jm = jm | J
∼
|jm
j(j + 1)}2

6
J∼ Suppose V∼ precesses around J∼ . The time averaged
...........
..........................................
........
value of the component normal to J∼ is zero. The
....... .....
.....
....
.
.....
....
....
....
...
time average of V∼ is therefore parallel to J∼ and has
.... ...

...
...
...
...
..
..
magnitude
.
..
...
- ..
..
. J ·V
...
...
...



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

....
....
...
..
|J|
.......
.......... .
...........
............................................


J∼ · V∼
V Hence, on this model, the average is
∼


|J|

 J∼ · V∼ (J · V ) J∼
-
V∼ = 2
J∼ =
J |J| |J|

The theorem is very useful, as it states that, for any vector operator V∼ ,
the matrix elements diagonal in j are simply proportional to the corre-
sponding matrix elements of J∼ itself. The proportionality constant, c0 (j) =


j |(J · V )|j (}j(j + 1)) is the same for all m–states. Therefore, we have
∼ ∼
via the Wigner-Eckart Theorem:

hj, m + 1|V+ |jmi = c0 (j)[j(j + 1) − m(m + 1)]1/2

hjm|Vz |jmi = c0 (j)m
hj, m − 1|V− |jmi = c0 (j)[j(j + 1) − m(m − 1)]1/2

with c0 (j) = hα0 jkV kαji a reduced matrix element. In particular, we note
that all matrix elements of V∼ between j = 0 states vanish.

Case II: Elements non-diagonal in j

Now consider j 0 6= j, again take matrix elements of Equation (3). LHS gives

0 0 2 2
j m |[J , [J , V ]]|jm = j 0 m0 |J 2 (J 2 V − V J 2 ) − (J 2 V − V J 2 )J 2 |jm



2
= {j 0 (j 0 + 1)2 − 2j(j + 1)j(j + 1) + j 2 (j + 1)2 } hj 0 m0 |V |jmi .

20
Lecture #27 Supplement #1 Page 27S(1)-21

RHS gives
* +
2h2 j 0 m0 |J 2 V∼ − 2(J∼ · V∼ )J∼ +V∼ J 2 |jm = 2}2 {j 0 (j 0 + 1) + j(j + 1)} j 0 m0 |V∼ |jm


| {z }

drops out as j 0 m0 |J|jm = 0, because j 0 =


∼
6 j.

Equating LHS = RHS and rearranging gives

{(j 0 − j)2 − 1} {(j 0 + 1 + 1)2 −1} j 0 m0 |V∼ |jm = 0


| {z }

This factor > 0 since j 0 =

6 j and j 0 ≥ 0, j ≥ 0

Therefore

0 0 
j m |V∼ |jm = 0

unless (j 0 − j)2 − 1 = 0 or j 0 = j ± 1.
The complete selection rules for any vector operator thus are:

0 0 
j m |V∼ |jm = 0 unless
j0 = j =
6 0 or j 0 = j ± 1

and, for any j 0 , j

m0 = m or m0 = m ± 1.

We have already found (page 19) the matrix element for j 0 = j. Now we will
do j 0 = j + 1.

Consider [J+ , V+ ] = [Jx + iJy , Vx + iVy ]

= [Jx , Vx ] + i[Jx , Vy ] + i[Jy , Vx ] − [Jy , Vy ]
↓ ↓ ↓ ↓
0 i}Vz − i}Vz 0
= 0.

21
Lecture #27 Supplement #1 Page 27S(1)-22

Take matrix element and use hj, m + 1|J+ |j, mi = }[j(j + 1) − m(m + 1)]1/2
= }[(j + m + 1)(j − 1)]1/2

0 = hj + 1, m + 1|(J+ V+ − V+ J+ |j, m − 1i
= hj + 1, m + 1|J+ |j + 1, mi hj + 1, m|V+ |j, m − 1i
− hj + 1, m + 1|V+ |jmi hjm|J+ |j, m − 1i

where we use the Δm = +1 selection rule for V+ and J+ .

This provides a recurrence relation for the matrix elements.

hj + 1, m + 1|J+ |j + 1, mi hj + 1, m|V+ |j, m − 1i = hj + 1, m + 1|V+ |jmi

× hjm|J+ |j, m − 1i
[(j(− 1)(j + m + 2)]1/2 [(j(− 1)(j + m)]1/2
( (
}( (m +(
(( }( (m +(
((

So
hj + 1, m|V+ |j, m − 1i hj + 1, m + 1|V+ |j, mi
1/2
= .
(j + m) (j + m + 2)1/2
This takes on a simple pattern if we divide both sides by (j + m + 1)1/2 :
hj + 1, m|V+ |j, m − 1i hj + 1, m + 1|V+ |j, mi
−c+ (j, m) ≡ 1/2
=
[(j + m + 1)(j + m)] [(j + m + 2)(j + m + 1)]1/2
= −c+ (j, m + 1).

Since m was arbitrary, c+ (j, m) = c+ (j, m + 1) = c+ (j, any other m) so the

ratio c+ (j) must be independent of m. The m–independence of the matrix
element is therefore given by

hj + 1, m + 1|V+ |j, mi = −c+ (j)[(j + m + 2)(j + m + 1)]1/2 ,

with c+ (j) = α0 , j + 1kV∼ kα, j a reduced matrix element that depends on



the detailed nature of V∼ , not merely on its vector character. However, it can
be evaluated if the matrix element of V∼ can be evaluated for any single m
value, e.g., m = j or m = 0, for which the evaluation is often simpler than
in the general case.

22
Lecture #27 Supplement #1 Page 27S(1)-23

Now determine the j 0 = j + 1 elements of VZ using the above result for V+ .

Start with
−2}VZ = [J− , V+ ] which expresses VZ in terms of J− and V+ ,
whose matrix elements we now know.

−2} hj + 1, m|VZ |j, mi = hj + 1, m|J− |j + 1, m + 1i hj + 1, m + 1|V+ |j, mi

− hj + 1, m|V+ |j, m − 1i hj, m − 1|J− |j, mi
= }[(j + m + 2)(j − m + 1)]1/2 (−c+ (j))[(j + m + 2)(j + m + 1)]1/2
− }(−c+ (j))[(j + m + 1)(j + m)]1/2 [(j + m)(j − m + 1)]1/2
1/2
= −}c+ (j)[(
j+
 + 2) − (j
m  +m)][(j + m + 1)(j − m + 1)]


= −2}c+ (j)[(j + m + 1)(j − m + 1)]1/2

Thus,
hj + 1, m|VZ |j, mi = c+ (j)[(j + m + 1)(j − m + 1)]1/2 .

Similarly, from
}V− = [J− , VZ ]

we find

hj + 1, m − 1|V− |j, mi = c+ (j)[(j − m + 2)(j − m + 1)]1/2 .

Results for j 0 = j − 1 are derived in analogous fashion and involve a third re-
duced matrix element, c− (j) = hα0 , j − 1kV kα, ji. Hence the m–dependence
of a scalar or vector operator follows from its scalar or vector character only.
Classification of operators by their transformation properties under rotation
can be extended to tensors of any rank. In each case the form of the matrix
elements is determined except for factors that depend on α and j.

23
Lecture #27 Supplement #1 Page 27S(1)-24

SUMMARY: Non-zero Matrix Elements of a Vector Op-

erator, V∼

Δj = +1 hj + 1, m ± 1|V± |jmi = c+ (j)[(j ± m + 2)(j ± m + 1)]1/2

hj + 1, m|VZ |jmi = c+ (j)[(j + m + 1)(j − m + 1)]1/2

Δj = 0 hj, m ± 1|V± |jmi = c0 (j)[(j ± m + 1)(j m)]1/2

hjm|VZ |jmi = c0 (j)m

Δj = −1 hj − 1, m ± 1|V± |jmi = ±c− (j)[(j m)(j m − 1)]1/2

hj − 1, m|VZ |jmi = c− (j)[(j − m)(j + m)]1/2

24
MIT OpenCourseWare
https://ocw.mit.edu/

5.73 Quantum Mechanics I

Fall 2018

For information about citing these materials or our Terms of Use, visit: https://ocw.mit.edu/terms.

25