Angular Momentum

Lecture 14
Angular momentum operator

algebra
In this lecture we present the theory of angular momentum operator algebra

in quantum mechanics.
14.1 Basic relations

Consider the three Hermitian angular momentum operators Jˆx , Jˆy and Jˆz ,
which satisfy the commutation relations
Jˆx , Jˆy = i~ Jˆz , Jˆz , Jˆx = i~ Jˆy , Jˆy , Jˆz = i~ Jˆx .

(14.1)
The operator
Jˆ2 = Jˆx2 + Jˆy2 + Jˆz2 , (14.2)
is also Hermitian and it commutes with Jˆx , Jˆy and Jˆz :
Jˆ , Jˆx = Jˆ , Jˆy = Jˆ , Jˆz = 0.

2 2 2
(14.3)
These relations are not difficult to prove using the operator identity

Â, B̂ Ĉ = Â, B̂ Ĉ + B̂ Â, Ĉ . (14.4)
1
2 LECTURE 14. ANGULAR MOMENTUM OPERATOR ALGEBRA
For example,
Jˆ , Jˆz = Jˆx + Jˆy2 + Jˆz2 , Jˆz

2 2
= Jˆx2 , Jˆz + Jˆy2 , Jˆz ,

(14.5)
because by definition Jˆz commutes with itself and, using (14.4),
Jˆz , Jˆz = − Jˆz , Jˆz Jˆz − Jˆz Jˆz , Jˆz = 0.

2
(14.6)
The remaining two commutators in the last row of (14.5) can be calculated
using again (14.4):
Jˆx , Jˆz = Jˆx , Jˆz Jˆx + Jˆx Jˆx , Jˆz

2

= − i~ Jˆy Jˆx + Jˆx Jˆy , (14.7)
and
Jˆy , Jˆz = Jˆy , Jˆz Jˆy + Jˆy Jˆy , Jˆz

2

= i~ Jˆx Jˆy + Jˆy Jˆx . (14.8)
The sum of Jˆx2 , Jˆz and Jˆy2 , Jˆz is therefore zero and from (14.5) it follows

that Jˆ2 , Jˆz = 0.

Rather then working with the Hermitian operators Jˆx and Jˆy , it is more
convenient to work with the non-Hermitian linear combinations,
Jˆ+ = Jˆx + iJˆy , (14.9a)
Jˆ− = Jˆx − iJˆy , (14.9b)
where, by definition, (Jˆ− )† = Jˆ+ . For reasons that will become clear later,
Jˆ+ and Jˆ− are called ladder operators. Using (14.1) and (14.3), it is straight-
14.1. BASIC RELATIONS 3
forward to show that
Jˆz , Jˆ+ = ~ Jˆ+ ,

(14.10a)
Jˆz , Jˆ− = − ~ Jˆ− ,

(14.10b)
Jˆ+ , Jˆ− = 2~ Jˆz ,

(14.10c)
Jˆ , Jˆ+ = Jˆ2 , Jˆ− = 0.

2
(14.10d)
The operators Jˆ+ and Jˆ− often appear in the products Jˆ+ Jˆ− and Jˆ− Jˆ+ , which
are equal to

Jˆ+ Jˆ− = Jˆx + iJˆy Jˆx − iJˆy
= Jˆx2 + Jˆy2 − i Jˆx , Jˆy

= Jˆx2 + Jˆy2 + ~ Jˆz , (14.11)
and

Jˆ− Jˆ+ = Jˆx − iJˆy Jˆx + iJˆy
= Jˆx2 + Jˆy2 + i Jˆx , Jˆy

= Jˆx2 + Jˆy2 − ~ Jˆz , (14.12)
respectively. After noticing that (14.2) implies Jˆx2 + Jˆy2 = Jˆ2 − Jˆz2 , we can
straightforwardly rewrite Jˆ+ Jˆ− and Jˆ− Jˆ+ as
Jˆ+ Jˆ− = Jˆ2 − Jˆz2 + ~ Jˆz , (14.13a)
Jˆ− Jˆ+ = Jˆ2 − Jˆz2 − ~ Jˆz . (14.13b)
Adding equations (14.13) side-by-side and rearranging the terms, we obtain
1ˆ ˆ
Jˆ2 = J+ J− + Jˆ− Jˆ+ + Jˆz2 . (14.14)
2
Two additional useful relations are:

n n
Jˆz , Jˆ+ = n~ Jˆ+ ,

(14.15a)
n n
Jˆz , Jˆ− = − n~ Jˆ− ,

(14.15b)
where n = 0, 1, 2, . . ., is an integer number. We demonstrate, for example,

the first relation (14.15a). The proof is by iteration. If n = 0, then (Jˆ+ )0 = Iˆ
and (14.15a) becomes a trivial identity. If n = 1, we simply recover (14.10a).
If n > 1, we first rewrite
Jˆz (Jˆ+ )n = Jˆz Jˆ+ (Jˆ+ )n−1 , (14.16)
and then use (14.10a) to write Jˆz Jˆ+ = Jˆ+ Jˆz + ~Jˆ+ . This permits us to find
the following recursion relation:
n n−1
Jˆz Jˆ+ = Jˆ+ Jˆz + ~Jˆ+ Jˆ+

n h n−1 i
= ~ Jˆ+ + Jˆ+ Jˆz Jˆ+ . (14.17)
If we rewrite this equation replacing n with n − 1, we easily find

n−1 n−1 h n−2 i
ˆ ˆ
Jz J+ ˆ
= ~ J+ ˆ ˆ
+ J+ Jz J+ˆ . (14.18)
Substituting this expression into (14.17), we obtain

n n n n−1 h n−2 io
Jˆz Jˆ+ = ~ Jˆ+ + Jˆ+ ~ Jˆ+ + Jˆ+ Jˆz Jˆ+
n 2 h n−2 i
= 2~ Jˆ+ + Jˆ+ Jˆz Jˆ+ . (14.19)
This procedure can be iterated again and again. After p iterations we find
n n p h n−p i
Jˆz Jˆ+ = p~ Jˆ+ + Jˆ+ Jˆz Jˆ+ , (14.20)
which becomes, for p = n,

n n n
Jˆz Jˆ+ = n~ Jˆ+ + Jˆ+ Jˆz . (14.21)
14.2. COMPATIBLE AND INCOMPATIBLE OBSERVABLES 5
This proves (14.15a). Equation (14.15b) can be obtained in the same way
using (14.10b).
Finally, from Jˆ2 , Jˆ± = 0 and the equation1

Â, B̂ n = Â, B̂ B̂ n−1 + B̂ Â, B̂ B̂ n−2 + . . . + B̂ n−1 Â, B̂ ,

(14.22)
where n is a positive integer number, it follows that
Jˆ , (Jˆ± )n = 0.
2
(14.23)
14.2 Compatible and incompatible observables

In quantum mechanics, two observables A and B are said to be compatible
when the corresponding operators Â and B̂ commute,

Â, B̂ = 0, (14.24)
and incompatible when they do not

Â, B̂ 6= 0. (14.25)
If A and B are compatible and |ψi is an eigenvector of Â associated with the

eigenvalue a,
Â|ψi = a|ψi, (14.26)
then B̂|ψi is also an eigenvector of Â associated with the same eigenvalue

a. This can be easily seen applying B̂ to both sides of (14.26):
B̂ Â|ψi = aB̂|ψi. (14.27)

Since, by hypothesis, Â, B̂ = 0, we can replace B̂ Â on the left side of this
equation with ÂB̂, thus obtaining

Â B̂|ψi = a B̂|ψi . (14.28)
Now, there are two possibilities:

1
This equation can be easily proved by induction using (14.4).
1) If a is a non-degenerate eigenvalue, then all vectors |ψi satisfying

(14.26) are parallel2 and B̂|ψi is necessarily proportional to |ψi, that
is
B̂|ψi = b|ψi. (14.29)
Therefore, |ψi is also an eigenvector of B̂.
2) If a is a degenerate eigenvalue, then the set of all vectors |ψi satisfying

(14.26) spans a subspace Ea associated with the eigenvalue a. Then, if
|ψi ∈ Ea , all what we can say is that B̂|ψi belongs to Ea :
B̂|ψi ∈ Ea . (14.30)
In the first case, the knowledge of the common eigenvector |ψi of Â and B̂,
uniquely determines the values of both a and b. This means that a and b are
not independent variables and the knowledge of a fixes the value of b and
vice versa. In the second case, instead, we need to specify both values a and
b to uniquely identify the common eigenvectors of Â and B̂. A fundamental
theorem of quantum mechanics states that3 :
When two observables A and B are compatible, it is always pos-

sible to find an orthonormal basis of the vector space spanned by
the eigenvectors common to the corresponding operators Â and
B̂.
We denote with {|a, bi} such a basis, where the eigenvectors |a, bi are defined
by the properties
Â|a, bi = a|a, bi, (14.31a)
B̂|a, bi = b|a, bi, (14.31b)

2
Here and hereafter, the statement that two vectors |ψi and |φi are parallel, signifies
that |ψi = γ|φi, where γ is any nonzero complex number.
3
For a demonstration see, e.g., C. Cohen-Tannoudji, B. Diu, F. Laloë, Quantum Me-
chanics, Vol. I (Wiley-Vch, 2005), Chapter II, § D-3-a.
14.2. COMPATIBLE AND INCOMPATIBLE OBSERVABLES 7
and
ha0 , b 0 |a, bi = δaa0 δbb 0 . (14.32)
If the ordered pair of eigenvalues (a, b) determine a unique common eigenvec-

tor |a, bi of Â and B̂, then the set of observables {A, B} is called a Complete
Set of Commuting Observables (CSCO), or a maximal set of commuting
observables; that is there is not a third observable C such that

Â, B̂ = Â, Ĉ = B̂, Ĉ = 0. (14.33)
The eigenvalues of operators Â and B̂ may still be degenerate, but if we

specify a pair (a, b), then the corresponding eigenvector |a, bi common to Â
and B̂ is uniquely specified.
The Hermitian operator Â possess at least one degenerate eigen-

value when there are two observables B and C compatible with A
but incompatible each other.
To prove this statement, consider three observables A, B and C such that

{A, B} is a CSCO, with

Â, B̂ = 0 = Â, Ĉ , and B̂, Ĉ 6= 0. (14.34)
According to our nomenclature, equations (14.34) imply that A is compatible

with both B and C, but B and C are incompatible. Let |a, bi be a common
eigenvector of Â and B̂ associated with the eigenvalues a and b, respectively.

From Â, Ĉ = 0 it follows that Ĉ|a, bi is also eigenvectors of Â associated
with the same eigenvalue a. Therefore, we have

Â B̂|a, bi = a B̂|a, bi , (14.35a)

Â Ĉ|a, bi = a Ĉ|a, bi , (14.35b)
Suppose that a is a non-degenerate eigenvalue. Then Ĉ|a, bi must be

proportional to B̂|a, bi, otherwise according to (14.35) Ĉ|a, bi and B̂|a, bi
would be two different eigenvectors associated with the same eigenvalue

a, which is incompatible with the hypothesis that a is a non-degenerate
eigenvalue. Therefore, we can write

Ĉ|a, bi = γ B̂|a, bi = γ b|a, bi, (14.36)
because B̂|a, bi = b|a, bi, and γ is a numerical constant. Multiplying by B̂

both sides of this equation, we obtain:

B̂ Ĉ|a, bi = γ b B̂|a, bi = γ b2 |a, bi. (14.37)
On the other hand, applying Ĉ to both sides of

B̂|a, bi = b|a, bi, (14.38)
we find

Ĉ B̂|a, bi = b Ĉ|a, bi = γ b2 |a, bi, (14.39)
where (14.36) has been used. Subtracting (14.39) from (14.37) we obtain

B̂, Ĉ |a, bi = 0. (14.40)
This equality cannot be true for all the basis vectors {|a, bi}, because this
would imply

B̂, Ĉ = 0, (14.41)
in contradiction to the assumption (14.34). Therefore, the presupposition
that the eigenvalue a is non-degenerate cannot be valid for all eigenvalues of
the operator Â. This means that Â has at least one degenerate eigenvalue.
14.3 Definitions and notation for the eigen-

values of Jˆ2 and Jˆz
From the definition (14.2) it follows that for any ket |ψi, the expectation
value hψ|Jˆ2 |ψi is nonnegative, because
hψ|Jˆ2 |ψi = hψ|Jˆx2 |ψi + hψ|Jˆy2 |ψi + hψ|Jˆz2 |ψi
2 2 2
= Jˆx |ψi + Jˆy |ψi + Jˆz |ψi ≥ 0,

(14.42)
14.3. THE EIGENVALUES OF Jˆ2 AND JˆZ 9
2
where hψ|Jˆx2 |ψi = hψ|Jˆx† Jˆx |ψi = Jˆx |ψi , et cetera (remember that Jˆx is

Hermitian, therefore hψ|Jˆx† = hψ|Jˆx ). If |ψi is an eigenvector of Jˆ2 associated

with the eigenvalue λ~2 (λ is thus dimensionless), then (14.42) implies
λ ≥ 0. (14.43)
It is conventional (but not mandatory) to introduce a nonnegative number
j ≥ 0 defined by
λ = j (j + 1) . (14.44)
For j ≥ 0 the function j (j + 1) is positive or null and monotonically increas-
ing, as shown in Fig. 14.1.
λ = j( j + 1)
2.0
1.5
1.0
0.5
0.2 0.4 0.6 0.8 1.0

j
Figure 14.1: √
Graph of the function λ = j (j + 1), for j ≥ 0. The vertical gray line mark
the value j = ( 5 − 1)/2, which gives λ = 1.
Therefore, if necessary, we can invert (14.44) to obtain

1 √
j= 4λ + 1 − 1 . (14.45)
2
According to Sec. 14.2, since Jˆ2 and Jˆz commute, it is possible to find a
set of common eigenvectors {|j, mi}, such that
Jˆ2 |j, mi = j(j + 1)~2 |j, mi, (14.46a)
Jˆz |j, mi = m~ |j, mi, (14.46b)

and
hj 0 , m 0 |j, mi = δjj 0 δmm 0 , (14.47)
where the eigenvalue of Jˆz is traditionally written as m~, and m is a dimen-

sionless number4 .
From inspection of (14.1-14.3), we can easily convince ourselves that us-
ing only Jˆx , Jˆy and Jˆz , it is not possible to build an additional (nontrivial)
Hermitian operator commuting with both Jˆ2 and Jˆz . Therefore, from the
results of Sec. 14.2, it follows that {J 2 , Jz } is a CSCO. This can be also
rigorously demonstrated using group theory, but this is outside the scope of
these lectures. In short, in the language of group theory, the set of the three
Hermitian operators Jˆx , Jˆy and Jˆz , is closed under commutation, as shown
by (14.1). These operators are the generators of a Lie group. Specifically, the
commutation relations (14.1) represent the Lie algebra of the SO(3) group.
The rank of a Lie group is defined as the largest number of generators com-
muting with each other. Since none of the operators Jˆx , Jˆy and Jˆz commute
with any other, the SO(3) group has rank 1 and it has a single independent
Casimir element 5 , which is Jˆ2 = Jˆx2 + Jˆy2 + Jˆz2 . Therefore, there are three
independent pairs of commuting operators: {Jˆ2 , Jˆx }, {Jˆ2 , Jˆy } and {Jˆ2 , Jˆz }.
We can choose any one of these pairs to build a CSCO. Traditionally, the
choice is {Jˆ2 , Jˆz }. So, assigning the two eigenvalues of Jˆ2 and Jˆz (that is,
fixing j and m), completely defines the eigenvector |j, mi. However, accord-
ing to the last result of Sec. 14.2, since the J 2 is compatible with Jx , Jy and
Jz , but Jx , Jy and Jz are reciprocally incompatible, we expect the eigenvalue
j(j + 1)~2 of Jˆ2 , to be degenerate.
4
It should be noted that at this point we are not making any assumption upon the either
discrete or continuous nature of the eigenvalues j and m. Therefore, δjj 0 and δmm0 can be
interpreted either as discrete Kronecker’s delta, or continuous Dirac’s delta functions.
5
In our non-technical parlance, we can say that a Casimir operator is an operator that
commutes with Jˆx , Jˆy and Jˆz .
14.4. EIGENVALUES AND EIGENVECTORS OF Jˆ2 AND JˆZ 11
14.4 Properties of the eigenvalues and eigen-

vectors of Jˆ2 and Jˆz
Let j(j + 1)~2 and m~ be the eigenvalues of Jˆ2 and Jˆz associated with the
eigenvector |j, mi. Then j and m satisfy the following inequality:
−j ≤ m ≤ j. (14.48)
To prove this assertion, consider first the two vectors Jˆ+ |j, mi and Jˆ− |j, mi.
By definition, their squared norms are nonnegative, that is
Jˆ+ |j, mi 2 = hj, m|Jˆ− Jˆ+ |j, mi ≥ 0,

(14.49a)
Jˆ− |j, mi 2 = hj, m|Jˆ+ Jˆ− |j, mi ≥ 0,

(14.49b)
where we have used Jˆ− = (Jˆ+ )† . Substituting (14.12) and (14.13) in the
equations above, we find
hj, m|Jˆ− Jˆ+ |j, mi = hj, m|Jˆ2 − Jˆz2 − ~ Jˆz |j, mi
= j(j + 1) − m(m + 1) ~2 hj, m|j, mi ≥ 0,

(14.50a)
hj, m|Jˆ+ Jˆ− |j, mi = hj, m|Jˆ2 − Jˆz2 + ~ Jˆz |j, mi
= j(j + 1) − m(m − 1) ~2 hj, m|j, mi ≥ 0.

(14.50b)
These equations imply
j(j + 1) − m(m + 1) = (j − m)(j + m + 1) ≥ 0, (14.51a)
j(j + 1) − m(m − 1) = (j + m)(j − m + 1) ≥ 0, (14.51b)
where to obtain the right sides we have added the null term jm − mj = 0 to
the left sides. For example,
j(j + 1) − m(m + 1) = j(j + 1) − m(m + 1) + jm − mj
= j(j + 1 + m) − m(m + 1 + j)
= (j − m)(j + m + 1). (14.52)

Then, from (14.51a) and (14.51b) it follows that
−(j + 1) ≤ m ≤ j, (14.53a)
−j ≤ m ≤ j + 1, (14.53b)
respectively. From j ≥ 0, it follows that −(j + 1) < −j ≤ j < j + 1.

Therefore, equations (14.53a) and (14.53b) are satisfied together only if m
satisfies the inequality
−j ≤ m ≤ j. (14.54)
This proves (14.48).

A necessary and sufficient condition for a ket |ψi to be a null vector,
that is |ψi = 0, is that its norm vanishes: hψ|ψi = 0. Therefore, from
(14.49a,14.50a,14.51a) it follows that
Jˆ+ |j, mi = 0 if and only if (j − m)(j + m + 1) = 0. (14.55)
The solutions of the algebraic equation (j − m)(j + m + 1) = 0 are m = j and

m = −j − 1 < −j. Since the values of m are constrained by −j ≤ m ≤ j,
only the solution m = j is acceptable and (14.55) implies:
Jˆ+ |j, ji = 0. (14.56)
Similarly, from (14.49b,14.50b,14.51b) it follows that
Jˆ− |j, mi = 0 if and only if (j + m)(j − m + 1) = 0, (14.57)
and, therefore,
Jˆ− |j, −ji = 0. (14.58)
The converse relations are also true, namely
Jˆ± |j, mi = 0 =⇒ m = ±j. (14.59)

14.4. EIGENVALUES AND EIGENVECTORS OF Jˆ2 AND JˆZ 13
This is easy to prove. Applying Jˆ∓ to both sides of Jˆ± |j, mi = 0 and using
(14.13) and (14.51), we obtain
Jˆ∓ Jˆ± |j, mi = Jˆ2 − Jˆz2 ∓ ~Jˆz |j, mi

= ~2 j(j + 1) − m(m ± 1) |j, mi

= ~2 (j ∓ m)(j ± m + 1) |j, mi = 0.

(14.60)
The last equality together with (14.48) implies m = ±j. What if m 6= ±j?
Then, the following two statements are true:
1) If m > −j, Jˆ− |j, mi is a non-null eigenvector of Jˆ2 and Jˆz associated
with the eigenvalues j(j + 1)~2 and (m − 1)~, respectively.
2) If m < j, Jˆ+ |j, mi is a non-null eigenvector of Jˆ2 and Jˆz associated

with the eigenvalues j(j + 1)~2 and (m + 1)~, respectively.
To prove 1), first we notice that from (14.49b,14.50b) it follows that Jˆ− |j, mi
is a non-null vector because its norm is positive for m > −j.
Then, we show that Jˆ− |j, mi is an eigenvector of Jˆ2 . Using (14.10d), we
can write
Jˆ , Jˆ− |j, mi = 0.
2
(14.61)
This can be rewritten as

J J− |j, mi = Jˆ− Jˆ2 |j, mi
ˆ2 ˆ

= j(j + 1)~2 Jˆ− |j, mi , (14.62)
which signifies that Jˆ− |j, mi is an eigenvector of Jˆ2 associated with the eigen-
value j(j + 1)~2 .
Now, we prove that Jˆ− |j, mi is an eigenvector of Jˆz . If we multiply from
right both sides of (14.10b) by |j, mi, we obtain
Jˆz , Jˆ− |j, mi = −~Jˆ− |j, mi,

(14.63)
which is equivalent to

ˆ
Jz J− |j, mi = Jˆ− Jˆz |j, mi − ~Jˆ− |j, mi
ˆ

ˆ
= (m − 1)~ J− |j, mi . (14.64)
This shows that Jˆ− |j, mi is an eigenvector of Jˆz with the eigenvalue (m−1)~.
If m < j, we can use arguments similar to the ones leading to (14.62)
and (14.64), to prove statement 2). Of course, we must replace Jˆ− |j, mi with
Jˆ+ |j, mi, work again with (14.10d) and use (14.10a) instead of (14.10b). The
final result is:

ˆ2 ˆ 2 ˆ
J J+ |j, mi = j(j + 1)~ J+ |j, mi ,

Jˆz Jˆ+ |j, mi = (m + 1)~ Jˆ+ |j, mi . (14.65)
These relations may be straightforwardly generalized to include powers

of J± , that is (Jˆ± )n , where n = 1, 2, . . . is a positive integer. Specifically, we
ˆ
assert that
3) If −j + n ≤ m ≤ j, then (Jˆ− )n |j, mi is an eigenvector of Jˆ2 and Jˆz

associated with the eigenvalues j(j + 1)~2 and (m − n)~, respectively.
4) If −j ≤ m ≤ j − n, then (Jˆ+ )n |j, mi is an eigenvector of Jˆ2 and Jˆz

associated with the eigenvalues j(j + 1)~2 and (m + n)~, respectively.
The proof is very simple. First, we use (14.23) to write:

n
Jˆ , Jˆ± |j, mi = 0,
2
(14.66)
which can be rewritten as

h n i n
Jˆ2 Jˆ± |j, mi = Jˆ± Jˆ2 |j, mi
h n i
= j(j + 1)~2 Jˆ± |j, mi . (14.67)
14.5. THE SPECTRUM OF Jˆ2 AND JˆZ 15
n
This relation means that Jˆ± |j, mi is an eigenvector of Jˆ2 with eigenvalue
j(j + 1)~2 . Next, if we multiply from right both sides of equations (14.15)
by |j, mi, we obtain
n n
Jˆz , Jˆ± |j, mi = ±n~ Jˆ± |j, mi,

(14.68)
namely,
h n i n n
Jˆz ˆ
J± |j, mi = Jˆ± Jˆz |j, mi ± n~ Jˆ± |j, mi
n n
= m~ Jˆ± |j, mi ± n~ Jˆ± |j, mi
h n i
= (m ± n)~ Jˆ± |j, mi . (14.69)
n
Therefore, Jˆ± |j, mi is an eigenvector of Jˆz with eigenvalue (m ± n)~.
However, since (14.48) requires that −j ≤ m ≤ j, we must demand
−j ≤ m + n ≤ j ⇐⇒ − j ≤ m ≤ j − n, (14.70a)
−j ≤ m − n ≤ j ⇐⇒ − j + n ≤ m ≤ j. (14.70b)
This concludes the demonstration of statements 3) and 4).
14.5 The spectrum of Jˆ2 and Jˆz

Now we are able to determine the possible values of j and m, that is, the
spectrum of Jˆ2 and Jˆz . According to our previous findings, since the set
{J 2 , Jz } is a CSCO, the knowledge of j and m uniquely identify the eigen-
vector |j, mi common to Jˆ2 and Jˆz . Therefore, once we know the spectrum of
Jˆ2 and Jˆz , we also know the common eigenvectors. Thus, let j(j + 1)~2 and
m~ be the eigenvalues of Jˆ2 and Jˆz associated with the eigenvector |j, mi.
We do not make any hypothesis about the values of j and m; we only require,
according to (14.48), that j and m satisfy the inequality −j ≤ m ≤ j, with
j ≥ 0. So, at this stage m can be any real number between −j and j.
Consider the two vectors (Jˆ+ )p |j, mi and (Jˆ− )q |j, mi, where p and q are
nonnegative integers. According to 3) of the previous section, (Jˆ− )q |j, mi
is an eigenvector of Jˆ2 and Jˆz with the eigenvalues j(j + 1)~2 and (m − q)~,
respectively. Similarly, from 4) it follows that (Jˆ+ )p |j, mi is an eigenvector
of Jˆ2 and Jˆz with the eigenvalues j(j + 1)~2 and (m + p)~, respectively. As
usual, (14.48) requires that
−j ≤ m − q, (14.71a)
m + p ≤ j. (14.71b)
Now, let us choose p and q to be the greatest nonnegative integers such

that:
m − (q + 1) < − j, (14.72a)
m + (p + 1) > j, (14.72b)
as illustrated in Fig. 14.2.
m-q m-1 m m+1 m+p

m
a b
–j j
Figure 14.2: Graphical representation of the inequalities (14.72). By hypothesis, the

nonnegative integers q and p are chosen to satisfy a = j +(m−q) < 1 and b = j −(m+p) <
1. The intervals a and b are pictured as gray bands.
The unnormalized eigenvectors of Jˆ2 and Jˆz with the eigenvalues of Jˆz pro-
portional to
m − q, . . . , m − 1, m, m + 1, . . . , m + p, (14.73)
are, according to 3) and 4) of the previous section,

q p
Jˆ− |j, mi, . . . , Jˆ− |j, mi, |j, mi, Jˆ+ |j, mi, . . . , Jˆ+ |j, mi. (14.74)
14.5. THE SPECTRUM OF Jˆ2 AND JˆZ 17
It is important to understand that these are the only possible vectors with
these properties, because the pair (j, m) uniquely identify (up to an irrelevant
multiplicative numerical constant) the eigenvector |j, mi. Given the leftmost
q p
and the rightmost vectors Jˆ− |j, mi and Jˆ+ |j, mi, we can calculate
h q+1 i h q+1 i
Jˆz Jˆ− |j, mi = (m − q − 1)~ Jˆ− |j, mi , (14.75a)
h p+1 i h p+1 i
Jˆz Jˆ+ |j, mi = (m + p + 1)~ Jˆ+ |j, mi . (14.75b)
q+1 q+1
According to (14.75a), either Jˆ− |j, mi = 0 or Jˆ− |j, mi is an eigen-
vector of Jˆz with eigenvalue m − q − 1. However, from (14.72) it follows
that m − q − 1 < −j, in contradiction with (14.48). Therefore, we conclude
q+1 p+1
that Jˆ− |j, mi = 0. An analogous argument yields to Jˆ+ |j, mi = 0.
When a vector is null, its norm is equal to zero, that is
2
ˆ q+1 q+1 q+1
J− |j, mi = hj, m| Jˆ+ Jˆ− |j, mi = 0, (14.76a)

2
ˆ p+1 p+1 p+1
J+ |j, mi = hj, m| Jˆ− Jˆ+ |j, mi = 0. (14.76b)

It is convenient to rewrite these equations as:

q+1 q+1 q q
hj, m| Jˆ+ Jˆ− |j, mi = hj, m| Jˆ+ Jˆ+ Jˆ− Jˆ− |j, mi = 0, (14.77a)

p+1 p+1 p p
hj, m| Jˆ− Jˆ+ |j, mi = hj, m| Jˆ− Jˆ− Jˆ+ Jˆ+ |j, mi = 0. (14.77b)

Substituting (14.12) and (14.13) in the equations above, we find
hα|Jˆ+ Jˆ− |αi = hα|Jˆ2 − Jˆz2 + ~ Jˆz |αi
= j(j + 1) − (m − q)(m − q − 1) ~2 hα|αi

= 0, (14.78a)
hβ|Jˆ− Jˆ+ |βi = hβ|Jˆ2 − Jˆz2 − ~ Jˆz |βi
= j(j + 1) − (m + p)(m + p + 1) ~2 hβ|βi

= 0, (14.78b)
where we have used the shorthand notation,

q p
|αi ≡ Jˆ− |j, mi, |βi ≡ Jˆ+ |j, mi. (14.79)
Equations (14.78) imply that:
j(j + 1) − (m − q)(m − q − 1) = (j + m − q)(j − m + q + 1) = 0, (14.80a)
j(j + 1) − (m + p)(m + p + 1) = (j − m − p)(j + m + p + 1) = 0. (14.80b)
The solution of these two algebraic equations are:
m = q − j, or m = j + (q + 1) > j, (14.81a)
m = j − p, or m = −j − (p + 1) < −j. (14.81b)
According to the condition (14.48), the only acceptable solutions are:
m =q − j ⇐⇒ m − q = −j, (14.82a)
m =j − p ⇐⇒ m + p = j. (14.82b)
Clearly, these two equalities are satisfied simultaneously only if

q+p
q−j =j−p ⇐⇒ j= . (14.83)
2
Therefore, j is equal to a positive or zero integer (by hypothesis q and p are
nonnegative integers) divided by 2:
1 3
j = 0,, 1, , . . . (14.84)
2 2
Equations (14.82) imply that if j is an integer, all m values are integers; if j
is a half-integer, all m values are half-integers. The allowed values of m for
a given j are therefore:
−j = m − q, m − q + 1, . . . , m + p − 1, m + p = j, (14.85)
that is
m = −j, −j + 1, . . . , j − 1, j . (14.86)
| {z }
2j + 1 values
14.6. MATRIX ELEMENTS OF ANGULAR-MOMENTUM OPERATORS19
This equation expresses the fact that each eigenvalue j(j + 1)~2 of Jˆ2 , is
2j + 1 times degenerate. We predicted this degeneracy at the end of Sec.
14.3, on the ground of commutation relations (14.1).
To illustrate the procedure of building the set {|j, mi} of 2j + 1 eigenvec-
tors of Jˆ2 and Jˆz with a given j, suppose to fix a value of j and to consider
the eigenvector |j, −ji with m = −j. This implies q = 0 and p = 2j, that
is one needs 2j steps to reach the other side of the spectrum with m = j.
Therefore, the eigenvalues of Jˆz are:
m = −j, −j + 1, . . . , j − 1, j, (14.87)
and the associated (unnormalized) eigenvectors of Jˆ2 and Jˆz , are obtained
by applying 2j times6 the ladder operator Jˆ+ to |j, −ji:
2j−1 2j
|j, −ji, Jˆ+ |j, −ji, . . . , Jˆ+ |j, −ji, Jˆ+ |j, −ji. (14.88)
Vice versa, if we start from the eigenvector |j, ji with m = j, then we have
q = 2j and p = 0. In this case, obviously the eigenvalues of Jˆz are still
m = −j, −j + 1, . . . , j − 1, j, (14.89)
and the associated (unnormalized) eigenvectors of Jˆ2 and Jˆz , are obtained
by applying 2j times the ladder operator Jˆ− to |j, ji:
2j 2j−1
Jˆ− |j, ji, Jˆ− |j, ji, . . . , Jˆ− |j, ji, |j, ji. (14.90)
14.6 Matrix elements of angular-momentum

operators
Let us assume that the kets |j, mi are normalized:
hj 0 , m 0 |j, mi = δjj 0 δmm 0 . (14.91)

6
Note that from (14.83) it follows that the number 2j = q+p is necessarily a nonnegative
integer.
Then, from equations (14.46) we clearly have
hj 0 , m 0 |Jˆ2 |j, mi = j(j + 1)~2 δjj 0 δmm 0 , (14.92)
and
hj 0 , m 0 |Jˆz |j, mi = m~ δjj 0 δmm 0 . (14.93)
To find the matrix elements of Jˆx and Jˆy , we first calculate the matrix ele-
ments of Jˆ± . We have from (14.60)
hj, m|Jˆ∓ Jˆ± |j, mi = ~2 (j ∓ m)(j ± m + 1) .

(14.94)
Since Jˆ± |j, mi is an eigenvector of Jˆz with eigenvalue (m ± 1)~, then it must
be the same as |j, m ± 1i up to a multiplicative constant7 . Thus, we set
Jˆ± |j, mi = c±
jm |j, m ± 1i, (14.95)
where the coefficients c±jm have to be determined. Substituting this expression

in the left side of (14.94), we obtain
± 2
c = ~2 (j ∓ m)(j ± m + 1) .

jm (14.96)
This equation determines c± jm up to an arbitrary phase factor. Convention-

ally, this phase is chosen to be zero. This choice guarantees that the matrix
elements of Jˆ± are all nonnegative; this is called the Condon-Shortley phase
convention. So, we found
Jˆ± |j, mi = (j ∓ m)(j ± m + 1) ~ |j, m ± 1i.

p
(14.97)
Therefore, the matrix elements of Jˆ± are
hj 0 , m 0 |Jˆ± |j, mi = (j ∓ m)(j ± m + 1) ~ δjj 0 δm±1,m 0 .

p
(14.98)
7
Remember that by definition, the eigenvalues a, b, . . . , z, of the operators Â, B̂, . . . , Ẑ
corresponding to a CSCO {A, B, . . . , Z}, uniquely determine the eigenvectors |a, b, . . . , zi
up to a multiplicative constant. Therefore, since Jˆ± |j, mi and |j, m ± 1i are associated
with the same pair of eigenvalues j(j + 1)~2 and (m ± 1)~, it must be Jˆ± |j, mi ∝ |j, m ± 1i.
14.6. MATRIX ELEMENTS OF ANGULAR-MOMENTUM OPERATORS21
Finally, we determine the matrix elements of Jˆx and Jˆy to be
1h 0 i
hj 0 , m 0 |Jˆx |j, mi = hj , m 0 |Jˆ+ |j, mi + hj 0 , m 0 |Jˆ− |j, mi , (14.99)
2
and
1 h 0 i
hj 0 , m 0 |Jˆy |j, mi = hj , m 0 |Jˆ+ |j, mi − hj 0 , m 0 |Jˆ− |j, mi . (14.100)
2i

Angular Momentum

Uploaded by

Document Information

Original Description:

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Angular Momentum

Uploaded by

Copyright:

Available Formats

Lecture 14

Angular momentum operator

In this lecture we present the theory of angular momentum operator algebra

14.1 Basic relations

Jˆx , Jˆy = i~ Jˆz , Jˆz , Jˆx = i~ Jˆy , Jˆy , Jˆz = i~ Jˆx .

Jˆ2 = Jˆx2 + Jˆy2 + Jˆz2 , (14.2)

is also Hermitian and it commutes with Jˆx , Jˆy and Jˆz :

Jˆ , Jˆx = Jˆ , Jˆy = Jˆ , Jˆz = 0.

Jˆ , Jˆz = Jˆx + Jˆy2 + Jˆz2 , Jˆz

= Jˆx2 , Jˆz + Jˆy2 , Jˆz ,

because by definition Jˆz commutes with itself and, using (14.4),

Jˆz , Jˆz = − Jˆz , Jˆz Jˆz − Jˆz Jˆz , Jˆz = 0.

Jˆx , Jˆz = Jˆx , Jˆz Jˆx + Jˆx Jˆx , Jˆz

Jˆy , Jˆz = Jˆy , Jˆz Jˆy + Jˆy Jˆy , Jˆz

that Jˆ2 , Jˆz = 0.

Jˆ+ = Jˆx + iJˆy , (14.9a)

Jˆ− = Jˆx − iJˆy , (14.9b)

forward to show that

Jˆz , Jˆ+ = ~ Jˆ+ ,

Jˆz , Jˆ− = − ~ Jˆ− ,

Jˆ+ , Jˆ− = 2~ Jˆz ,

Jˆ , Jˆ+ = Jˆ2 , Jˆ− = 0.

= Jˆx2 + Jˆy2 − i Jˆx , Jˆy

= Jˆx2 + Jˆy2 + ~ Jˆz , (14.11)

= Jˆx2 + Jˆy2 + i Jˆx , Jˆy

= Jˆx2 + Jˆy2 − ~ Jˆz , (14.12)

Jˆ+ Jˆ− = Jˆ2 − Jˆz2 + ~ Jˆz , (14.13a)

Jˆ− Jˆ+ = Jˆ2 − Jˆz2 − ~ Jˆz . (14.13b)

Adding equations (14.13) side-by-side and rearranging the terms, we obtain

Two additional useful relations are:

where n = 0, 1, 2, . . ., is an integer number. We demonstrate, for example,

Jˆz (Jˆ+ )n = Jˆz Jˆ+ (Jˆ+ )n−1 , (14.16)

If we rewrite this equation replacing n with n − 1, we easily find

Substituting this expression into (14.17), we obtain

which becomes, for p = n,

Â, B̂ n = Â, B̂ B̂ n−1 + B̂ Â, B̂ B̂ n−2 + . . . + B̂ n−1 Â, B̂ ,

where n is a positive integer number, it follows that

14.2 Compatible and incompatible observables

and incompatible when they do not

If A and B are compatible and |ψi is an eigenvector of Â associated with the

Â|ψi = a|ψi, (14.26)

then B̂|ψi is also an eigenvector of Â associated with the same eigenvalue

B̂ Â|ψi = aB̂|ψi. (14.27)

Now, there are two possibilities:

1) If a is a non-degenerate eigenvalue, then all vectors |ψi satisfying

B̂|ψi = b|ψi. (14.29)

Therefore, |ψi is also an eigenvector of B̂.

2) If a is a degenerate eigenvalue, then the set of all vectors |ψi satisfying

When two observables A and B are compatible, it is always pos-

Â|a, bi = a|a, bi, (14.31a)

B̂|a, bi = b|a, bi, (14.31b)

ha0 , b 0 |a, bi = δaa0 δbb 0 . (14.32)

If the ordered pair of eigenvalues (a, b) determine a unique common eigenvec-

The eigenvalues of operators Â and B̂ may still be degenerate, but if we

The Hermitian operator Â possess at least one degenerate eigen-

To prove this statement, consider three observables A, B and C such that

According to our nomenclature, equations (14.34) imply that A is compatible

Suppose that a is a non-degenerate eigenvalue. Then Ĉ|a, bi must be

would be two different eigenvectors associated with the same eigenvalue

because B̂|a, bi = b|a, bi, and γ is a numerical constant. Multiplying by B̂

On the other hand, applying Ĉ to both sides of

14.3 Definitions and notation for the eigen-

Hermitian, therefore hψ|Jˆx† = hψ|Jˆx ). If |ψi is an eigenvector of Jˆ2 associated