Angular Momenta

In this section, we begin the study of the quantum theory of angular momentum, concen-
trating initially on orbital angular momentum. The approach taken here is algebraic, i.e.
we try to derive as many things as possible from the algebra of the angular momentum
operators. Suppose our system is one particle in three dimensions. The orbital angular
momentum operator is
⃗ =X
L ⃗ × P⃗ ,

or
L1 = X2 P3 − X3 P2 , etc
Here Xj and Pk are the usual coordinate and momentum operators, satisfying

[Pj , Xk ] = −ih̄δjk

From this equation, we obtain

[Lj , Lk ] = ih̄ϵjkn Ln ,

or
[L1 , L2 ] = ih̄L3 , etc
In what follows, we will not refer to configuration space very often. However, we do need
one result, namely the formula for L3 in spherical coordinates. The wave function in
spherical coordinates is < r, θ, ϕ|Ψ > . The action of L3 is specified by

h̄ ∂
< r, θ, ϕ|L3 |Ψ >= < r, θ, ϕ|L3 |Ψ >
i ∂ϕ
If the particle is in an eigenstate of L3 the ϕ dependence of the wave function will be

< r, θ, ϕ|L3 |Ψ >∼ exp(imϕ),

where h̄m is the eigenvalue of L3 . For the motion of a particle, we demand that the wave
function be periodic in ϕ, so we must have

exp(im2π) = 1,

or m must be an integer (positive or negative or zero.) So when the type of angular

momentum under discussion is orbital, the third component of angular momentum has
eigenvalues which are integral multiples of h̄.
Scalar and Vector Operators
It is familiar in classical physics, that if we perform a rigid rotation of a system, that cer-
tain quantities remain unchanged whereas others do change in a regular way. Quantities
which remain unchanged under a rotation are called (rotational) scalars. An example
is the kinetic energy of a particle. If we rotate the “system” by rotating the velocity
vector, the kinetic energy does not change. On the other hand, the components of the
velocity certainly do change. (NOTE In all of our discussion of rotations, we will take
the so-called “active” viewpoint, where the rotation is applied to the object of interest,
and the coordinate system is held fixed.) If we use a ′ symbol to denote the rotated
object, and no ′ to denote the original, then for a scalar quantity, we have
S ′ = S. (1)
For a vector quantity, let us consider how the components transform if the system is
rotated (counter-clockwise) around the 3-axis by an angle α. By drawing a few pictures,
it is easy to see that the correct transformation properties for the components of a vector
V⃗ are given by
V1′ = cos αV1 − sin αV2 (2)
V2′ = cos αV2 + sin αV1
V3′ = V3
Now let us turn to quantum mechanics and require that expected values of scalar and
vector operators transform as in Eqs.(1) and (2). In the active viewpoint, we consider
that transformed state |Ψ′ >, which is related to the original state by the unitary trans-
formation representing (in the present case) the rotation. For simplicity, we will continue
to take our rotation to be one of angle α around the 3-axis. So for a scalar, we have
< S >′ ≡< Ψ′ |S|Ψ′ >=< Ψ|S|Ψ >≡< S > .
From transformation theory, the relation between |Ψ′ > and |Ψ > is
iL3 α
|Ψ′ >= exp(− )|Ψ >
h̄
Using our equation, we have
iL3 α iL3 α
< Ψ′ |S|Ψ′ >=< Ψ| exp( )S exp(− )|Ψ >=< Ψ|S|Ψ >,
h̄ h̄
or stripping off the states,
iL3 α iL3 α
exp( )S exp(− ) = S.
h̄ h̄
Now, if we differentiate, we get
iL3 α iL3 α iL3 α iL3 α
ih̄∂α exp( )S exp(− ) = exp( )[S, L3 ] exp(− ) = 0,
h̄ h̄ h̄ h̄
since the right hand side of the equation is independent of α. The result is

[S, L3 ] = 0.

Using other axes of rotation, we will clearly get

[S, Lk ] = 0, k = 1, 2, 3.

The conclusion is that an operator representing a rotational scalar must commute with
all components of the angular momentum.
⃗ Again taking the rotation to
For a vector operator, let us take the components of X.
be one of angle α around the 3 axis, from Eqs.(2), we have
iL3 α iL3 α
< Ψ| exp( )X1 exp(− )|Ψ > = cos α < Ψ|X1 |Ψ > − sin α < Ψ|X2 |Ψ > (3)
h̄ h̄
iL3 α iL3 α
< Ψ| exp( )X2 exp(− )|Ψ > = cos α < Ψ|X2 |Ψ > + sin α < Ψ|X1 |Ψ >
h̄ h̄
iL3 α iL3 α
< Ψ| exp( )X3 exp(− )|Ψ > = < Ψ|X3 |Ψ >
h̄ h̄
As in the case of the scalar operator, these equations must hold for an arbitrary state.
This is guaranteed if we demand that they hold for the operators themselves. Writing
out the first of Eqs.(3), we have
iL3 α iL3 α
exp( )X1 exp(− ) = cos αX1 − sin αX2
h̄ h̄
Differentiating both sides with respect to α, we obtain
iL3 α iL3 α
exp( )[X1 , L3 ] exp(− ) = −ih̄(sin αX1 + cos αX2 ).
h̄ h̄
This equation holds for all α. If we set α = 0 we obtain

[L3 , X1 ] = ih̄X2 .

Using other axes of rotation and considering an arbitrary three-vector operator V⃗ , we

can derive
[Lj , Vk ] = ih̄ϵjkn Vn . (4)
So turning the argument around, we can define a three-vector operator as one whose
components satisfy Eq.(4). If we have a single particle, then the list of vector operators
⃗ P⃗ , L.
is X, ⃗
To summarize, demanding that quantities transform under rotations in the same way
as they do in classical physics leads to a definite form for the commutation relations
between the operators representing these quantities and the angular momentum of the
system. The discussion was carried out for a single particle, but applies to a arbitrarily
complicated quantum system.
Eigenvalues of L2
We will determine the eigenvalues of L2 by making use of the operators L± , defined by
L± = L1 ± L2 .
The standard angular momentum commutation rules give
[L3 , L± ] = ±h̄L±
The action of L± changes the eigenvalue of L3 by ±h̄. For example,
L3 (L+ |m >) = (L+ L3 + [L3 , L+ ])|m >= h̄(mL+ |m > +L+ |m >) = h̄(m + 1)(L+ |m >).
We will get our result for the eigenvalue of L2 by expressing L2 in terms of the L± . We
have
L+ L− = (L1 + iL2 )(L1 − iL2 ) = L1 L1 + L2 L2 − i[L1 , L2 ] = L1 L1 + L2 L2 + h̄L3 , (5)
and
L− L+ = L1 L1 + L2 L2 − h̄L3
Relating these quantities to L2 , we have
L+ L− = L2 − L3 L3 + h̄L3 , (6)
and
L− L+ = L2 − L3 L3 − h̄L3 . (7)
Now assume the eigenvalue of L is a real number λ ≥ 0, and consider the matrix element
2

< λ, m|L+ L− |λ, m > . Inserting a complete set of eigenstates of L3 , we have

∑ ∑
< λ, m|L+ L− |λ, m >= < λ, m|L+ |λ, m′ >< λ, m′ |L− |λ, m >= | < λ, m′ |L− |λ, m > |2
m′ m′

From the last equality, we have that < λ, m|L+ L− |λ, m > ≥ 0. Using Eq.(5), we have
< λ, m|L+ L− |λ, m >=< λ, m|L2 − L3 L3 + h̄L3 |λ, m >= λ + h̄2 (m − m2 ).
It is clear that this expression will go negative for large enough m2 . The resolution is that
there must be a minimum value of m, denoted as M IN, for which L− |λ, M IN >= 0.
We then have
0 = λ + h̄2 (M IN − M IN 2 ), or λ = h̄2 M IN (M IN − 1)
A parallel argument applied to the matrix element of < λ, m|L− L+ |λ, m > shows that
there must be a maximum value of m, denoted as M AX, and related to λ by
λ = h̄2 M AX(M AX + 1).
The two expressions we have for λ must agree. The only acceptable relation between
M AX and M IN is M IN = −M AX. It is normal to set M AX = l, so the value of L2
becomes
L2 = h̄2 l(l + 1), l = 0, 1, 2, . . . , (8)
and states are labeled as |l, m > . The allowed values of m are then
−l ≤ m ≤ l. (9)
Matrix Elements of L± .
From Eq.(7), we can write

< l, m|L− L+ |lm >= | < l, m+1|L+ |l, m > |2 = hbar2 [l(l+1)−m(m+1)] = h̄2 [(l−m)(l+m+1)].

Choosing a phase and taking the square root, we have

√
< l, m + 1|L+ |l, m >= h̄ (l − m)(l + m + 1). (10)

Finally, replacing m by m − 1, we obtain

√
< l, m|L+ |l, m − 1 >=< l, m − 1|L− |l, m >= h̄ (l + m)(l − m + 1) (11)

At this point, we have the full set of results. The eigenvalues of L2 are given in Eq.(8).
Eq.(9) allows L3 to be written out as a (2l + 1) × (2l + 1) matrix, and finally by taking
combinations of Eqs.(6) and (7) the matrices for L1 and L2 can be constructed. With
the (standard) choice of phases we have made, the matrix for L1 is purely real, while
that for L2 is purely imaginary.
It is useful to work out a simple example. Consider the case of l = 1. The matrices
are 3 × 3 for this case. The matrix for L3 , is
 
1 0 0
 
L3 = h̄  0 0 0 
0 0 −1

From Eq.(10), we have

√ √
L+ |1, 0 >= h̄ 2|1, 1 >, L+ |1, −1 >= h̄ 2|1, 0 > .

These results allows us to obtain

 √ 
0 2 √0
 
L+ = h̄  0 0 2 .
0 0 0

Taking the adjoint, we have

 
0 0 0
 √ 
L− = h̄  2 √0 0  .
0 2 0

From these last two equations, L1 and L2 are obtained as

1 1
L1 = (L+ + L− ), L2 = (L+ − L− )
2 2i
Beyond Orbital Angular Momentum
In the above sections, the fact that the angular momentum being discussed was orbital
came in when we restricted the value of the third component of angular momentum to be
an integer multiple of h̄. This gives integer values of l and angular momentum matrices
which are (2l + 1) × (2l + 1) dimensional. This does not exhaust the possible sets of
matrices satisfying angular momentum commutation rules. To see this suppose we have
an angular momentum vector J, ⃗ with the following commutation rules,

[J 2 , Jk ] = 0, [Jk , Jn ] = ih̄ϵknp Jp .

Without assuming J3 has eigenvalues which are integral multiples of h̄, all the steps
of previous sections go through. In particular there must be a maximum value of J3 ,
denoted as M AX, and a minimum value, denoted as M IN, where M IN = −M AX, and
setting j = M AX, we have that the eigenvalue of J 2 can be written as

h̄2 j(j + 1).

The final question is, what are the allowed values of j? To settle this question, imagine
starting at the state |j, M IN > and successively applying the raising operator J+ until
the state |j, M AX > is reached. This must take an integer number of steps. Call the
integer n. Then we can write
M AX = M IN + n.
Using M IN = −M AX, we have
n
M AX = .
2
This allows half-integer solutions, so choosing j = 1/2, 3/2, 5, 2, . . . leads to a set of even
dimensional matrices which satisfy all the angular momentum commutation rules. It is
easy to check that trying to find further fractions, e.g. M AX = 1/4, cannot work.

Summary Including all possible cases, the total angular momentum quantum number
is restricted to j = 0, 1/2, 1, 3/2, 2, . . . . The square of the angular momentum takes values

h̄2 j(j + 1),

while J3 has eigenvalues h̄m, with

−j ≤ m ≤ j.

The matrix elements of J± satisfy

√
< j, m + 1|J+ |j, m >= h̄ (j − m)(j + m + 1), (12)

and √
< j, m − 1|J− |j, m >= h̄ (j + m)(j − m + 1) (13)
 The simplest example is j = 1/2, where we have
( )
h̄ 1 0
J3 =
2 0 −1

Using Eqs.(12) and (13), we have

( )
h̄ 0 1
J+ = ,
2 0 0

and ( )
h̄ 0 0
J− = ,
2 1 0
so ( )
h̄ 0 1
J1 = ,
2 1 0
( )
h̄ 0 −i
J2 = .
2 i 0
The coefficients of h̄/2 in the formulas for J1 , J2 , J3 , are the famous Pauli matrices
σ1 , σ2 , σ3 .