Some notes on “Modern Quantum Mechanics,” by Sakurai

and Napoletano

Maarten Golterman,

Department of Physics and Astronomy, San Francisco State University,

San Francisco, CA 94132, USA

These notes spell out some details of the arguments in the textbook
Ref. [1]. They do not replace the textbook, so they should be studied
along with the textbook.

1. Section 2.7: quantum mechanics and electromagnetism

1.1. Gauge transformations in Electromagnetism

Let us start with the hamiltonian for a free particle,
p~2
H= . (1.1)
2m
~ = −∇φ
The claim is that if we wish to couple this particle to an EM field with E ~
~ ~
and B = ∇ × A, the correct hamiltonian describing this is
1 e~ 2
 
H = p~ − A + eφ (1.2)
2m c
2
!
1 e ~+A e
~ · p~) + A ~ 2 + eφ .
= p~2 − (~p · A (1.3)
2m c c2
~ x) do not commute, so that the cross terms cannot be simplified.
Note that p~ and A(~
Note also that, with this definition, H is hermitian.
We will now derive the Heisenberg equations of motion, and show that they lead
to the Lorentz force law on our particle, thus validating our choice of hamiltonian.
First, using the Heisenberg commutation rules, it is straightforward to show that
dxi 1 1 e Πi
 
= [xi , H] = pi − Ai ≡ , (1.4)
dt ih̄ m c m
where we defined the kinematical momentum Πi . Let us work this out in more detail.
To find the commutator of xi with H, we only need the terms in H which depend on
the components pj of the canonical momentum p~:
!
1 e 1 h̄ h̄ e h̄ e
 
[xi , p2j − (pj Aj + pj Aj )] = −2 pi + 2 Ai = − pi − Ai , (1.5)
2m c 2m i ic mi c
1
and, multiplying this by 1/(ih̄), we find the right-hand side of Eq. (1.4).
Because Πi depends on both pi and the coordinates xj (through A), ~ the Πi do
not commute with themselves; using [pi , Aj ] = (h̄/i)∂i Aj (∂i ≡ ∂/∂xi ),
e e
[Πi , Πj ] = [pi − Ai , pj − Aj ] (1.6)
c c
h̄e
= (−∂i Aj + ∂j Ai ) (1.7)
ic
ih̄e
= ijk Bk , (1.8)
c
where we use the convention that a repeated index is summed over. Using this, we
find that
dΠi 1
= [Πi , H] (1.9)
dt ih̄
1 1 e
= [Πi , Πj Πj ] + [Πi , φ] (1.10)
ih̄ 2m ih̄
1 ih̄e e h̄
= ijk (Πj Bk + Bk Πj ) + ∂i φ (1.11)
2ih̄m c ih̄ i
e
= ijk (Πj Bk + Bk Πj ) − e∂i φ . (1.12)
2mc
Combining the two Heisenberg equations, we find that
d 2 xi
!
e dxj dxk
= ijk Bk − Bj + eEi , (1.13)
dt2 2c dt dt
which is nothing else than the Lorentz force equation for a charged particle (with
charge e) in an EM field. Note that we switched the summed-over indices j and k in
the term Bk Πj thus picking up a minus sign from the Levi–Civita tensor. Also, note
that, since dxj /dt = Πj does not commute with Bk , we cannot simplify this equation
more. We can do so in the classical limit, of course.
In the discussion above, we used that the commutation relations between opera-
tors in the Heisenberg picture (if they are all taken at the same time) work the same
as in the Schrödinger picture. Suppose we have three operators X, Y and Z, with

A(H) (t) = U † (t)A(S) U (t) , (1.14)

and likewise for B and C, and suppose that

[A(S) , B (S) ] = C (S) , (1.15)

then

[A(H) (t), B (H) (t)] = [U (t)† A(S) U (t), U † (t)B (S) U (t)] (1.16)
† (S) † (S) † (S) † (S)
= U (t)A U (t)U (t)B U (t) − U (t)B U (t)U (t)A U (t)
 
= U † (t) A(S) B (S) − B (S) a(S) U (t)
= U † (t)C (S) U (t) = C (H) (t) ,
2
which is the same commutation rule as Eq. (1.15) with all Heisenberg operators at
the same time t.

1.2. Continuity equation with EM fields

In the 2nd problem of the 5th problem set, you will verify Eq. (2.7.30) of Ref. [1].
Here, let us look at the expression in Eq. (2.7.33). Writing
√ iS/h̄
ψ= ρe , (1.17)
√
which just expresses the complex function in terms of its absolute value ρ and phase
S/h̄. It should be straightforward to derive Eq. (2.7.33) from this. Furthermore, by
√
using Eq. (2.7.55), one sees that ρ does not change under a gauge transformation
(and, since ρ is the probability density, it should not!), but the phase, S, does; it
transforms as
iS/h̄ → iS/h̄ + ieΛ/(h̄c) → S → S + eΛ/c . (1.18)
Using now also Eq. (2.7.36), we conclude that ∇S − eA/c ~ is gauge invariant, and
thus that ~j defined in Eq. (2.7.33) is gauge invariant. Note that, since ~j has to be
gauge invariant, and is given by Eq. (2.7.33), we need to have S transform as given
in Eq. (1.15) above. This gives us thus an alternative way to derive Eq. (2.7.55) in
Ref. [1]!

1.3. Aharonov–Bohm effect

Here are some details about the magnetic field for the Aharonov–Bohm effect. With
ρa being the radius of the cylinder, we define
( 1
~= B
2 0
ẑ × ~r , ρ < ρa
A 2 , (1.19)
1
B ρa ϕ̂ ,
2 0 ρ
ρ > ρa

where we use cylindrical coordinates ρ, ϕ and z. Remember that

ρ̂ = cos ϕ x̂ + sin ϕ ŷ , (1.20)

ϕ̂ = − sin ϕ x̂ + cos ϕ ŷ , (1.21)

so that (~r = ρ~ + z ẑ)

ẑ × ~r = ẑ × ρ~ = ρϕ̂ . (1.22)
Inside the cylinder, we have that Ax = − 21 B0 y, Ay = 12 B0 x and Az = 0. This leads to
Bx = By = 0m, Bz = B0 . (Check these claims!) Outside the cylinder, we have that

1 ρ2 1 ρ2 y
Ax = − B0 a sin ϕ = − B0 2 a 2 , (1.23)
2 ρ 2 x +y
2
1 ρa 1 ρ2 x
Ay = B0 cos ϕ = B0 2 a 2 , (1.24)
2 ρ 2 x +y
Az = 0 , (1.25)
3
hence (check!)
~ =0
B (outside cylinder) . (1.26)

2. Section 3.1: rotations

2.1. The rotation group

Rotations leave the inner product ~r1 · ~r2 between two (position) vectors ~r1 and ~r2
invariant.1 Using summation convention for repeated indices, the components of a
vector ~r transform under a rotation as

ri0 = Rij rj , (2.1)

where Rij are the components of a 3 × 3 matrix representing the rotation and ri0 are
the transformed components. Requiring the invariance of the inner product:

r~0 1 · r~0 2 = (R~r1 ) · R~r2 = ~r1 · RT R~r2 = ~r1 · ~r2 , (2.2)

in index notation:
0 0 T
r1i r2i = (Ri jr1j )(Rik r2k = r1j Rji Rik r2k = r1j r2j , (2.3)

and using that this has to be true for any pair ~r1 , ~r2 , we find that

RT R = 1 , (2.4)
T
with 1 the 3 × 3 unit matrix. In index notation, we need that Rji Rik = δjk , which
is Eq. (2.4) in index notation. Equation (2.4) implies that R is orthogonal (unitary
and real). This group is denoted as O(3): the group of orthogonal 3 × 3 matrices. If
we also require, as we will do in this chapter, that det(R) = 1, it is the group SO(3)
of “special” orthogonal 3 × 3 matrices.
All elements of SO(3) can be written in the form

ϕ) = eϕi Ti ,
R(~ (2.5)
     
0 0 0 0 0 1 0 −1 0
     
Tx = 0 0 −1 , Ty = 0 0 0 , Tz = 1 0 0  ,
     
    
     
0 1 0 −1 0 0 0 0 0

with Tx,y,z the generators of the group. Indeed, SO(3) is a group: it is associative
(because matrix multiplication is associative), there is a unit element R(~ ϕ = 0) = 1,
−ϕi Ti
each element R(~ ϕ) has an inverse R(−~ ϕ) = e (rotate around the same axis by
the opposite angle), and it is closed under multiplication (this means that the product
of two rotations is also a rotation, i.e., that the product of two elements of the group
1
Of course, they also leave the inner product between two momentum vectors, etc., invariant.

4
is an element of the group). Closure is not trivial, and we will return to that below.
First, let us work out an example:

Rz (ϕ) = eϕTz (2.6)

     
0 −1 0 −1 0 0 0 1 0
  1 2  1 3 
= 1 + ϕ 1 0 0 + ϕ  0 −1 0 + ϕ  −1 0 0  + . . .
     
  2   6  
0 0 0 0 0 0 0 0 0
   
1 − 21 ϕ2 + . . . −ϕ + 16 ϕ3 + . . . 0 cos ϕ − sin ϕ 0
   
 +ϕ − 61 ϕ3 + . . .
=  1 − 12 ϕ2 + . . . 0  =
 sin ϕ cos ϕ 0  ,
 
   
0 0 1 0 0 1

which is a rotation around the z-axis by an angle ϕ. A general rotation with param-
~ is a rotation around an axis with direction n̂ = ϕ̂ over an angle ϕ = |~
eters ϕ ϕ|.
To prove closure, we need the following two ingredients. First, it is straightforward
to check that

[Tx , Ty ] = Tz , [Ty , Tz ] = Tx , [Tz , Tx ] = Ty . (2.7)

(This can be written succinctly as [Ti , Tj ] = ijk Tk .) Second, we need the Baker–
Campbell–Hausdorff formula:
1 1
eA eB = eA+B+ 2 [A,B]+ 12 ([A,[A,B]]+[[A,B],B])+... . (2.8)

There is no closed-form expression, but one can prove that all terms in the exponent on
the right-hand side are “commutators of commutators,” like the ones explicitly shown.
Now, if we apply this to A = ϕi Ti and B = ψi Ti for two rotations R(~ ϕ) = eϕi Ti and
~ = eψi Ti , the exponent on the right-hand side of Eq. (2.8) becomes
R(ψ)
1 ~ i , (2.9)
(ϕi + ψi )Ti + ijk ϕi ψj Tk + . . . = (ϕi + ψi + jki ϕj ψk + . . .)Ti ≡ αi (~
ϕ, ψ)T
2
and the product of R(~ ϕ) and R(ψ) ~ is thus a rotation R(~ α), with α~ a complicated
function of ϕ ~
~ and ψ. We see that if we know the commutation rules of the generators
Ti , we know how elements of the group multiply. In other words, the algebra (2.7)
defines the structure of the group SO(3). Note that, if [A, B] = 0, Eq. (2.8) collapses
to eA eB = eA+B . For rotations, this happens when we consider two rotations around
the same rotation axis. A group for which all generators commute is an abelian group.
An example of an abelian group is the translation group: recall that translations are
generated by the momenta pi , and we have that [pi , pj ] = 0.
In fact, it is sufficient to consider only infinitesimal rotations, i.e., rotations over
an infinitesimal angle ϕ = |~ ϕ|. If we define  = ϕ/N with N very large, so that  is
very small, we have that
N
ϕ

N
Rz (ϕ) = lim R() = lim 1 + Tz = eϕTz . (2.10)
N →∞ N →∞ N
5
A group like this is called a Lie group, and the corresponding algebra (2.7) a Lie
algebra.

2.2. Rotations in quantum mechanics

Now, let us see how this all works in quantum mechanics. In the discussion,
note the similarity to the case of translations! The only thing that makes the case
of rotations more complicated (and more interesting!) is that rotations in general do
not commute, while translations always do.
If a system is represented by a ket |αi, the rotated system is represented by some
different ket
|αiR = D(R)|αi , (2.11)
which defines a linear operator D(R) for each rotation R. Note that, of course, D(R)
is represented by a matrix on the Hilbert space (all operators are), but the dimension
of D(R) is equal to the dimension of the Hilbert space, which is not (necessarily)
equal to 3. The operators D(R) have to follow the same rules as the rotations R
themselves; in jargon, the D(R) form a representation of the group SO(3). There is
a homeomorphism R → D(R) such that

R1 R2 = R3 ⇒ D(R1 )D(R2 ) = D(R3 ) . (2.12)

This implies that

|αiR1 R2 = D(R1 )|αiR2 = D(R1 )D(R2 )|αi . (2.13)

In words: if I first rotate a physical system, described by a ket |αi, by R2 , and then I
rotate it again by R1 , the ket describing this system should transform in accordance
with that.
Here is a trivial example: it could be that the physical system represented by |αi
does not change at all under rotations, which means that for such a system, for all R,
D(R) = 1! Clearly, Eq. (2.12) is satisfied, and this one-dimensional representation is
called the trivial representation. Referring to Eq. (3.1.15) of Ref. [1], for such a state
~
we have that J|αi = 0. In other words, such a state has zero angular momentum.
A physical example is the hydrogen atom in the ground state: you may recall that
the ”electron cloud” in the ground state looks spherically symmetric, so that indeed
rotations have no effect.

3. Section 3.2: SU (2)

The simplest example of a non-trivial representation is described in great detail

in Sec. 3.2 of Ref. [1]. Consider, in particular, Eq. (3.2.45) in Ref. [1], which works
out an explicit form of a 2-dimensional representation of the rotation group. Note
that the matrices on the right-hand side of this equation are 2 × 2 unitary matrices
with determinant equal to 1 (check this – you can do this by explicit calculation!).
6
These matrices form the group SU (2), i.e., the group of 2 × 2 unitary matrices with
determinant equal to 1. What we discover is that this group is homeomorphic with
the group SO(3), they are (almost) the same group, if we think about this group
in terms of their mathematical structure. They are not quite the same: clearly, for
ϕ = 2π, we have that R(2π) = 1, but if you substitute ϕ = 2π in Eq. (3.2.45), you
will find that !  
−iσ · n̂2π −1 0
exp =  . (3.1)
2 0 −1
In a sense, the group SU (2) is “twice as big” as SO(3), because the SU (2) matrices
for ϕ and ϕ + 2π are different: they differ by an oveall minus sign, i.e., D(n̂, ϕ) =
D(n̂, ϕ + 2π) (check this!), but get mapped onto the same element of SO(3). In this
sense, in quantum mechanics, the group SU (2) is the more fundamental group.
Note that we already know that nature makes use of this representation: the spin
of the electron transforms according to this representation! According to quantum
mechanics, the spin of the electron is thus a special example of angular momentum.

4. Section 3.5: j = 1 representation

In Sec. 3.5 of Ref. [1] we found a 3-dimensional representation of the rotation

group, the j = 1 representation. An example of a particle with spin 1 is the W
meson, which is electrically charged (in fact, W mesons with charges ±e exist). One
can therefore imagine subjecting it to a Stern–Gerlach type of experiment, just like
the electron, and measure its possible polarizations along some axis. (In practice, a
W meson decays pretty quickly, but that is an experimental problem!) Unlike the
case of the electron, the possible outcomes for the spin polarization are now +h̄, 0
and −h̄. So, if we consider the spin of the W , it lives in a 3-dimensional Hilbert
space. Operators acting on kets in this space are represented by 3 × 3 matrices.
However, even before thinking about quantum mechanics, we already knew a
3 × 3 matrix representation of the rotation group — it is how we defined the rotation
group in the first place, as the group of rotations on vectors in 3-dimensional physical
space leaving inner products of these vectors invariant. This raises the mathematical
question whether these two 3-dimensional representations are the same (or, more
precisely, equivalent). Are there two different 3-dimensional representations, or is
there only one? In the systematic analyis of Sec. 3.5, only one j = 1 representation was
found, so there is only one, and we would like to see how rotations in 3-dimensional
physical space can be brought into the form of the j = 1 representation found in
Sec. 3.5.
Let us work this out with a simple example, considering a rotation around the
z-axis. The generator we found in Sec. 3.5, Eq. (3.3.35b) is, writing out the matrix
for j = j 0 = 1:  
1 0 0
 
Jz = h̄  0 0 0  . (4.1)
 
 
0 0 −1
7
However, this does not coincide with (ih̄ times) Tz , given in Eq. (2.5)! We note that
the eigenvalues of Jz are h̄, 0, −h̄, while the eigenvalues of ih̄Tz are also equal to
h̄, 0, −h̄ (check this!). So, these two matrices are in fact the “same” (more precisely,
mathematically equivalent), but differ by a basis transformation. Below, we work out
what this basis transformation is.
Start with a rotation over an angle ϕ generated by Tz ; we have seen that this
takes the form ~r 0 = eTz ϕ~r, or, in components (cf. Eq. (2.6)):

x0 = x cos ϕ − y sin ϕ , (4.2)

y 0 = x sin ϕ + y cos ϕ ,
z0 = z .

Now define new variables

√ √
u = (x + iy)/ 2 , v = (x − iy)/ 2 . (4.3)

It is straightforward to work out that in terms of the variables u, v and z, Eq. (4.2)
takes the form

u0 = ueiϕ , (4.4)
z0 = z .
v 0 = ve−iϕ ,

and then that this transformation can be written as

      
u0 u eiϕ 0 0 u
      
z0  = eiJz ϕ/h̄  z  =  0 1 0 z  . (4.5)
      
 
      
0 −iϕ
v v 0 0 e v

The expression on the right-hand side is just Eq. (4.4) in matrix form, but it is also
what you get if you work out eiJz ϕ/h̄ in matrix form, so the middle expression and the
expression on the right-hand side of Eq. (4.5) coincide! Therefore, in this form, the
rotation around the z-axis is indeed generated by iJz /h̄, so the generators iJz /h̄ and
Tz only differ by a (unitary) basis transformation. This can be proven more generally
(i.e., that a basis transformation exists that maps the Ti into the iJi /h̄), and thus
the two 3-dimensional representations are equivalent. Mathematically, we say there
is only one 3-dimensional representation. Of course, nothing prevents nature from
making use of this representation in different physical applications, such as rotations
in physical 3-dimensional space, but also the existence of a 3-dimensional Hilbert space
corresponding to the angular momentum states of a particle with angular momentum
j = 1. Problem 3.14 of Ref. [1] asks you to work out the complete proof: a basis
transformation exists that takes the set ih̄Tx,y,z into the set Jx,y,z , respectively.

5. Section 3.6: orbital angular momentum

8
 This section discusses a few technical points that arise in Sec. 3.6 of Ref. [1] in
some more detail.

5.1. Derivation of Eq. (3.6.11)

Let us derive Eq. (3.6.11) of Ref. [1] in more detail, following the same logic that led
to Eq. (3.6.9). The idea is the same, but technically it is a little more involved. The
reason is that a rotation around the z-axis corresponds to a change only of the polar
angle φ, while θ remains constant. With a rotation around the x-axis, both φ and θ
change, and one thus expects terms involving both ∂/∂θ and ∂/∂φ.
Let us consider a rotation over an infinitesimal angle δϕ around the x-axis. This
infinitesimal rotation is represented by the operator 1 − iδϕLx /h̄, and we have that
i
 
0 0 0 0 0 0
hx , y , z | (1 − iδϕLx /h̄) |αi = hx , y , z | 1 − δϕ(ypz − zpy ) |αi (5.1)
h̄ !
∂ 0 ∂
= 1 − δϕy 0 + δϕz 0 hx0 , y 0 , z 0 |αi
0
∂z ∂y
0 0 0 0 0
= hx , y + z δϕ, z − y δϕ|αi .

Since rotations only change the polar angles, and not r, we can express the operator
Lx in terms of derivatives with respect to φ and θ, instead of derivatives with respect
to y 0 and z 0 . Assume, therefore, that
!
h̄ ∂ ∂
Lx = A +B . (5.2)
i ∂θ ∂φ

The task is to find A and B. For Lz this was done in Ref. [1], where it was found that
in that case A = 0, and B = 1. Now, because a rotation around the x-axis changes
both θ and φ, both A and B will be non-zero, and more complicated. Start from
!
∂ ∂
1 − δϕA − δϕB hr, θ, φ|αi = hr, θ − δϕA, φ − δϕB|αi , (5.3)
∂θ ∂φ

where we transitioned to write the wave function hx0 , y 0 , z 0 |αi in polar coordinates.
Comparing with Eq. (5.1), and remembering the relation between the cartesian and
polar coordinates, this gives us the equations

x0 ≡ r sin θ cos φ = r sin (θ − δϕA) cos (φ − δϕB) (5.4)

= r sin θ cos φ − r cos θ cos φ δϕA + r sin θ sin φ δϕB ,
0 0
y + z δϕ = r sin θ sin φ + r cos θδϕ = r sin (θ − δϕA) sin (φ − δϕB)
= r sin θ sin φ − r cos θ sin φ δϕA − r sin θ cos φ δϕB ,
0 0
z − y δϕ = r cos θ − r sin θ sin φδϕ = r cos (θ − δϕA)
= r cos θ − r sin θ δϕA .

From the last of these equations, we find that A = − sin φ, and substituting this into
the first or second equation, we then find that B = − cot θ cos φ (check both these
9
equations!). We thus arrive at
!
h̄ ∂ ∂
Lx = − sin φ − cot θ cos φ , (5.5)
i ∂θ ∂φ
which is Eq. (3.6.11) in position representation. A very similar calculation gives the
expression for Ly in Eq. (3.6.12).

5.2. Comment on Eq. (3.6.18)

In Eq. (3.6.18) of Ref. [1], we use the polar version of the gradient operator:

~ = r̂ ∂ + θ̂ 1 ∂ + φ̂ 1 ∂ .
∇ (5.6)
∂r r ∂θ r sin θ ∂φ
With ~r = rr̂, we see that
~ =r ∂ ,
~r · ∇ (5.7)
∂r
because, of course, r̂ · θ̂ = r̂ · φ̂ = 0.

5.3. Comment on Eq. (3.6.23)

For the discussion of spherical harmonics starting on page 202 of Ref. [1], it may
help to refer to the discussion on page 207, noting that there the energy eigenvalue
is denoted by E, and not by n. Also note that this n has nothing to do with the
directional unit vector n̂!
The directional eigenket |n̂i can also be written as

|n̂i = |θ, φi . (5.8)

It makes sense to consider this ket, because for a spherically symmetric problem, the
total energy and angular momentum eigen state (in position space) h~x 0 |n, `, mi =
hr, θ, φ|n, `, mi in Eq. (3.6.22) factorizes into two parts, with one only depending on
r, and the other only on the polar angles θ and φ.

6. Sec. 3.8: addition of angular momentum

6.1. Definitions
Suppose we have a system with two different angular momentum operators, J~1 and J~2 .
Examples are the electron in the hydrogen atom, which has both an orbital angular
momentum J~1 = L ~ and a spin J~2 = S,
~ or a system with two electrons, each with their
own spin, J~1 = S~1 and a spin J~2 = S~2 . Since these two angular momenta operators
are unrelated to each other (i.e., they correspond to different, unrelated, observables
of the system), they commute:
[J1i , J2j ] = 0 , (6.1)
10
for all components i and j of both operators. Formally, the operators J~1 and J~2 act
on different Hilbert spaces. In the example of the hydrogen atom, L ~ acts on the
(angular part) of the spatial wave function, while S ~ acts on the spin of the electron.
To get a description of all properties of the electron, we can combine these two Hilbert
spaces: using a position eigenbasis for the spatial properties of the electron, and the
Sz eigenkets | ± 21 i for the spin of the electron, a combined basis is |~x 0 i ⊗ | ± 21 i, where
hopefully the use of the “direct product” symbol ⊗ will become clear in what follows.
If, instead of position eigenkets |~x 0 i, we only consider orbital angular momentum
eigenkets in position space, we may use the orbital angular momentum eigenkets |`, mi
(here we are ignoring any quantum numbers not related to angular momentum, i.e.,
quantum numbers which do not change under rotations), and a basis for the electron’s
“combined” Hilbert space is
1 1
|`, m; ± i = |`, mi ⊗ | ± i , (6.2)
2 2
where we show two equivalent ways of writing states in this combined Hilbert space.
(We say that the combined Hilbert space is the direct product of the two individual
Hilbert spaces.) In general, for a system with two different angular momenta J~1 and
J~2 , using an eigenbasis for each, the combined eigenbasis of the product Hilbert space
consists of the eigenkets

|j1 , j2 ; m1 , m2 i = |j1 , m1 i ⊗ |j2 , m2 i . (6.3)

Since J~1 acts only on the states |j1 , m1 i and J~2 only on the states |j2 , m2 i, the notation
on the right in Eq. (6.3) makes sense if instead of J~1 we write J~1 ⊗ 1, and instead of
J~2 we write 1 ⊗ J~2 , with2

J~1 |j1 , j2 ; m1 , m2 i = (J~1 ⊗ 1)(|j1 , m1 i ⊗ |j2 , m2 i) = J~1 |j1 , m1 i ⊗ |j2 , m2 i , (6.4)

J~2 |j1 , j2 ; m1 , m2 i = (1 ⊗ J~2 )(|j1 , m1 i ⊗ |j2 , m2 i) = |j1 , m1 i ⊗ J~2 |j2 , m2 i .

The notation with the ⊗ symbol exhibits the product structure of this Hilbert space
very clearly, but often, in practice, the notation on the left-hand side of Eq. (6.4) is
used. The notation with the ⊗ also makes it very clear that the two angular momenta
commute:
[J~1 ⊗ 1, 1 ⊗ J~2 ] = [J~1 , 1] ⊗ [1, J~2 ] = 0 . (6.5)
With a little less mathematical precision we can also write that [J~1 , J~2 ] = 0, which is
how we think about it in physics, but the notation with the direct product symbol
clarifies the mathematical structure of these operators more clearly. Note that this
is all just notation to give a mathematically precise description of the physics we are
interested in!

6.2. Total angular momentum

2
For the states on the left of this equation I follow the notation of Ref. [1], in which first the j,
and then the m quantum numbers are listed.

11
We can now define a total angular momentum, which acts on the product Hilbert
space,
J~ = J~1 ⊗ 1 + 1 ⊗ J~2 , (6.6)
or, more succinctly, J~ = J~1 + J~2 . First, this is an angular momentum: because
using [J1i , J1j ] = ih̄ijk J1k and [J2i , J2j ] = ih̄ijk J2k , it is straightforward to show
that [Ji , Jj ] = ih̄ijk Jk (show this!). Since the components Ji satisfy the angular
momentum algebra, it is an angular momentum by definition. Of course, it also
makes sense intuitively that the sum of two angular momenta is again an angular
momentum.
The basis |j1 , j2 ; m1 , m2 i = |j1 , m1 i ⊗ |j2 , m2 i is composed of simultaneous eigen-
kets of J~12 , J1z , J~22 and J2z . But, one can, of course, also try to find a basis composed
of eigenkets of J~2 and Jz . However, such eigenkets are not the same as the eigenkets
|j1 , j2 ; m1 , m2 i, because J1z and J2z do not commute with J~2 (because they do not
commute with J~1 · J~2 ; show this!), so we cannot find simultaneous eigenkets of J~2 ,
Jz , J1z and J2z ! In contrast, we can find simultaneous eigenkets of J~2 , Jz , J~12 and J~22 ,
because these operators all commute with each other (again, show this). We can thus
construct a basis for our Hilbert space either using the eigenkets |j1 , j2 ; m1 , m2 i, or
simultaneous eigenkets |j1 , j2 ; j, mi of the commuting set J~2 , Jz , J~12 and J~22 , with

J~12 |j1 , j2 ; j, mi = j1 (j1 + 1)h̄2 |j1 , j2 ; j, mi , (6.7)

J~22 |j1 , j2 ; j, mi = j2 (j2 + 1)h̄2 |j1 , j2 ; j, mi ,
J~2 |j1 , j2 ; j, mi = j(j + 1)h̄2 |j1 , j2 ; j, mi ,
Jz |j1 , j2 ; j, mi = mh̄|j1 , j2 ; j, mi .

Since these two bases span the same Hilbert space, there should be a unitary basis
transformation relating the two, which is written in Eq. (3.8.33) of Ref. [1]. The
elements of this basis transformation are hj1 , j2 ; m1 , m2 |j1 , j1 ; j, mi, and the task of
Sec. 3.8 is the construction of these elements, the Clebsch–Gordan coefficients, given
two values of j1 and j2 .
An example where the new basis is useful is the case of a spin-orbit coupling L· ~ S,
~
~ ~ ~ ~
where in this example L = J1 and S = J2 . Since

~ = 1 J~2 − 1 L
~ 2 − 1S~ 2 = 1 J~2 − J~2 − J~2 ,
 
~ ·S
L 1 2 (6.8)
2 2 2 2
~ ·S
the new basis eigenkets |j1 , j2 ; j, mi are eigenstates of L ~ = J~1 · J~2 :

1  
J~1 · J~2 |j1 , j2 ; j, mi = h̄2 j(j + 1) − j1 (j1 + 1) − j2 (j2 + 1) |j1 , j2 ; j, mi . (6.9)
2
In contrast, the states |j1 , j2 ; m1 , m2 i are not eigenstates of J~1 · J~2 = J1x J2x + J1y J2y +
J1z J2z . They are eigenstates of the last term in this sum, J1z J2z , but not of the other
two terms (convince yourself!).

6.3. Discussion of the addition of two angular momenta

12
Suppose we have a basis of simultaneous eigenkets |j1 , j2 ; m1 , m2 i of two angular
momenta, J~1 and J~2 :

J~12 |j1 , j2 ; m1 , m2 i = j1 (j1 + 1)h̄2 |j1 , j2 ; m1 , m2 i , (6.10)

J~22 |j1 , j2 ; m1 , m2 i = j2 (j2 + 1)h̄2 |j1 , j2 ; m1 , m2 i ,
J1z |j1 , j2 ; m1 , m2 i = m1 h̄|j1 , j2 ; m1 , m2 i ,
J2z |j1 , j2 ; m1 , m2 i = m2 h̄|j1 , j2 ; m1 , m2 i .

We want to find the basis transformation to another set of simultaneous eigenkets,

now of the commuting set of operators J~12 , J~22 , J~2 and Jz , where J~ = J~1 + J~2 :

J~12 |j1 , j2 ; j, mi = j1 (j1 + 1)h̄2 |j1 , j2 ; j, mi , (6.11)

J~22 |j1 , j2 ; j, mi = j2 (j2 + 1)h̄2 |j1 , j2 ; j, mi ,
J~2 |j1 , j2 ; j, mi = j(j + 1)h̄|j1 , j2 ; j, mi ,
Jz |j1 , j2 ; j, mi = mh̄|j1 , j2 ; j, mi .

In what follows, to avoid cluttered notation, we will use units in which h̄ = 1. Since
Jz = J1z + J2z ,
Jz |j1 , j2 ; m1 , m2 i = (m1 + m2 )|j1 , j2 ; m1 , m2 i , (6.12)
and therefore, that

|j1 , j2 ; j, ji = |j1 , j2 ; j, m = j = j1 + j2 i = |j1 , j2 ; m1 = j1 , m2 = j2 i . (6.13)

This state, with j = j1 +j2 is the unique state with the maximal Jz eigenvalue possible,
and thus is the highest m-value component of a multiplet with angular momentum
j = j1 +j2 . The largest possible values of j and m are thus j = j1 +j2 and m = j1 +j2 .
We find the other componets of this multiplet by acting with the lowering operator
J− = J1− + J2− :
1 1
√ J− |j1 , j2 ; j, ji = √ (J1− + J2− )|j1 , j2 ; j, ji (6.14)
2j 2j
1
= √ (J1− + J2− )|j1 , j2 ; j1 , j2 i
2j
1
q q 
= √ 2j1 |j1 , j2 ; j1 − 1, j2 i + 2j2 |j1 , j2 ; j1 , j2 − 1i ,
2j
where we used that
q
J± |j, mi = (j ∓ m)(j ± m + 1)|j, m ± 1i . (6.15)

We thus find that

1
q q 
|j1 , j2 ; j, j − 1i = √ 2j1 |j1 , j2 ; j1 − 1, j2 i + 2j2 |j1 , j2 ; j1 , j2 − 1i . (6.16)
2j
√ √ √ √
Note that this state is normalized because ( 2j1 / 2j)2 +( 2j2 / 2j)2 = (j1 +j2 )/j =
1. Also note that this state is a linear combination of the two possible states with
13
m = j1 + j2 − 1, namely the states |j1 , j2 ; j1 − 1, j2 i and |j1 , j2 ; j1 , j2 − 1i. Up to an
overall phase, there is a unique state orthogonal to this state:
1
q q 
|j1 , j2 ; j − 1, j − 1i ≡ √ 2j2 |j1 , j2 ; j1 − 1, j2 i − 2j1 |j1 , j2 ; j1 , j2 − 1i . (6.17)
2j
Excercise: check that the states on the right-hand sides of Eqs. (6.16) and (6.17) are
indeed orthogonal. Moreover, if we act with J+ on the latter state, we find

J+ |j1 , j2 ; j − 1, j − 1i = (J1+ + J2+ )|j1 , j2 ; j − 1, j − 1i (6.18)

1
q q 
= √ (J1+ + J2+ ) 2j2 |j1 , j2 ; j1 − 1, j2 i − 2j1 |j1 , j2 ; j1 , j2 − 1i
2j
1
q q q q 
=√ 2j1 2j2 |j1 , j2 ; j1 , j2 i − 2j2 2j1 |j1 , j2 ; j1 , j2 i = 0 .
2j
This justifies the name we gave this state in Eq. (6.17): it is the component with
maximal m-value of a multiplet with angular momentum j1 + j2 − 1.
So far, we have thus identified two total angular momentum multiplets: one with
j = j1 + j2 , and one with j = j1 + j2 − 1. Of course, we can find the other components
(states) in these two multiplets by “laddering down,” i.e., applying the operator J−
repeatedly.
Let us carry out one more step of this exercise by applying the operator J− to
the state we found in Eq. (6.16) (omitting the overall normalization, which is not
relevant for the rest of the argument):
1 q  q 
J− |j1 , j2 ; j, j − 1i = (J1− + J2− ) √ 2j1 |j1 , j2 ; j1 − 1, j2 i + 2j2 |j1 , j2 ; j1 , j2 − 1i
2j
1 q q
= √ 2j1 2(2j1 − 1)|j1 , j2 ; j1 − 2, j2 i (6.19)
2j
q q
+ 2 2j1 2j2 |j1 , j2 ; j1 − 1, j2 − 1i
q !
q
+ 2j2 2(2j2 − 1)|j1 , j2 ; j1 , j2 − 2i .

Acting with J− on the state (6.17) will also give a linear combination of the three states
|j1 , j2 ; j1 −2, j2 i, |j1 , j2 ; j1 −1, j2 −1i and |j1 , j2 ; j1 , j2 −2i, but it will be a different linear
combination, orthogonal to the state (6.19); this is the state |j1 , j2 ; j − 1, m = j − 2i.
Excercise: Why? Work this out! But, since these two states are linear combinations
of three states, there is a third independent linear combination of these states, which
is orthogonal to both |j1 , j2 ; j, j − 2i and |j1 , j2 ; j − 1, j − 2i. This new state will be
the maximal m-value component of a multiplet with j = j1 + j2 − 2! It is thus the
state |j1 , j2 ; j − 2, j − 2i. Exercise: Convince yourself ! Of course, the minimal value
to which we can lower m1 is −j1 , and the minimal value to which we can lower m2 is
−j2 ; acting with Ji− on a state with mi = −ji will give zero (i = 1, 2).
We see that m1 or m2 get lowered each time we apply J− = J1− +J2− ; in Eq. (6.19)
to minimally m1 = j1 − 2 or m2 = j2 − 2. Continuing with the tower of states we are
14
obtaining starting with Eq. (6.13), i.e., our first multiplet, we find, ultimately, the
state
|j1 , j2 ; j, −ji = |j1 , j2 ; j, m = −j = −j1 − j2 i = |j1 , j2 ; m1 = −j1 , m2 = −j2 i . (6.20)
One way to see that this is where we have to end up is that we could have started with
this state, and obtain the same tower of states by acting with the raising operator
J+ = J1+ + J2+ .
Collecting what we know thus far, we find J~2 , Jz multiplets with j = j1 + j2 ,
j = j1 + j2 − 1, j = j1 + j2 − 2, etc. Where does this end, i.e., what is the lowest j
value we will find, carrying out this procedure to the end? In constructing our first
(j = j1 + j2 ) multiplet, we started with one state from the collection in Eq. (6.10) (cf.
Eq. (6.13)), we then found a linear combination of two of them (cf. Eq. (6.16)), and
then a linear combination of three of them (cf. Eq. (6.19)). Each time we laddered
down one step we found one more state to show up in the linear combination, and
each time we also found a new multiplet: this is how we found the multiplets with
j = j1 + j2 − 1 and j = j1 + j2 − 2. Now, without loss of generality, take j1 ≥ j2 .
Then, the number of states in these “growing” linear combinations stops growing
once we have encountered the state |j1 , j2 ; m1 = j1 , m2 = −j2 i, because states with
m2 = −j2 −1 do not exist. (Note how the m2 quantum number in the last state in the
linear combinations in Eqs. (6.16) and (6.19) gets lowered by one unit in each step.)
That means that the “latest” new multiplet we found, by systematically carrying out
our procedure is the multiplet with j = j1 − j2 (which is non-negative, because we
chose j1 ≥ j2 ). When m2 = −j2 is reached in our lowering process, no new states
from the collection (6.10) will show up in our linear combinations anymore, so all
states with m1 and m2 such that m = m1 + m2 = j1 − j2 will have been used to
complete multiplets we already were constructing. Therefore, the possible j values
for the total angular momentum J~ = J~1 + J~2 are
j = j1 + j2 , j1 + j2 − 1 , ... , |j1 − j2 | , (6.21)
where we dropped the restriction j1 ≥ j2 . Since the two bases (6.10) and (6.11) have
to consist of an equal number of states, a simple counting argument, given in Ref. [1],
confirms that we found all states in the new basis (6.11). Problem 3.24 of Ref. [1]
asks you to work this out for the case j1 = 1, j2 = 1.

7. Section 3.11: tensor operators

7.1. Details of Eq. (3.11.13)

Let us check that the three terms on the right-hand side of Eq. (3.11.13) indeed only
transform among themselves, and not into each other. Start with the first term. First,
~ · V~ is the inner product of the vectors U
U ~ and V~ , so, it is invariant under rotations.
Furthermore
Rii0 RiT0 j = δij ,
X X X
δij → Rii0 Rjj 0 δi0 j 0 = Rii0 Rji0 = (7.1)
i0 j 0 i0 i0

15
so also δij is invariant under rotations, and thus the whole first term is invariant under
rotations. In more technical terms, this part transforms in the j = 0 representation of
the rotation group (i.e., it does not transform at all, just like the angular momentum
eigenstate |j = 0, m = 0i!). Note that δij , a two-index object, does not transform at
all either — such a tensor is called an invariant tensor.
Now let us consider the second term. Applying a rotation to it:
X
Ui Vj − Uj Vi = Ui Vj − Vi Uj → (Rii0 Rjj 0 Ui0 Vj 0 − Rii0 Rjj 0 Vi0 Uj 0 ) (7.2)
i0 j 0
X
= Rii0 Rjj 0 (Ui0 Vj 0 − Vi0 Uj 0 ) .
i0 j 0

We see that the anti-symmetrized product of Ui and Vj transform among each other,
and not into the other two terms in Eq. (3.11.13). In fact, we recognize the compo-
~ × V~ , which has three components, and transforms like a
nents of the cross product U
vector. Thus, the three anti-symmetrized products in the second term of Eq. (3.11.13)
transform in the j = 1 representation of the rotation group.
Likewise, the symmetrized products in the third term of Eq. (3.11.13) also trans-
form among themselves:
X
Ui Vj + Uj Vi = Ui Vj + Vi Uj → (Rii0 Rjj 0 Ui0 Vj 0 + Rii0 Rjj 0 Vi0 Uj 0 ) (7.3)
i0 j 0
X
= Rii0 Rjj 0 (Ui0 Vj 0 + Vi0 Uj 0 ) .
i0 j 0

Now, note that if we set i = j in the third term, and we sum over i, we get
X 1 1~ ~ X 
(Ui Vi + Ui Vi ) − U · V δii = Ui Vi = 0 , (7.4)
i 2 3 i

so the third term is traceless, and the “33” component of this combination is thus
equal to minus the sum of the “11” and “22” components. Therefore, there are only
five independent components in the third term of Eq. (3.11.13). They all transform
into each other, in a five dimensional representation of the rotation group; thus, they
transform in the j = 2 representation. These j = 0, j = 1 and j = 2 tensors are
examples of spherical tensors.

7.2. Elements of the derivation of Eq. (3.11.25)

First, recall that the ket |n̂i is just the angular part of the position eigenket. We can
express the position eigenket |~x 0 i in terms of polar coordinates: |r, θ, φi. If we only
consider rotations, the coordinate r does not play a role, and we can just consider
the “abbreviated” ket |n̂i = |θ, φi. The Y`m (θ, φ) are nothing else than the (orbital)
angular momentum eigenstates |`mi expressed on a position eigenbasis:

Y`m (θ, φ) = hθ, φ|`mi . (7.5)

The steps leading from Eq. (3.11.18) to Eq. (3.11.22b) in Ref. [1] should be quite
straightforward; note that Eq. (3.11.18) is the same equation as Eq. (3.5.49), but now
16
we consider orbital angular momentum, so we relabel j → `, and instead of taking
the rotation to be R, we take it to be R−1 (which is also a rotation of course, so we
are free to do that).
Note that Eq. (3.11.22a) is a definition, inspired by Eq. (3.11.21): it defines a
spherical tensor Tq(k) by requiring it to transform in the same way as Ykq . Note the
(annoying!) choice of making the label k ↔ ` the superscript on T , even though
it is the subscript on Y , and likewise, the label q ↔ m is a subscript on T , but a
superscript on Y !
Then, since according to Eq. (3.11.4)

D(R) = 1 − i J~ · n̂ , (7.6)
h̄
the signs in Eq. (3.11.23) would be different if you derive it directly from Eq. (3.11.22a):
k
  
     
(k)
1 − i J~ · n̂ Tq(k) 1 + i J~ · n̂ = Tq0 hkq 0 | 1 − i J~ · n̂ |kqi .
X
(7.7)
h̄ h̄ q 0 =−k
h̄
This still leads to Eq. (3.11.24), as it should, of course. From there it should also be
straightforward to derive Eq.(3.11.25), following the “directions” given in Ref. [1], if
in Eqs. (3.5.35b) and (3.5.41) you substitute j → k and m → k.

7.3. Solving the recursion relation for Clebsch–Gordan coefficients: an

example
Clebsch–Gordan coefficients satisfiy the recursion relation (3.8.49) of Ref. [1]:
q
(j ∓ m)(j ± m + 1)hj1 j2 ; m1 , m2 |j1 j2 ; j, m ± 1i (7.8)
q
= (j1 ± m1 )(j1 ∓ m1 + 1)hj1 j2 ; m1 ∓ 1, m2 |j1 j2 ; j, mi
q
+ (j2 ± m2 )(j2 ∓ m2 + 1)hhj1 j2 ; m1 , m2 ∓ 1|j1 j2 ; j, mi .

Here, we work out an example. We take j1 = j2 = 21 , so that j can take the values 0
and 1. Taking the lower sign everywhere in Eq. (7.8) for j = 1, m = 0, m1 = 12 and
m2 = 12 yields:
√ √ 1 1 1 1  1 1 √ 1 1 1 1  1 1
*  + *  + *  +
1 1 1 1  1 1
2 ; ,  , ; 1, 1 = 1 ; − ,  ; 1, 0 + 1 ; , −  ; 1, 0 ,
2 2 2 22 2 2 2 2 22 2 2 2 2 22 2
(7.9)
1 1
while taking the lower sign, j = 1, m = 1, m1 = 2 and m2 = − 2 yields:
√ √ 1 1 3 1  1 1 √ 1 1 1 1  1 1
*  + *  + *  +
1 1 1 1  1 1
2 ; , −  , ; 1, 0 = 0 ; , −  ; 1, 1 + 1 ; ,  ; 1, 1 .
2 2 2 22 2 2 2 2 22 2 2 2 2 22 2
(7.10)
Note that, of course, m1 cannot be equal to 3/2, and indeed, that coefficient is
multiplied by zero in Eq. (7.10). Now take
*  +
1 1 1 1  1 1
; ,  ; 1, 1 = 1 . (7.11)
2 2 2 22 2
17
This happens to be the correct value (why?), but since the recursion relation (7.8) is
homogeneous, we have to fix one coefficient to find all the others. Substituting this
in Eq. (7.10), we find *  +
1 1 1 1  1 1 1
; , −  , ; 1, 0 = √ , (7.12)
22 2 2 2 2 2
and substituting both Eq. (7.11) and Eq. (7.12) into Eq. (7.9) then yields
*  +
1 1 1 1  1 1 1
; − ,  , ; 1, 0 = √ . (7.13)
22 2 2 2 2 2
Note that Eq. (7.11) is equivalent to
 +  +
1 1 1 1 1 1

 ; 1, 1 = 
 ; , , (7.14)
2 2 2 2 2 2

while Eqs. (7.12) and (7.13) are equivalent to

 +   !
1 1 1  11 1 1  11 1 1

 , ; 1, 0 = √ ; ,− +  ;− , . (7.15)
 

2 2 2  22 2 2  22 2 2

These are just equations (3.8.15a) and (3.8.15b) of Ref. [1], in more “pedantic” no-
tation. It is straightforward to find other Clebsch–Gordan coefficients with j = 1,
by choosing other values of m, m1 and m2 in Eq. (7.8) (note that we always should
choose m = m1 + m2 − 1 for the “upper” equation, and m = m1 + m2 + 1 for the
“lower” equation—why?).
Let us consider also the case j = 0, which is a separate set of recursion relations,
because rotations do not mix different j values. If j = 0 then also m = 0, and the
left-hand side of Eq. (7.8) will now vanish. Choosing m1 = 12 and m2 = 21 in the
“upper” equation thus yields:
√ √ √ 1 1 1 1  1 1
*  + *  +
1 1 1 1  1 1
0= 1 ; − ,  ; 0, 0 + 1 ; , −  ; 0, 0 , (7.16)
2 2 2 22 2 2 2 2 22 2

and thus *  + *  +
1 1 1 1  1 1 1 1 1 1  1 1
; − ,  ; 0, 0 = − ; , −  ; 0, 0 . (7.17)
2 2 2 22 2 2 2 2 22 2
We can choose the righ-hand side of this equation to be equal to a constant c, and
that fixes the other j = 0 Clebsch–Gordan coefficient. We thus find that
 +  +  +!
1 1 1 1 1 1 1 1 1 1
; 0, 0 =c ; ,− −  ; − , , (7.18)
  
 
2 2 2 2 2 2 22 2 2
√
which is nothing else than Eq. (3.8.15d) of Ref. [1] if we choose c = 1/ 2. Try finding
the remaining equation, Eq. (3.8.15c) from the recursion relations!

8. Section 4.2: parity

18
 Here is another proof of Theorem 4.1 (which is just an application of Theorem
1.2). Suppose that H commutes with π, and that H is not degenerate. It follows
that, for two energy eigenkets |Em i and |En i

0 = hEn |[H, π]|Em i = (En − Em )hEn |π|Em i . (8.1)

Thus, if n 6= m we find that hEn |π|Em i = 0, because then Em 6= En . Now, take

m = n and define pn = hEn |π|En i. We have that
X
π|En i = |Em ihEm |π|En i = pn |En i , (8.2)
m

because only the term m = n in the sum contributes. Thus, |En i is a parity eigenstate,
with eigenvalue pn . Since the only possible eigenvalues of π are ±1, we thus conclude
that energy eigenstates are parity even (pn = +1), or parity odd (pn = −1). An
example is the one-dimensional harmonic oscillator. Note that non-degeneracy of the
spectrum of H is important!

9. Section 5.1: non-degenerate perturbation theory

Consider a hamiltonian
H = H0 + λV , (9.1)
, where we have solved the eigenvalue problem for H0 , i.e., we know the eigenvalues
En(0) and eigenkets |n(0) i, with

H0 |n(0) i = En(0) |n(0) i . (9.2)

We do not know the eigenvalues En and eigenkets |ni of the full hamiltonian, and we
want to approximate them in perturbation theory in λ, which we will assume to be a
small parameter.
We expand

En = En(0) + λ∆(1) 2 (2)

n + λ ∆n + . . . , (9.3)
|ni = |n(0) i + λ|n(1) i + λ2 |n(2) i + . . . ,

and substitute this into Eq. (9.1):

(H0 + λV )(|n(0) i + λ|n(1) i + λ2 |n(2) i + . . .) (9.4)

(0) (1) 2 (2) (0) (1) 2 (2)
= (En + λ∆n + λ ∆n + . . .)(|n i + λ|n i + λ |n i + . . .) .

Terms of order λk should be equal on both sides of this equation. At order λ0 = 1,

we recover Eq. (9.2), as one would expect. At order λ, we find:

V |n(0) i + H0 |n(1) i = ∆(1) (0) (0) (1)

n |n i + En |n i , (9.5)

where we already divided by λ. This is a vector equation, with the vectors kets in our
Hilbert space. Therefore, let us consider the components of this vector equation on
19
the basis of normalized energy eigenkets of the unperturbed hamiltonian H0 , {|k (0) i}.
Taking first |k (0) i = |n(0) i, we find, taking the inner product with |n(0) i,

hn(0) |V |n(0) i + hn(0) |H0 |n(1) i = ∆(1) (0) (0) (1)

n + En hn |n i . (9.6)

Using that hn(0) |H0 = hn(0) |En(0) two terms cancel between the left- and right-hand
sides, and we find a very simple expression for the first-order energy shift:

∆(1) (0) (0)

n = hn |V |n i . (9.7)

Now let us take the inner product with |k (0) i =

6 |n(0) i:

hk (0) |V |n(0) i + hk (0) |H0 |n(1) i = En(0) hk (0) |n(1) i , (9.8)

(0)
where we used that hk (0) |n(0) i = 0. Using that hk (0) |H0 = hn(0) |Ek , we find
1
hk (0) |n(1) i = (0) (0)
hk (0) |V |n(0) i , (9.9)
En − Ek
or
1
|ni = cn (λ)|n(0) i + λ |k (0) ihk (0) |V |n(0) i + O(λ2 ) .
X
(0) (0)
(9.10)
k6=n En − Ek
Let us pause to consider what we learned. First, our calculation does not provide
us with a solution for the component of |n(1) i in the |n(0) i direction. We thus have
an unknown constant cn (λ) multiplying |n(0) i in Eq. (9.10); all we know is that
cn (λ = 0) = 1. For now, we will arbitrarily set cn (λ) = 1 for all λ, but we will return
to this issue below. Second, we observe that the result is singular if unperturbed
(0)
6 |n(0) i exist with Ek = En(0) , i.e., if the n-th level is degenerate! For
eigenkets |k (0) i =
now, we will assume that the n-th level is not degenerate, and postpone a discussion
of the degenerate case till later.
Let us also consider what happens at order λ2 . From Eq. (9.4) we obtain

V |n(1) i + H0 |n(2) i = ∆(2) (0) (1) (1) (0) (2)

n |n i + ∆n |n i + En |n i , (9.11)

where we divided by λ2 . Proceeding as we did before, we first take the inner product
with |n(0) i; using again that hn(0) |H0 = hn(0) |En(0) the terms with |n(2) i cancel again,
and we find
∆(2) (0) (1)
n = hn |V |n i . (9.12)
Here we used that, with choosing cn (λ) = 1, |n(1) i does not have a component along
|n(0) i, i.e., hn(0) |n(1) i = 0 (so we made a clever choice for cn (λ)!). Substituting the sum
in Eq. (9.10), which equals |n(1) i into this, we get our final result for the second-order
energy shift:
1
∆(2) hn(0) |V |k (0) ihk (0) |V |n(0) i .
X
n = (0) (0)
(9.13)
k6=n En − Ek

Defining a short-hand Vkn = hk (0) |V |n(0) i, we can also write this as

|Vkn |2
∆(2)
X
n = (0) (0)
, (9.14)
k6=n En − Ek
20
 ∗
if we recall that Vnk = Vkn . Again, this result works only if the n-th level is not
degenerate. Finally, take the inner product of Eq. (9.11) with |k (0) i =
6 |n(0) i:

hk (0) |V |n(1) i + hk (0) |H0 |n(2) i = ∆(1) (0) (1) (0) (0) (2)
n hk |n i + En hk |n i , (9.15)
(0)
and using again that hk (0) |H0 = hn(0) |Ek and hk (0) |n(0) i = 0 we obtain
1  
hk (0) |n(2) i = (0) (0)
hk (0) |V |n(1) i − ∆(1) (0) (1)
n hk |n i (9.16)
En − Ek
1 1
hk (0) |V |`(0) ih`(0) |V |n(0) i
X
= (0) (0) (0) (0)
En − Ek `6=n En − E`
1
− (0) (0)
hn(0) |V |n(0) ihk (0) |V |n(0) i
(En − Ek )2
1 X Vk` V`n Vkn Vnn
= (0) (0) (0) (0)
− (0) (0)
.
En − Ek `6=n En − E` (En − Ek )2

Again, we see that the component hn(0) |n(2) i does not follow from our calculation,
but, recall that we chose hn(0) |ni = 1 for the full solution, to all orders in λ.
Let us now address this latter point. We want the state |ni to be normalized,
but the state we constructed actually is not normalized! So, let us define
q
|niN = Zn |ni , (9.17)

for some real number Zn > 0 such √ that |niN is normalized, N hn|niN = 1. Observing
that hn(0) |ni = 1, we see that Zn = hn(0) |niN , so that the probability to find |niN
in the state |n(0) i is equal to Zn .
Using Eq. (9.17) we can calculate Zn from 1 = N hn|niN = Zn hn|ni:

Zn−1 = hn|ni (9.18)

  
(0) (1) 2 (2) (0) (1) 2 (2)
= hn | + λhn | + λ hn | + . . . |n i + λ|n i + λ |n i + . . .
= 1 + λ2 hn(1) |n(1) i + O(λ3 ) ,

where in the last step we used that |n(1) i and |n(2) i were cleverly chosen to be or-
thogonal to |n(0) i. Substituting |n(1) i into this, we find (prove this!)
|Vkn |2
Zn = 1 − λ2 + O(λ3 ) .
X
(0) (0)
(9.19)
k6=n (En − E k )2

10. Sec. 5.2: degenerate perturbation theory

Here we review a of of the technical details of Sec. 5.2 of Ref. [1], considering first
the projectors needed to project on the degenerate subspace D and its “complement”
(which contains all states orthogonal to any of the states inside D), and then a few
calculational details of the example of the linear Stark effect.
21
10.1. Projectors
In Sec. 5.2 of Ref. [1], the projector P0 onto the subspace D spanned by the degenerate
H0 eigenstates is defined as
g
X (0) (0)
P0 = |mi ihmi | , (10.1)
i=1

where we use a shorthand notation

(0) (0)
|mi i ≡ |ED , ii , (10.2)
(0)
where ED is the common energy eigenvalue, and the index i labels the g degenerate
eigenstates. The projector P1 onto the space orthogonal to D is defined as

P1 = 1 − P0 (10.3)
g
X (0) (0) X (0) (0)
= |Ek ihEk | − |mi ihmi |
k i=1
X (0) (0)
= |Ek ihEk | . (10.4)
k∈{m
/ 1 ,...,mg }

We have that
g X
g
(0) (0) (0) (0)
P02 =
X
|mi ihmi |mj ihmj | (10.5)
i=1 j=1
g X g g X
g
X (0) (0) X (0) (0)
= |mi iδij hmj | = |mi ihmi | = P0 ,
i=1 j=1 i=1 j=1

P12 = (1 − P0 )2 = 1 − 2P0 + P02 = 1 − P0 = P1 , (10.6)

and
P 0 P1 = P1 P 0 = 0 , (10.7)
(0) (0) (0) (0)
because hmj |Ek i = 0 if k ∈
/ {m1 , . . . , mg }, and thus Ek 6= ED .

10.2. Linear Stark effect

Here are a few more details on the linear Stark effect. Let us derive Eq. (5.2.19)
of Ref. [1] in more detail. First, notation. Historically, instead of labeling hydrogen
energy eigenstates with `, the labels s (` = 0), p (` = 1), d (` = 2), etc. were used,
so

|2si ≡ |n = 2, ` = 0, m = 0i , |2p, m = 0i ≡ |n = 2, ` = 1, m = 0i . (10.8)

The two matrix elements in Eq. (5.2.19) are complex conjugates of each other (since
the operator V is hermitian), so we need to calculate only one of them:
~
h2s|V |2p, m = 0i = −e|E|h2s|z|2p, m = 0i (10.9)
22
 ~
= −e|E|h2s|r cos θ|2p, m = 0i
Z ∞ Z 1 Z 2π
~
= −e|E| drr2 d(cos θ) dφ
0 −1
h i0∗ h i
× R20 (r)(Y00 )(θ, φ) r cos θ R21 (r)Y10 (θ, φ)
Z ∞ Z 1
~
= −2πe|E| drr 3 ∗
R20 (r)R21 (r) d(cos θ)(Y00 )∗ (θφ) cos θY10 (θ, φ)
0 −1
Z ∞
1 r r
 
~
= −e|E| drr 2− 3
√ e−r/a0
3
0 8a0 a0 3a
s 0
Z 1
1 3
× 2π d(cos θ) √ cos2 θ
−1 4π 4π
 √
!
~
 1 ~ 0,
= −e|E| 3 3a0 × √ = 3e|E|a
3
where a0 is the Bohr radius, we used Eqs. (B.5.7) and (B.6.7) of Ref. [1], and we used
that Z ∞
dx xn e−x = Γ(n + 1) = n! (10.10)
0
after substituting y = r/a0 . Therefore, the matrix (5.2.18) of Ref. [1] is equal to
 
0 1 0 0
 
1 0 0 0 
 
P0 V P0 = ~ 0
3e|E|a   , (10.11)
0 0 0 0 
 

 
0 0 0 0

on the basis of the four degenerate H0 eigenstates

 
|2si
 
|2p, m = 0i 
 

  . (10.12)
|2p, m = 1i 
 

 
|2p, m = −1i

The 4 × 4 matrix P0 V P0 has eigenvalues

~ 0,
+3e|E|a ~ 0,
−3e|E|a 0, 0. (10.13)

The energy levels of the perturbed p-wave states thus become

e2 ~ 0,
E2+ = − + 3e|E|a 1 state , (10.14)
8a0
e2 ~ 0,
E2− = − − 3e|E|a 1 state ,
8a0
e2
E2 = − , 2 states .
8a0

23
 The eigenstates for the first two eigenvalues are, respectively,
1 1
√ (|2si + |2p, m = 0i) , √ (|2si − |2p, m = 0i) . (10.15)
2 2
(Check this!) See Eqs. (5.2.20) and (5.2.21), as well as Fig. 5.1 in Ref. [1].
For perturbation theory to work, we want the energy shifts to be small compared
to the differences between energy levels of the unperturbed states. The smallest
energy difference for the n = 2 energy level is that with the n = 3 energy level
e2 e2 5e2
E3 − E2 = − + = . (10.16)
18a0 8a0 72a0
~ 0 5e2 ~  5e2
We thus want 3e|E|a 72a0
, or |E| 216a20
≈ 2 × 10−19 V/m.

Appendix A. Typos in Ref. [1]

Eq. (3.5.4b) should read

Jz |a, bi = b|a, bi . (A.1)
Eq. (3.5.33) should read

m = −j, −j + 1, . . . , j − 1, j .
| {z }
2j + 1 states

References
[1] J. J. Sakurai and J. Napoletano, Modern Quantum Mechanics, 2nd edition,
Addison-Wesley, 2011

24