∗

II Principles of Quantum Mechanics

Rain Zhu
[email protected]
Presented by Dr Enrico Pajer

Michaelmas 2024

Contents
1 Hilbert Space 3
1.1 Definition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.2 Physical interpretation . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.3 Dirac notation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.4 Tensor products of Hilbert spaces . . . . . . . . . . . . . . . . . . . . 6
1.5 Operators . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
1.5.1 Eigenstates and eigenvalues . . . . . . . . . . . . . . . . . . . 8
1.5.2 Operators in infinite dimensional Hilbert space . . . . . . . . 9
1.5.3 Operators for composite systems . . . . . . . . . . . . . . . . 9
1.6 The generalised uncertainty principle . . . . . . . . . . . . . . . . . . 10
1.7 Quantum harmonic oscillator . . . . . . . . . . . . . . . . . . . . . . 11

2 Translations and symmetries 14

2.1 Translations of states and operators . . . . . . . . . . . . . . . . . . 14
2.2 Spatial translations . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
2.3 Rotations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
2.3.1 Translations around a circle . . . . . . . . . . . . . . . . . . . 18
2.3.2 Spins . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
2.4 Symmetries and conservation laws . . . . . . . . . . . . . . . . . . . 20
2.5 Dynamics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
2.5.1 Conserved quantities . . . . . . . . . . . . . . . . . . . . . . . 22
2.5.2 Symmetries . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
∗ Based on the notes of David Skinner here

1
3 Angular momentum 24
3.1 Angular momentum eigenstates . . . . . . . . . . . . . . . . . . . . . 24
3.2 Rotations and orientation . . . . . . . . . . . . . . . . . . . . . . . . 25
3.2.1 Rotation of molecules . . . . . . . . . . . . . . . . . . . . . . 26
3.3 Spins . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
3.3.1 Integer or half integer? . . . . . . . . . . . . . . . . . . . . . . 27
3.3.2 Spin matrices . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
3.3.3 The Sterm-Gerlach experiment . . . . . . . . . . . . . . . . . 28
3.3.4 Spin precession (Intermezzo) . . . . . . . . . . . . . . . . . . 29
3.4 Orbital angular momentum . . . . . . . . . . . . . . . . . . . . . . . 30
3.4.1 Spherical harmonics . . . . . . . . . . . . . . . . . . . . . . . 31
3.5 Central potentials . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
3.5.1 The 3D quantum harmonic oscillator . . . . . . . . . . . . . . 32
3.5.2 Hydrogen atom . . . . . . . . . . . . . . . . . . . . . . . . . . 34

4 Addition of angular momentum 36

4.1 Combining two states . . . . . . . . . . . . . . . . . . . . . . . . . . 36
4.2 Identical particles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
4.3 Ideal quantum gas . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
4.4 Exchange and parity . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
4.4.1 Parity of pion . . . . . . . . . . . . . . . . . . . . . . . . . . . 41

5 Perturbation theories 43
5.1 Time-independent perturbation theory . . . . . . . . . . . . . . . . . 43
5.2 Degenerate perturbation theory . . . . . . . . . . . . . . . . . . . . . 45
5.2.1 The Stark effect . . . . . . . . . . . . . . . . . . . . . . . . . 45
5.3 Time-dependent perturbation theory . . . . . . . . . . . . . . . . . . 46
5.4 Ionisation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48

6 Interpreting quantum mechanics 50

6.1 The density operator . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
6.1.1 The Bloch ball . . . . . . . . . . . . . . . . . . . . . . . . . . 51
6.1.2 No-cloning theorem (intermezzo) . . . . . . . . . . . . . . . . 52
6.2 Entropy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
6.3 Entanglement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
6.4 Bell’s inequalities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54

2
Postulates of quantum mechanics
I. The states of a system are vectors in a Hilbert space.

II. Observable quantities are Hermitian linear operators.

III. Time evolution form of Schrödinger equation: iℏ∂t (ψ) = H |ψ⟩

1 Hilbert Space
1.1 Definition
Definition (Hilbert space). A Hilbert Space H is a vector space over C where
• For ϕ, ψ ∈ H, aϕ + bψ = bψ + aϕ ∈ H ∀a, b ∈ C;
• ϕ + (ψ + ξ) = (ϕ + ψ) + ξ;

• ∃0 ∈ H s.t. ϕ + 0 = ϕ ∀ϕ ∈ H.
H is equipped with an inner product ( , ) : H × H → C where
• (ϕ, ψ) = (ψ, ϕ);
• (ϕ, aψ) = a(ϕ, ψ) ∀a ∈ C but (aϕ, ψ) = a(ϕ, ψ);

• (ϕ, ϕ) ⩾ 0 with equality if and only if ϕ = 0.

p
The inner product gives a norm ∥ψ∥ = (ψ, ψ). This satisfies the Cauchy-Schwarz
inequality: |(ϕ, ψ)| ⩽ ∥ϕ∥∥ψ∥.
H is complete with this norm: a convergent sequence {ϕn } converges in H, so we
can take limits and derivatives.

1.2 Physical interpretation

Each ϕ ∈ H represents a physical state:
• ϕ and λϕ represent the same physical state ∀λ ∈ C, λ ̸= 0. States correspond
to “rays” in H.
• Sometimes it is useful to work with normalised vectors: (ϕ, ϕ) = 1.

• If ϕ, ψ ∈ H are normalised, the inner product is called the probability ampli-

tude (∈ C) to transition from the state ϕ to ψ.
• All the “weirdness” of QM comes from this: alive + dead = valid state

3
Example 1.1 (Finite-dimensional Hilbert space). H ∼
= Cn for some n ∈ Z. Then
n
X
(V, U ) = V i Ui , ∀u, v ∈ Cn
i=1
Pn
and has norm ∥V ∥2 = i=1 |Vi |2 .

Example 1.2 (Infinite-dimensional Hilbert space). We have dim(H) = ∞. All

such space in QM are isomorphic to l2 , the spaceP of all infinite sequences U =
(u1 , u2 , · · · ) of complex numbers such that ∥U ∥2 = |Ui |2 < ∞.
• The inner product (V, U ) = V i Ui < ∞ by Cauchy-Schwarz.
P

• For example, consider a particle on a circle of radius 2π, then

X Z 2π ∞
X
inθ 2 2
ψ(θ) = an e ⇒ ∥ψ∥ = |ψ(θ)| dθ = 2π |an |2
n 0 n

• Another example is the bound states of a harmonic oscillator with

∞
X 2
ψ(x) = an Hn (x)e−x α

n=0

where Hn are the Hermite polynomials.

• The set of functions L2 (R, dx) ∼
= l2 . The inner product
Z
(ϕ, ψ) = ψ(x)ϕ(x)dx < ∞
R

Definition (Dual space). The dual H∗ of H is the space of linear maps f : H → C.

• For dim(H) < ∞, given ϕ ∈ H, we can always build a linear map fϕ : (ϕ, ).

• Theorem (IB Linear Analysis): If all f ∈ H∗ are of this type, then H∗ ∼

= H.
• ForRdim(H) = ∞ this is not true, for example the
R Dirac δ : H → C takes ψ(x)
to dxδ(x)ψ(x) = ψ(0) ∈ C. But ∥δ(x)∥2 = dxδ(x)δ(x) = δ(0), which is
undefined.

• For continuous maps it is true, see Riesz representation theorem from Part
II Analysis of Functions.

4
1.3 Dirac notation
Definition (Dirac notation). We’ll use the following notation from Dirac:

• If ψ ∈ H we place it inside a “ket” as |ψ⟩.

• If ϕ ∈ H∗ we place it inside a “bra” as ⟨ϕ|.
The inner product is ⟨ϕ | ψ⟩, which forms a bra-ket.
Given a basis |n⟩ for n = 1, 2, · · · and dim H = N < ∞, we write
• Vectors: |V ⟩ = n Vn |n⟩ where V = (v1 , · · · , vN )T ;
P

• Linear maps: ⟨U | =
P ∗
Um ⟨n| where U = (u∗1 , · · · , u∗N );
PN
• Inner products are written as matrix multiplications: ⟨U |V ⟩ = n=1 Un∗ Vn .

[Lecture 1 finish]
P
Usually, we expand in countable, orthonormal basis {ϕn } ∈ H s,t, |ψ⟩ = n an |ϕn ⟩
and ⟨ϕn |ϕm ⟩ = δnm . Sometimes we generalise this to continuum states (see II
Analysis of Functions for formal definition).
For x ∈ R let |x⟩ obey X̂ |x⟩ = x |x⟩ and ⟨x′ |x⟩ = δ(x − x′ ), where X̂ is the
position operator. Then we can expand any vector as
Z
|ψ⟩ = ψ(x′ ) |x′ ⟩ dx′
R

where ψ(x′ ) is the wavefunction in the position basis.

We can extract the wavefunction by projecting on |x⟩:
Z Z
′
⟨x|ψ⟩ = ψ(x ) ⟨x|x ⟩ dx = ψ(x′ )δ(x − x′ ) = ψ(x)
′ ′

Normalisation ensures that

Z Z
⟨ϕ|ψ⟩ = ϕ(x′ )ψ(x) ⟨x′ |x⟩ dxdx′ = ϕ(x)ψ(x)dx

?
Note that ⟨x|x⟩ = δ(0), so continuum states are not normalisable and so they are
not in L2 (R, dx).
Remark. The importance of using bra-ket notation is that it frees us from having
to pick a basis.

5
We could have equally well chosen momentum states P |p⟩ = p |p⟩, where P is the
momentum operator and p is real. Then
Z Z
|ψ⟩ = ψ̃(p) |p⟩ dp = ψ(x |x⟩ dx

where ψ̃(p) is the wavefunction in the continuum basis |p⟩. We’ll show later that
1
⟨x|p⟩ = eipx/ℏ √ = ⟨p|x⟩
2πℏ

ψ̃(p)eipx/ℏ √dP
R
Therefore, ψ(x) = ⟨x|ψ⟩ = 2πℏ
.

1.4 Tensor products of Hilbert spaces

More complicated systems require “longer” Hilbert spaces. We can build these
by taking tensor products. Let dim(H1 ), dim(H2 ) < ∞ and let {|ea ⟩} ∈ H1 and
{|fa ⟩} ∈ H2 be orthonormal basis. Then
X
H = H1 ⊗ H2 ∋ |ψ⟩ = caα |ea ⟩ ⊗ |fα ⟩
a,α

Note that |ψ⟩ =

̸ |ϕ⟩ ⊗ |ψ⟩. The inner product is defined on basis elements:

(⟨ea | ⊗ ⟨fα |)(|eb ⟩ ⊗ |fβ ⟩) = ⟨ea |eb ⟩ ⟨fα |fβ ⟩ ∈ C

Then extend to H by linearity.

If |x⟩ and |y⟩ are continuum states for a particle travelling in the x and y directions.
The state |ψ⟩ can be written as
Z
|ψ⟩ = ψ(x, y) |x⟩ ⊗ |y⟩ dxdy

Again note that ψ(x, y) ̸= ψ(y, x) in general. Again the inner product is
Z Z
′ ′ ′ ′ ′ ′
⟨ϕ|ψ⟩ = ϕ(x , y )ψ(x, y) ⟨x |x⟩ ⟨y |y⟩ dx dxdy dy = ϕ(x, y)ψ(x, y)d2 x

For particles in R3 we’ll use the short hand notation:

Z
|ψ⟩ = ψ(x) |x⟩ d3 x

where ψ(x) = ψ(x, y, z) and |x⟩ = |x⟩ ⊗ |y⟩ ⊗ |z⟩.

6
Example. A Hydrogen atom consists of negative electrons (e− ) and positive pro-
tons (p+ ), so we have
Z
|ψ⟩ = ψ(xe , xp ) |xe ⟩ ⊗ |xp ⟩ d3 xe d3 xp ∈ L2 (R6 , d3 xe , d3 xp )

me xe +mp xp
Remark. We can introduce xrel = xe − xp and xcm = me +mp . Then expand:
Z
|ψ⟩ = ψ̃(xcm , xrel ) |xcm ⟩ ⊗ |xrel ⟩ d3 xcm d3 xrel

This can be convenient if there are no external forces, we also have

ψ̃(xcm , xrel ) = eik·xcm ψ̃(xrel )

For an N -particle system the Hilbet space is

L2 (R3N , d3N x) = L2 (R3 , d3 x) ⊗ L2 (R3 , d3 x) ⊗ · · · ⊗ L2 (R3 , d3 x)

| {z }
N times

Spin means electrons are actually described by a pair of wavefunctions:

|ψ⟩ ∈ L2 (R3 , d3 x) ⊗ C2 , ψ(x = (ψ↑ (x), ψ↓ (x))

Proposition (Postulate I of quantum mechanics).

• The state of a system is a vector |ψ⟩ ∈ H, |ψ⟩ =
̸ |0⟩.
• Any complete set of orthonormal states {|ψn ⟩} is in 1−1 correspondence with
possible outcome of a measurement, see Postulate II.
• The probability to observe outcome corresponding to |ϕn ⟩ is

| ⟨ϕn |ψ⟩ |2
Prob(|ψ⟩ → |ϕn ⟩) = ∈R
⟨ϕn |ϕn ⟩ ⟨ψ|ψ⟩
also known as the Born rule.

[Lecture 2 finish]

1.5 Operators
Definition (Linear operators). For dim H = n < ∞, a linear operator is an n × n
matrix A : H → H. Operators form an associative but not commutative algebra
over C called the operator algerba.

(αA + βB) : |ψ⟩ → αA |ψ⟩ + βB |ψ⟩ ∀α, β ∈ C, |ψ⟩ ∈ H

7
Definition (Product of operators). AB : |ψ⟩ → A(B |ψ⟩).
Definition (Commutator of operators). The commutator of A, B is

[A, B] = AB − BA

This obeys:
• Anti-symmetry: [A, B] = −[B, A]
• Linearity: [αA + βB, C] = α[A, C] + β[B, C]
• Leibniz rule: [A, BC] = [A, B]C + B[A, C]
• Jacobi identity: [A, [B, C]] + [B, [C, A]] + [C, [A, B]] = 0
• The inner product on H allows us to define the adjoint A† of A by
∗
⟨ϕ|A† |ψ⟩ = ⟨ψ|A|ϕ⟩ ∀ |ϕ⟩ , |ψ⟩ ∈ H
+
This satisfies (A + B)+ = A† + B + , (AB)+ = B + A† , (A† ) = A.

1.5.1 Eigenstates and eigenvalues

Definition (Eigenstates and eigenvalues). A state (vector) |ψ⟩ is an eigenstate of
A if A |ψ⟩ = a |ψ⟩ for some a ∈ C. Then a is called the eigenvalue. If there exists
a unique eigenvector for each eigenvalue, i.e. dim(ker(A − aI)) = 1, then we can
label eigenvectors by its eigenvalue: A |a⟩ = a |a⟩. Otherwise, we say the eigenvalue
has degenrate eigenvectors.
Proposition. If Q = Q+ (Hermitian) and Q |q⟩ = q |q⟩ then q ∈ R.
∗
Proof. We have q ⟨q|q⟩ = ⟨q|Q|q⟩ = ⟨q|Q|q⟩ = q ∗ ⟨q|q⟩ i.e. q = q ∗ .
Proposition. If Q = Q+ and ⟨Q|q1 ⟩ = ⟨q1 |q1 ⟩, ⟨Q|q2 ⟩ = ⟨q2 |q2 ⟩ with q1 ̸= q2 ,
then ⟨q2 |q1 ⟩ = 0.
∗
Proof. We have 0 = ⟨q2 |Q − Q+ |q1 ⟩ = ⟨q2 |Q|q1 ⟩−⟨q1 |Q|q2 ⟩ = (q1 −q2 ) ⟨q2 |q1 ⟩.
Proposition. If Q = Q+ , then Q is diagonalisable, so eigenvectors span H.
Definition (Basis of eigenvectors). If all eigenvectors are non-degenerate
P with
Q |n⟩ = qn |n⟩, we can use {|n⟩} to form an orthonormal basis: Q = n qn |n⟩ ⟨n|
so that ! !
X X X
′
Q |ψ⟩ = qn |n⟩ ⟨n| ψn′ |n ⟩ = qn ψn |n⟩
n n′ n
P
The identity operator on H is then 1 = n |n⟩ ⟨n|.

8
 P
Definition. f (Q) = n f (qn ) |n⟩ ⟨n| defines f (Q) if f (ψn ) exists.
Matrix elements of A inP this basis are Ann′ = ⟨n|A|n′ ⟩. Also, (AB)nn′ = ⟨n|AB|n′ ⟩ =
′
P
k ⟨n|A|k⟩ ⟨k|B|n ⟩ = k Ank Bkn . Also, recall that adjoints are transpose of con-
′
∗
jugates, so (A† )nn′ = ⟨n|A† |n′ ⟩ = ⟨n′ |A|n⟩ = (An′ n )∗ .
When eigenvectors of Q are degenerate, we look for a “complete set of commuting
observables” {Q, P, · · · } s.t. Q+ = Q and [Q, P ] = 0 = [P, Q] ∀Q, P . Then they
are simultaneously diagonalisable (in some basis), then form the set {|qn , pm , · · ·⟩}.

1.5.2 Operators in infinite dimensional Hilbert space

Let’s mention and ignore some issues in dim H = ∞ such as L2 (R, dx).
Definition. An operator A is bounded if there exists M ∈ R s.t. ∥A |ψ⟩∥ < M ∥|ψ⟩∥
for all |ψ⟩ ∈ H.
Note. In finite dimensional Hilbert spaces, all operators are bounded.
Example. The position operator X given as X |ψ⟩ = xψ(x) and momentum op-
erator P given as P |ψ⟩ = −iℏ∂x ψ(x) are unbounded.
When A is unbounded, it is only defined in dom(A) ⊂ H, its domain.

1.5.3 Operators for composite systems

Definition. Let {|ea ⟩} be a basis of H1 and A : H1 → H1 a linear operator.
Similarly for {|fa ⟩} and B for H2 . We define A ⊗ B : H1 ⊗ H2 → H1 ⊗ H2 by

(A ⊗ B)(|ea ⟩ ⊗ |fa ⟩) = (A |ea ⟩) ⊗ (B |fa ⟩)

on a basis and by linearity on the whole of H1 ⊗ H2 .

Note that [A ⊗ 1H1 , 1H2 ⊗ B] = 0 ∀A, B.

[Lecture 3 finish]

For example, Hydrogen is described by the Hamiltonian:

P2p P2 q2 1
H= ⊗ 1e + 1p ⊗ e −
2mp 2me 4πε0 |xe − xp |
P2cm
 2
q2

Prel 1
= ⊗ 1rel + 1cm ⊗ −
2M 2µ 4πε0 |xrel |
me mp
where µ = me +mp is the reduced mass and M = me + mp . Also Pcm = Pp + Pe
mp Pe −me Pp
and the relative momentum is Prel = M .

9
Proposition (Postulate II of quantum mechanics). Observable quantities are rep-
resented by Hermitian (linear) operators on H. The “expectation value”

⟨ψ|Q|ψ⟩
⟨Q⟩ψ =
⟨ψ|ψ⟩

gives the property corresponding to the observable Q in the state |ψ⟩.

Remark. To build a quantum theory, we can:
• Start from H and define observables as operators.
• Start from operator algebra and define H as a representation.
Proposition (Postulate III of quantum mechanics). The dynamical evolution of a
quantum system is governed by the time-dependent Schrödinger equation (TDSE):

∂
iℏ |ψ⟩ = H |ψ⟩
∂t
for some Hamiltonian operator H.
Remark.
• Time in Hilbert space is viewed by the change in vector |ψ⟩ in H, and H does
not involve time.
• The form of H depends on the system, often H = H(X, P, · · · ).

1.6 The generalised uncertainty principle

Definition. The root mean squared deviation or uncertainty ∆Q of Q = Q+ in
state |ψ⟩ is q
2
∆ψ Q = ⟨Q2 ⟩ψ − ⟨Q⟩ψ

Since Q is Hermitian,
2
⟨Q2 ⟩ψ − ⟨Q⟩ψ = ⟨ψ|(Q − ⟨Q⟩ψ )2 |ψ⟩
= ∥(Q − ⟨Q⟩ψ )| |ψ⟩∥2 by Hermitian
⩾0

So ∆ψ Q is well-defined and vanishes if and only if Q |ψ⟩ = q |ψ⟩ for some p.

Definition. For two Hermitian operators A, B, define

|ψA ⟩ = A |ψ⟩ − ⟨A⟩ψ |ψ⟩ , |ψB ⟩ = B |ψ⟩ − ⟨B⟩ψ |ψ⟩

10
By Cauchy-Schwarz,
2
(∆ψ A)2 (∆ψ B)2 = ∥|ψA ⟩∥2 ∥|ψB ⟩∥2 ⩾ | ⟨ψA |ψB ⟩
Then we have ∥|ψA ⟩∥ = ∆ψ A and
⟨ψA |ψB ⟩ = ⟨ψ|(A − Aψ )(B − Bψ )|ψ⟩ = ⟨ψ|AB|ψ⟩ − ⟨A⟩ψ ⟨B⟩ψ
Hence
∗
2iIm(⟨ψA |ψB ⟩) = ⟨ψ|AB|ψ⟩ − ⟨ψ|AB|ψ⟩ = ⟨ψ|[A, B]|ψ⟩
Again by Cauchy-Schwarz,
1
∆p siA∆ψ B = ∥|ψA ⟩∥∥|ψB ⟩∥ ⩾ | ⟨ψA |ψB ⟩ ⩾ ⟨[A, B]⟩ψ
2
This is the generalised uncertainty principle.

1.7 Quantum harmonic oscillator

Harmonic oscillators are ubiquitous in physics. The Hamiltonian for QHO in 1D
is
P2 mω 2 2
H= + X
2m 2
Definition (Lowering and raising operators). The lowering operator is
1
A= √ (mωX + iP )
2mωℏ
and the raising operator is its adjoint:
1
A† = √ (mωX − iP )
2mωℏ
Since
1
A† A = (mωX − iP )(mωX + iP )
2mωℏ
1
= (P 2 + m2 ω 2 X 2 +imω [X, P ])
2mωℏ | {z } | {z }
2mH iℏ
H 1
= −
ℏω 2
The Hamiltonian becomes
   
1 1
H = ℏω A† A + = ℏω N +
2 2
where N = A† A is the (Hermitian) number operator.

11
Start from operator algebra {X n P m } and the commutation [X, P ] = iℏ (recall IB
QM). In terms of A† , A− ,

1 −imω
[A, A† ] = [mωX + iP, mωX − iP ] − ([X, P ] − [P, X]) = 1
2mℏω 2mℏω
Also, [A, A] = [A† , A† ] = 0. Consequently

[N, A† ] = [A† A, A] = A† [A, A† ] + [A† , A† ]A† = A†

and similarly [N, A] = A.

[Lecture 4 finish]

Let’s build a Hilbert space that represent this operator algebra: Let |n⟩ be an
eigenvector of N with eigenvalue N |n⟩ = n |n⟩ and ⟨n|n⟩ = 1. Then consider
A† |n⟩, we know that

N (A† |n⟩) = [N, A† ] + A† N |n⟩ = (A† + nA† ) |n⟩ = (n + 1)A† |n⟩

and similarly N (A |n⟩) = ([N, A] + AN ) |n⟩ = (n − 1)A |n⟩. Provided that they
don’t vanish, A† |n⟩ and A |n⟩ are also new eigenvectors of N with eigenvalues n±1.
The question is, what can n ∈ R be? Note that

n = n ⟨n|n⟩ = ⟨n|N |n⟩ = (⟨n| A† )(A |n⟩) = ∥A |n⟩∥2 ⩾ 0

The condition n ⩾ 0 requires that there exists nmin s.t. A |nmin ⟩ = 0. But this
requires nmin = 0. Since if A |n⟩ =
̸ 0 then it is an eigenstate with eigenvalue n − 1,
for a minimum of 0 we conclude that n is a non-negative integer. For normalisation,
let A† |n⟩ = cn |n + 1⟩ for some cn , so

|cn |2 = ∥A† |n⟩∥2 = ⟨n|AA† |n⟩ = ⟨n|N + [A, A† ]|n⟩ = n + 1

√
Hence we choose cn = n + 1.
In summary, we built the Fock Hilbert space: H = Span{|0⟩ , |1⟩ , · · · }.
Definition. The ground state |0⟩ is defined to satisfy A |0⟩ = 0. Normalised energy
H- or number N -eigenstates satisfy
1
|n⟩ = √ (A† )n |0⟩
n!
with energy levels  
1
⟨n|H|n⟩ = ℏω n + , n ∈ N0
2

12
We can derive the wavefunctions for energy eigenstates:
Definition (Position-space wavefunction). Let ψ0 (x) = ⟨x|0⟩ be position-space
wavefunctions of the ground state, then
0 = ⟨x|A|0⟩ ∝ ⟨x|mωX + iP |0⟩ = mωXψ0 (x) + ℏψ0′ (x)
Solving this first order ODE gives
 mω 1/4 h mω i
ψ0 (x) = exp − x2
πℏ 2ℏ
This is then normalised such that ⟨0|0⟩ = R dx|ψ0 (x)|2 = 1.
R

Higher energy states can be obtained just by differentiating:

√
ψ1 (x) = ⟨x|1⟩ = ⟨x|A† |0⟩ = [mωXψ0 (x) − ℏψ0′ (x)] 2mωℏ ∝ 3Xe−mω/2ℏ x2
Repeated application of this trick generates the Hermite polynomials Hn (x).
The uncertainty principle tells us that
ℏ ℏ
∆ψ X∆ψ P ⩾ [X, P ]ψ =
2 2
for all |ψ⟩ ∈ H. Let’s compute this in quantum harmonic oscillator ground state.
Recall
2
(∆ψ X)2 = ⟨X 2 ⟩ψ − ⟨X⟩ψ
q
In terms of raising and lowering operators, X = 2mωℏ
(A + A† ). In any eigenstate,

⟨n|X|n⟩ ∝ ⟨n|A + A† |n⟩ = 0

Also,
 
2mω
⟨0|X 2 |0⟩ = ⟨0|(A + A† )2 |0⟩ = ⟨0|2
A + AA† + 
A† A†
A + A† |0⟩
 
ℏ
= ⟨0|AA† |0⟩ = ∥|1⟩∥ = 1
( = ⟨0|[A, A† ] + N |0⟩ = 1)
mℏω
Similarly we find ⟨0|P |0⟩ = 0 and ⟨0|P 2 |0⟩ = 2 . Combining these:
p ℏ
∆0 X∆0 P = ⟨0|X 2 |0⟩ ⟨0|P 2 |0⟩ =
2
So it saturates the uncertainty principle. Also other states have minimum uncer-
tainty.
[Lecture 5 finish]

13
2 Translations and symmetries
2.1 Translations of states and operators
Consider transformations of the state R of a system to state R′ that conserve
probabilities: Prob(ϕ → ψ) = Prob(ϕ′ → ψ ′ ).
Theorem (Wigner theorem). In quantum mechanics, these transformations are
represented by a linear and unitary operator |ψ ′ ⟩ = U |ψ⟩ with U † = U −1 , or an
anti-linear and anti-unitary operator, which we will not discuss in this course.
The operator U : H → H satisfies U |ψ⟩ = |ψ ′ ⟩. Hence ⟨ψ ′ | = ⟨ψ| U † . Unitarity
U † = U −1 ensures that

Prob(ϕ → ψ) = ∥⟨ϕ|ψ⟩∥2 = ∥⟨ϕ|U + U |ψ⟩∥2 = ∥⟨ϕ′ |ψ ′ ⟩∥2 = Prob(ϕ′ → ψ ′ )

These tramsformations form a group G, so U is a “homomorphism” from G to

linear operators on H, in a sense that

U : g1 ∈ G 7→ U (g1 ) ∈ G

Moreover G has the group structure such that

U (g2 ◦ g1 ) = U (g2 )U (g1 ) ∀g1 , g2 ∈ G

Also U (e) = 1H . In mathematical term we study the unitary representations of G

on H. We can equivalently transform operators instead of states. Let A : H → H
with expectation ⟨ψ|A|ψ⟩. After the transformation, we should have

⟨ψ ′ |A|ψ ′ ⟩ = ⟨ψ|U † AU |ψ⟩

Definition (Similarity transformation). Alternatively, the transformation acts on

A as:
A 7→ A′ = U † AU with states unchanged
This is called a similarity (or conjugation) transformation.

Note. We have A′ B ′ = (U + AU )(U + BU ) = U + (AB)U . Also [A′ , B ′ ] = U + [A, B]U .

Similarity transformations preserve the spectrum: If A |α⟩ = α |α⟩ then U † |α⟩ is
the eigenvector of A+ = U + AU with eigenvalue α.
Definition. Let U depend smoothly on a parameter θ ∈ R with U (θ = 0) = 1H .
For δθ << 1 we expand U (δθ) = 1H −iδθT +O(δθ2 ). Here T is called the generator
of the transform.

14
U † = U −1 implies that T † = T i.e. T is Hermitian. To see this, note that
1 = U † U = (1 − iδθT )† (1 − iδθT + · · · )
= (1 + iδθT † )(1 − iδθT + · · · )
= 1 + iδθ(T † − T ) + O(δθ2 )
which implies T † = T .
Consider the effect on states:
|ψ ′ ⟩ = U |ψ⟩ ≈ (1 − iδθT + · · · ) |ψ⟩ ⇒ δ |ψ⟩ = −iδθT |ψ⟩
On operators, we have:
A′ = U † AU = (1 + iδθT )A(1 − iδθT ) + · · · ⇒ δA = A′ − A = iδθ[A, T ]
We can write finite transformations by exponentiation of infinitesimal ones:
 N
θ
U (θ) = lim lim 1 − i T = e−iθT
N →∞ N →∞ N

2.2 Spatial translations

Example (Translations in R3 ). Translating by a ∈ R3 is represented by U (a).
Translations commute with each other, forming an Abelian group:
U (a)U (b) = U (a + b) = U (b + a) = U (b)U (a) and U † (a) = U (−a)
For infinitesimal transformation δa,
i
U (δa) = 1 − δa · P + O(δa2 ), P = (Px , Py , Pz )
ℏ
and therefore
U (a) = e−ia·P/ℏ
By definition, P/ℏ are generators of translations.
Note that since U (δa)U (δb) = U (δb)U (δa), this suggests that [Pi , Pj ] = 0 ∀i, j.
Proposition (Canonical commutation relations). If |ψ ′ ⟩ = U (a) |ψ⟩ then we must
have ⟨X⟩ψ′ = ⟨ψ|U † (a)|ψ⟩ = ⟨x⟩ψ + a. Since this must be true for all |ψ⟩,

U † (a)XU (a) = X + a1H

Again from an infinitesimal perspective,
(1 + iδa · P/ℏ)X(1 − iδa · P/ℏ) + O(δa2 ) = X + δa
To O(δa), we obtain the result:
Xi Pj − Pj Xi = [Xi , Pj ] = iℏδij

15
This operator algebra requires dim H = ∞, since Tr(1H ) = dim H = Tr(− ℏi [X, P ])
which is zero unless the trace is undefined, i.e. dim H = ∞.
[Lecture 6 finish]
How does U (a) act on wavefunctions?
• On position eigenstates |x⟩, X |x⟩ = x |x⟩, we translate |x⟩ → |x′ ⟩ = U (a) |x⟩.
But then
X(U (a) |x⟩) = ([X, U (a)] + U (a)X) |x⟩ = (a + x)(U (a) |x⟩)
Where the second equality follows since U † (a)XU (a) = X + a1H and mul-
tiplying U (a). Consequently U (a) |x⟩ = c |x + a⟩, but we can just fix c = 1
(see example sheet). Let ψ(x) = ⟨x|ψ⟩ be the wavefunction in R3 . After a
translation,
∗ ∗
ψtransl (x) = ⟨x|U (a)|ψ⟩ = ⟨ψ|U † (a)|x⟩ = ⟨ψ|x − a⟩ = ⟨x − a|ψ⟩ = ψ(x − a)

• For infinitesimal translations, the change is

ψ(x − δa) − ψ(x) = −δa · ∇ψ
Whilst:
ψtransl (x) − ψ(x) = ⟨x|1 − iδa · P/ℏ|ψ⟩ = ⟨x|ψ⟩
i
= − ⟨x|δa · P|ψ⟩
ℏ
Comparing, we find
⟨x|P |ψ⟩ = −iℏ∇ψ(x)
Example. Consider translation eigenstate |p⟩ obeying P |p⟩ = p |p⟩. After a
translation,
ψp (x − a) = ⟨x|U (a)|p⟩ = ⟨x|e−ia·P/ℏ |p⟩ = e−ia·p/ℏ ⟨x|p⟩ = eia·p/ℏ ψp (x)
where the third equality follows since |p⟩ is an eigenstate of P. It has solution
1
ψp (x) = eip·x/ℏ
(2πℏ)3/2
where the coefficient is for normalisation such that ⟨p′ |p⟩ = δ (3) (p′ − p).
Example. The ground state |0⟩ of the quantum harmonic oscillator in 1D has
2
x2 /2
ψ(x) = ⟨x|0⟩ = Ce−α
The translated state U (a) |0⟩ is called a coherent state:
2
(x−a)2 /2
ψacoherent (x) = ⟨x|U (a)|0⟩ = Ce−α
There is some variance as before, but centered at x = a. This is not on an energy
eignstate, so it will oscillate.

16
2.3 Rotations
An anti-clockwise rotation by an amount |α| around the α̂-axis corresponds to
a matrix R(α) : v → v′ = R(α)v. Since v′ · v′ = v · v and det R(α) = 1, we
can see that they form a rotation group SO(3). SO(3) is a non-Abelian group:
R(β)R(α) ̸= R(α)R(β) in general.
For infinitesimal rotations,
R(δα)v = v + δα × v + O(|δα|2 )
α

v′
v

Small rotations around δβ × δα give

R(δα)R(δβ)v = R(δα)(v + δβ × v) + O(|δβ|2 )
= (v + δα × v) + δα × (δβ × v) + O(|δα|2 )
Using the identity of a × (b × c) = b(a · c) − c(a · b), the difference is
[R(δα), R(δβ)]v = δα × (δβ × v) − δβ × (δα × v) + O(δα2 , δβ 2 )
= (δα × δβ) × v + · · ·
This is just R(δα × δβ) − 1R3 to the lowest order.
Definition (Rotation operator). Let U (α) be the corresponding unitary operator
in H. Let
i
U (δα) = 1 − δα · J + O(|δα|2 )
ℏ
where J/ℏ are generators of rotations with J = (Jx , Jy , Jz ). For finite rotations,
U (α) = e−iα·J/ℏ
The relation
[R(δα), R(δβ)]v = (δα × δβ) × v
implies
[U (δα), U (δβ)] = U (δα × δβ) − 1H
Solving this constraint for the generator gives
[Ji , Jj ] = iℏεijk Jk
[Lecture 7 finish]

17
If U (α) really describes rotations, then certainly,

U † (α)XU (α) = R(α)X

Taking infinitesimal δα, this is

   
δα δα
1+i ·J X 1−i · J = X + δα × X
ℏ ℏ
i
⇒ [δα · J, X] = δα × X to first order in δα
ℏ
⇒ [Ji , Xj ] = iℏεijk Xk

Any operator V transforms as a vector under rotations if

U † (α)VU (α) = R(α)V ⇔ [Ji , Vj ] = iℏεijk Vk

Example. If V, W transform as vectors under rotations, then

(V · W)V − V × W → U † (α)[(V · W)V − V × W]U (α)

= U † (α)VU (α) · U † (α)WU (α) U † (α)VU (α) − U † (α)VU (α) × U † (α)WU (α)




= (R(α)V) · (R(α)W) R(α)V − (R(α)V) × (R(α)W)
= (V′ · W′ )V′ − (V′ × W′ )

Definition. An operator S transforms under rotations as a scalar if

U † (α)SU (α) = S ⇒ [Ji , S] = 0

Example. Recall IB Quantum Mechanics, we have L = X × P. This transforms

as a vector under rotations. Therefore,

[Ji , Lj ] = iℏεijk Lk

Also, [Li , Lj ] = iℏεijk Lk , [Li , Xj ] = iℏεikj Xk , [Li , Pj ] = iℏεijk Pk and similarly for
the J’s. You may prove this or review notes from IB.

2.3.1 Translations around a circle

However, in a composite system J ̸= L. To see the relation, consider a succession
of translations around an N -sided regular polygon:

18
 2π
a= N n ×X
x
2π
N

As N → ∞, we get translations around a circle. Then

U † (δa)XU (δa) = X + δa
i.e.
   
δa δa
1+i ·P X 1−i · P = X + δa
ℏ ℏ
So
i 2π i 2π
[δa · P, X] = δa ⇒ [(n × X) · P, X] = n×X
ℏ N ℏ =n·(X×P) N
In other words,
[Li , Xj ] = iℏεijk Xk
So how do L and J differ? L generates translations around circular paths. exp(−iα·
L/ℏ) preserves the “heading” of object, but exp(−iα · J/ℏ) rotate the object itself.

2.3.2 Spins
Definition. The spin operator is defined as
S=J−L
i.e. exp(−iα · S/ℏ) spins the object itself.
Maybe J = particles a X(a) × P(a) ? The real distinction is, between which Hilbert
P

space do J and L act on. If H = L2 (R3 , d3 x), then J, L are identical. However, do
we really know which Hilbert space to use to describe, say, one electron? This is
no longer a mathematical problem, but something that requires experiments.
To see that S only acts on the internal state of a quantum system, compute
[Si , Xj ] = [Ji − Li , Xj ] = [Ji , Xj ] − [Li , Xj ]
= iℏεijk Xk − iℏεijk Xk = 0

19
Similarly,
[Si , Pj ] = [Ji − Li , Pj ] = [Ji , Pj ] − [Li , Pj ] = 0
However,

[Si , Sj ] = [Ji − Li , Jj − Lj ]
= [Ji , Jj ] − [Ji , Lj ] − [Li , Jj ] + [Li , Lj ]
= iℏεijk Jk − iℏεijk Lk + iℏεjik Lk + iℏεijk Lk
= iℏεijk (Jk − Lk )

So it also obeys SO(3) algebra, confirming that S as being some rotation.

Definition. The point inversion/reflection through 0 ∈ R3 P : x → −x again gives
a unitary operator Π : H → H, but because Z2 is discreet, there is no generator
for this.
We have
Π2 = idH
It follows that Π has eigenvalues ±1. Moreover, Π† = Π−1 = Π, and

Π† XΠ = −X ⇒ Π2 XΠ = −ΠX ⇒ 0 = XΠ + ΠX = {X, Π}

[Lecture 81 finish]

2.4 Symmetries and conservation laws

If the Hamiltonian H does not depend on time explicitly i.e. ∂t H = 0, the
Schrödinger equation is solved by

|ψ(t)⟩ = U (t) |ψ(0)⟩ , U (t) = e−iHt/ℏ

Note that U † = U −1 . Indeed,

 
H
iℏ∂t |ψ(t)⟩ = iℏ −i |ψ(t)⟩ = H |ψ(t)⟩
ℏ

as required. U is the unitary time-evolution operator. Since U (∆t) = 1 − i H

ℏ ∆t +
O(∆t2 ), the Hamiltonian is the generator of time-translations:

time
−−−−−→
|ψ(0)⟩ |ψ(t)⟩
1 Lecture given by Professor David Skinner

20
This is known as the Schrödinger picture: states or vectors evolve while operators
don’t.
Alternatively, define
|ψ⟩H ≡ |ψ(0)⟩S
Heisenberg Schrödinger

Definition. Define

QH (t) = U † (t)QS U (t), U (t) = e−iHt/ℏ , QS = Q − QH (0)

This gives a different but equivalent description of time evolution:

⟨Q⟩ψ =H ⟨ψ|QH (t)|ψ⟩H = S ⟨ψ(0)⟩ U † (t) QS U (t) |ψ(0)⟩S

| {z } | {z }
S ⟨ψ(t)| |ψ(t)⟩S

Differentiating QH (t) in time to obtain

dQH (t) i ∂QS
= U † (t)[H, QS ]U (t) + U † (t) U (t)
dt ℏ ∂t
where an explicit time-dependence of QS (t) is allowed. Simplifying:
dQH (t) i ∂QH
= [H, QH ] +
dt ℏ ∂t H

2.5 Dynamics
What is the Hamiltonian? For motion in R3 we choose H = H(X, P, · · ·). The
simplest choice is H = P2 /2m, which represents free particle. To model obsticles
we include a potential :
P2
H= + V (X)
2m
So in the Heisenberg picture,
dXH (t) PH (t) dPH (t)
= , = −∇V (t)
dt m dt
where ∇V (t) = U † (t)∇X V (X)U (t).
The translation generator P is now the momentum.
Definition. ⟨0, x0 ⟩ = e−ix0 P/ℏ |0⟩.
Example (Quantum harmonic oscillator). We compute 1D ⟨P ⟩ in a displaced
ground state of quantum harmonic oscillator. Solving the Heisenberg-picture equa-
tions,
PH (t) = U † (t)P U (t) = P cos(ωt) − mωX sin(ωt)

21
Then,

⟨P ⟩ = ⟨0, x0 |P (t)|0, x0 ⟩
= ⟨0|eix0 P/ℏ (P cos(ωt) − mωX sin(ωt)) e−ix0 P/ℏ |0⟩
= cos(ωt)⟨0|P |0⟩ − mω sin(ωt) × ⟨0|eix0 P ℏ ([X, e−ix0 P/ℏ ] + e−ix0 P/ℏ X)|0⟩
=0
−ix0 −ix0 P/ℏ
= −mω sin(ωt) ⟨0|eix0 P ℏ iℏ e |0⟩
ℏ
From Example Sheet 1, we know that [X, f (P )] = iℏf ′ (P ), so

⟨P ⟩ = −mω sin(ωt)x0

as expected from classical intuition.

2.5.1 Conserved quantities

Definition. If dQH (t)/dt = 0 in the Heisenberg picture, we say Q is conserved.
Suppose for simplicity that ∂t QS = 0. Then in the Heisenberg picture:

dQH (t) i i
= [H, QH (t)] = U † (t)[H, Q]U (t)
dt ℏ ℏ
Since U is unitary, dQH (t)/dt = 0 if and only if [H, Q] = 0. Hence conserved
quantities must commute with the Hamiltonian.
Example. Consider [H, Q] = 0 and Q has eigenstates |q⟩ at t = 0. At time t,
|q; t = 0⟩ becomes |q ′ ; t⟩ = U (t) |q, t = 0⟩. But

Q |q, t⟩ = QU (t) |q, t = 0⟩ = q |q, t⟩

Hence |q⟩ is a useful basis.

[Lecture 9 finish]

2.5.2 Symmetries
Symmetries lead to conserved quantities. Consider transformation U (θ) = e−iθT
so that H 7→ U † (θ)HU (θ).
Definition. A transformation is a symmetry if it leaves H unchanged i.e.

U † HU = H

or equivalently, [T, H] = 0. Infinitesimally,

22
 |ψ(0)⟩ time |ψ(t)⟩

U U†

|ψ ′ (0)⟩ |ψ ′ (t)⟩

This says the generator is a conserved quantity.

Example. If H(X, P, · · · ) is translation invariant then P is conserved. If it is

rotation invariant then J is conserved. Similarly for time for discrete symmetries
(e.g. parity).

23
3 Angular momentum
If L2 (R, dx) describes position, what Hilbert space describes angular momentum?
It must represent the SO(3) group of rotations, or equivalently the SO(3) algebra
of rotation generators J = (Jx , Jy , Jz ). Recall [Ji , Jj ] = iℏεijk Jk , we can build
finite-dimensional (unitary) representations (unlike X and P). If dim(H) < 0 then
0 = Tr([Ji , Jj ]) = iℏεijk Tr(Jk ). So J is represented by trace-less matrices.

3.1 Angular momentum eigenstates

How should we build H? We build a basis of eigenvectors of J. But Jx,y,z cannot be
simultaneously diagonalised because [Ji , Jj ] ̸= 0. However J2 = J·J = Jx2 +Jy2 +Jz2 ,
so

[Ji , J2 ] = [Ji , Jj ]Jj + Jj [Ji , Jj ] = iℏ(εijk Jk Jj + εijk Jj Jk ) = 0

Let |β, m⟩ obey

J2 |β, m⟩ = βℏ2 |β, m⟩ , Jz |β, m⟩ = ℏm |β, m⟩ , ⟨β ′ , m′ |β, m⟩ = δβ ′ β δm′ m

We’ll proceed as for the quantum harmonic oscillator. Define the new raising and
lowering operators as
J± = Jx ± iJy
then
[Jz , J± ] = [Jz , Jx ] ± i[Jz , Jy ] = iℏ(Jy ∓ iJx ) = ±ℏJ±
We have

J2 (J± |β, m⟩) − [J2 , J± ] + J± J2 |β, m⟩ = βℏ2 (J± |β, m⟩)

Jz (J± |β, m⟩) = ([Jz , J± ] + J± Jz ) |β, m⟩ = (m ± 1)ℏ(J± |β, m⟩)

As long as it’s not vanishing, J± |β, m⟩ is an eigenstate of J2 and J2 with eigenvalues

ℏ2 β and (m ± 1)ℏ, respectively.
To fix β, m we examine the norm:

∥J± |β, m⟩∥2 = ⟨β, m|J− J+ |β, m⟩ = ⟨β, m|(Jx − iJy )(Jx + iJy )|β, m⟩
= ⟨β, m|Jx2 + Jy2 + i[Jx , Jy ]|β, m⟩
= ⟨β, m|J2 − Jz2 − ℏJz |β, m⟩
= ℏ2 (β − m(m + 1)) ⩾ 0

J+ increases m while preserving β, so β = J(J + 1) for some J that is the largest

values of m:
∥J+ |β, J⟩∥2 = 0 ⇒ J+ |β, m⟩ = 0

24
States |β, J⟩ are called the largest weight states in II Representation Theory.
Similarly for ∥J− |β, m∥ = ℏ2 (β − m(m − 1)), we also need β = J ′ (J ′ − 1) where
J ′ is the minimum of m. Comparing β = J(J + 1) = J ′ (J ′ − 1), we find that
J ′ = −J or J ′ = J + 1. Since J ′ ⩽ J we must have J ′ = −J with J ⩾ 0 and
J − J ′ = 2J ∈ N0 because J± changes m by unit steps.
In summary:
• β = J(J + 1) where 2J ∈ N0 . We’ll just label our states as |j, m⟩.
• Hj = Span(|j, m⟩) where m ∈ (−J, −J + 1, · · · , J − 1, J).

• There are 2J + 1 basis so dimH = 2J + 1.

• We see that

Jz |j, m⟩ = ℏm |j, m⟩ , J2 |j, m⟩ = ℏ2 J(J + 1) |j, m⟩

and by assuming |J, m⟩’s are normalised, we conclude that

p
J+ |j, m⟩ = J(J + 1) − m(m + 1) |j, m + 1⟩
p
J− |j, m⟩ = J(J + 1) − m(m − 1) |j, m − 1⟩

And we can deduce that

J+ |j, J⟩ = 0, J− |j, −J⟩ = 0

[Lecture 10 finish]

Remark. Since J± = Jx ± iJy , it follows that J± maps Hj to Hj .

3.2 Rotations and orientation

Consider |j, J⟩, it has maximum angular momentum along ẑ. How much angular
momentum is there in x, y directions? Note that
1
⟨j, J|Jx,y |j, J⟩ = ⟨j, J|(J+ ± J− ) |j, J⟩ = 0
2
However,

⟨j, J|Jx2 + Jy2 |j, J⟩ ⟨j, J|J2 − Jz2 |j, J⟩ j(j + 1) − j 2 1

= = =
⟨j, J|Jz2 |j, J⟩ ℏ2 j 2 j2 j
√
So in |j, J⟩ roughl 1/ j of angular momentum is on the x-y plane.

25
To think about the classical limit (|J >> ℏ so J >> 1), let’s measure J along
n+(sin θ, 0, cos θ) in the state |j, J⟩. Classically we would find ℏjcosθ. In Quantum
Mechanics this is the same on average:

⟨j, J|n · J|j, J⟩ = ⟨j, J| cos θJz + sin θJx |j, J⟩ = ℏj cos θ

However the uncertainty is

2 2
⟨(n · J)2 ⟩ = cos2 θ ⟨Jz ⟩ + sin2 θ ⟨Jx ⟩ + sin θ cos θ ⟨
Jx
J
z +
Jzx⟩
J
1
= ℏ2 j 2 cos θ + sin2 θ ⟨J+ J− ⟩
 4 
2 2 j 2
= ℏ j cos θ + sin θ
2
p
So uncertainty is ∆jj (n · J) = ℏ| sin θ| j/2. In a maximally aligned state,
p
∆(n · J) ℏ| sin θ| j/2 1
= ∼√
⟨n · J⟩ ℏj cos θ j

3.2.1 Rotation of molecules

Consider an axisymmetric carbon-monoxide molecule. Then the moments of inertia
satisfy
I = Ix = Iy ̸= Iz
The Hamiltonian used to model its dynamics is then

Jx2 Jy2 Jz2 J2

 
2 1 1
H= + + = + Jz −
2Ix 2Iy 2Iz 2I 2Iz 2I

so |j, m⟩ is an energy eigenstate with

 
2 j(j + 1) 1 1
Ej,m = ℏ + m2 ℏ2 −
2I 2Iz 2I

Carbon-monoxide has Iz << I, so only m = 0 is accessible at low energies. Thus

the energies of the photons that can be emitted or absorbed by a rotating molecule
are
jℏ2
|Ej,0 − Ej−1,0 | =
I
with corresponding frequency
2πjℏ
νj = ≈ 113jGHz
I

26
3.3 Spins
3.3.1 Integer or half integer?
Consider |ψ⟩ ∈ Hj and expand in J2 , Jz basis:
J
X
|ψ⟩ = am |J, m⟩ , am ∈ C
m=−J

Let’s rotate this by

J
X J
X
U (αẑ) = am e−iαJz /ℏ |j, m⟩ = am e−iαm |j, m⟩
m=−J m=−J

When j, m ∈ N0 for α = 2π we have U (2πẑ) = 1H as expected. However, when

j ∈ N0 + 21 , we would get U (2πẑ) = −1H . In fact, U (2π n̂) = −1H for any n̂.
But there’s no need to worry. j ∈ N0 + 21 because |ψ⟩ and λ |ψ⟩ for λ ∈ C\0
represent the same physical state. Formally, λ1H acts trivially on PH.
Mathematically U (2π n̂) ̸= 1H means this is a “projective” representation of SO(3).
In fact, what we built is a representation of SO(3) which is universal covering
groups of SO(3).
In physics, half-integer representations are called spinors.
Definition. The spin operator S rotates to angular momentum by

J=L+S

where L = X × P is the orbital angular momentum. They all have the same algebra
as J:
[Si , Sj ] = iℏεijk Sk , [Li , Lj ] = iℏεijk Lk
They may have the same representations, but we will see for orbital angular mo-
mentum L, we only take j ∈ N, while for S, we can have j = N or N + 12 .

[Lecture 11 finish]

H admits a basis of eigenvectors of S2 and Sz (which commute: [S2 , Sz ] = 0):

S2 |s, σ⟩ = ℏ2 s(s + 1) |s, σ⟩ , Sz |s.σ⟩ = ℏσ |s, σ⟩

where 2s ∈ N0 and σ ∈ (−s, −s + 1, · · · , s − 1, s) and so dim Hs = 2s + 1. The

H-space of a particle with spin is L2 (R3 , d3 x) ⊗ Hs where Hs ∼
= C2s+1 .
Let’s build explicitly the representation of the algebra of spin operators i.e. Hs .

27
3.3.2 Spin matrices
Definition (Scalar). Particle with s = 0 is called a scalar : Hs=0 = C. The only
state up to rescaling for s = 0 is |s = 0, σ = 0⟩ = |0, 0⟩. Scalars are unchanged by
rotations.
Definition (Spin-half). For s = 1/2, there are two linearly independent states
| 12 , 12 ⟩ = |↑⟩ , | 12 , − 12 ⟩ = |↓⟩
It follows that ⟨↑ | ↑⟩ = ⟨↓ | ↓⟩ = 1, ⟨↑ | ↓⟩ = 0. For a generic state can be written
as
|ψ⟩ = a |↑⟩ + b |↓⟩
It if’s normalised, then |a|2 + |b|2 = 1.
In this basis,
   
⟨↑ |Sz | ↑⟩ ⟨↑ |Sz | ↓⟩ ℏ 1 0 ℏ
Sz = = = σz
⟨↓ |Sz | ↑⟩ ⟨↓ |Sz | ↓⟩ 2 0 −1 2
where σz is merely a customary notation, it has nothing to do with the eigenvalue.
Using Sx = 12 (S+ + S− ) and Sy = 2i1
(S+ − S− ), we can compute
   
⟨↑ |Sx | ↑⟩ ⟨↑ |Sx | ↓⟩ ℏ 0 1 ℏ
Sx = = = σx
⟨↓ |Sx | ↑⟩ ⟨↓ |Sx | ↓⟩ 2 1 0 2
 
ℏ 0 −i ℏ
Sy = = σy
2 i 0 2
1
When s = 2 we often write S = ℏ2 σ where σ are called Pauli matrices.

3.3.3 The Sterm-Gerlach experiment

Do spinors exist? Yes, for example: electrons, quarks, protons, neutrons, etc. The
Sterm-Gerlach experiment proved the existence of spinors and quantised angular
momentum.
Consider atoms of mass M and spin s. They travel through a magnetic field
B = Bẑ. The Hamiltonian is
P2 P2 µ 
H= −µ·B= − S · (ẑB)
2M 2M ℏs | {z }
=Sz ·B(x)

µS
where µ = ℏs is the atom’s magnetic dipole moment. Atoms in |s, σ⟩ feels a force
d ⟨P⟩  µ  µσ
= ⟨∇B · Sz ⟩ = ⟨∇B⟩
dt ℏs s

28
that depends on σ i.e. how well the atom’s spin is aligned with B. Classically we’d
expect the atoms to land anywhere, but in quantum mechanics, σ is discrete as
σ ∈ (−s, −s + 1, · · · , s), making the landing points discrete as well. It’s observed
that the beam splits into 2s + 1 rays. This was originally experimented on silver
atoms for s = 1/2, so the beam split into two.

3.3.4 Spin precession (Intermezzo)

Classically a magnetic diple µ in a magnetic field B experiences a torque:
Ṡ = µ × B
µ
If µ = ℏs S on B is constant, S precesses around B with angular velocity
µ
ω=− B
ℏs
For reference, proton has
µ gp e
=
ℏs 2mp c
If proton can’t move (e.g. fixed in a solid), then
µ
H = −µ · B = −B Sz
ℏs
Suppose our proton is initially (see Example Sheet 2) in |n ↑⟩ = e−iϕ/2 cos θ2 |↑⟩ +
eiϕ/2 sin θ2 |↓⟩ where
n̂ = (cos ϕ sin θ, sin ϕ sin θ, cos θ)
After some time t,
time evolution θ θ
|n̂, t = 0⟩ −−−−−−−−−→ e−iH |n̂ ↑⟩ = ei(ωt−ϕ)/2 cos |↑⟩ + ei(ωt−ϕ)/2 sin |↓⟩
2 2
µ
with ω = ℏs B. But this is just equal to |n̂(t) ↑⟩ where n̂ = (θ, ϕ 7→ ϕ − ωt). So
|n(t)⟩ precesses around B with ω.
[Lecture 12 finish]
Definition (Spin one). There are three orthonormal states with s = 1, which are
traditionally called
|+⟩ = |1, 1⟩ , |0⟩ = |1, 0⟩ , |−⟩ = |1, −1⟩
In this basis,
     
0 1 0 0 −i 0 1 0 0
ℏ ℏ
Sx = √ 1 0 1 , Sy = √  i 0 −i , Sz = ℏ 0 0 0
2 0 1 0 2 0 i 0 0 0 −1

29
which follow as Sz is diagonal in this basis and Sx , Sy are computed from Sx ± =
S± iSy using
√ √
S+ |0⟩ = 2ℏ |+⟩ , S+ |−⟩ = 2ℏ |0⟩ , S+ |+⟩ = 0
√ √
S− |0⟩ = 2ℏ |−⟩ , S− |+⟩ = 2ℏ |0⟩ , S− |−⟩ = 0

3.4 Orbital angular momentum

Recall J = L + S with L the orbital angular momentum satisfying [Li , Lj ] =
iℏεijk Lk . So particles are described by |ψ⟩ ∈ L2 (R3 , d3 x). Hence
L=X×P
For U (α) = e−iα·L/ℏ to describe a rotation we must have for i = 1, 2, 3:
U † (α)Xi U (α) = Rji (α)X j ⇒ U † XU = RX ⇒ XU = U RX
Then
X(U (α) |x⟩) = U (α)(R(α)X) |x⟩ = U (α)(R(α)x) |x⟩ = (R(α)x)(U (α) |x⟩)
Hence U (α) |x⟩ = |R(α)x⟩ on position eigenvectors. Under a rotation |ψ⟩ 7→ |ψ⟩ =
U (α) |ψ⟩, the wavefunction change is
ψ ′ (x) = ⟨x|ψ ′ ⟩ = ⟨x|U (α)|ψ⟩ = ⟨R−1 (α)x|ψ⟩ = ψ(R−1 (α)x)
Similarly,
⟨x|U2 U1 |ψ⟩ = ⟨R2−1 x|U1 |ψ⟩ = ⟨R1−1 R2−1 x|ψ⟩ = ⟨(R2 R1 )−1 x|ψ⟩
Thus representation is unitary:
Z Z
′ ′ −1 3
⟨ϕ |ψ ⟩ = −1
ϕ(R x)ψ(R x)d x = ϕ(y)ψ(y)d3 y = ⟨ϕ|ψ⟩
R3 R
3
where the second equality follows since d x is rotationally invariant.
Recall the relation [Li , Lj ] = iℏεijk Lk means we can simultaneously diagonalise
L2 and Lz . So we build a Hilbert space to represent L. We let |l, m⟩ denote the
eigenstates of L2 and Lz :
L2 |l, m⟩ = ℏ2 l(l + 1) |l, m⟩ , Lz |l, m⟩ = ℏm |l, m⟩
For R(α) ∈ SO(3) we have R(2π α̂) = 1. So

U (2π α̂) |x⟩ = |R(2πˆ(α)x⟩ = |x⟩

Hence we must have
U (2π α̂) = ei 2πα̂·L/ℏ = 1H ∀α̂
So unlike spins, only l ∈ N0 representations are allowed i.e. no half integers.

30
Where are these (2l + 1)-dimensional representations Hl inside L2 (R3 ) for which
dim L2 = ∞? To answer this, let’s use spherical coordinates. Since

Lz = Xx · Py − Xy Px

for |ψ⟩ ∈ L2 , we have

⟨x|Lz |ψ⟩ = ⟨x|Xx Py − Xy Px |ψ⟩

= −iℏ(X∂y − Y ∂x )ψ(x)
∂
= −iℏ ψ(x)
∂ϕ
Then
mℏ ⟨x|l, m⟩ = ⟨x|Lz |l, m⟩ = −iℏ∂ϕ ⟨x|l, m⟩
which has solution
⟨x|l, m⟩ = eimϕ Kl,m (r, θ)
for some Kl,m . Also L± as
  
∂ ∂ ∂
⟨x|L± |ψ⟩ = −iℏ (y ∓ ix) −z ∓i ψ(x)
∂z ∂y ∂x

or in spherical coordinates,
 
∂ ∂
⟨x|L± |ψ⟩ = ±iℏ ± i cot θ ψ(x) (∗)
∂θ ∂ϕ

Using ⟨x|L+ |l, l⟩ = 0 we can solve (∗) for ⟨x|l, l⟩ giving

⟨x|l, l⟩ eilϕ sin2 θRe (r)

for some Re .

[Lecture 13 finish]

3.4.1 Spherical harmonics

The eigenstates we’ve constructed are called spherical harmonics denoted as

Yl,m (θ, ϕ) = ⟨θ, ϕ|l, m⟩

where |θ, ϕ⟩ = |X⟩X2 =1 . They satisfy

1
l(l + 1) ⟨θ, ϕ|l, m⟩ = l(l + 1)Yl,m = ⟨θ, ϕ|L2 |l, m⟩ = −∇2S 2 Ylm
ℏ

31
where ∇2S2 is the Laplacian operator on the unit sphere. Note that Yl,m ’s are
orthonormal, in a sense that
Z
Yl′ ,m′ (θ, ϕ)Yl,m (θ, ϕ) sin θdθdϕ = δll′ δmm′
S2

Under the parity P : (x, y, z) 7→ (−x, −y, −z), θ 7→ π − θ and ϕ 7→ π + ϕ. So

Yl,m (−X̂) = (−1)l Yl,m (X̂), Yl,m (π − θ, π + ϕ) = (−1)l Yl,m (θ, ϕ)

3.5 Central potentials

Consider a scalar particle of mass M moving in a central potential:
P2
H= + V (|X|)
2M
Since [H, L2 ] = 0 = [H, Lz ], [Lz , L2 ] = 0 and we can use a basis |n, l, m⟩ of
eigenstates with H |n, l, m⟩ = En,l,m |n, l, m⟩. Energy, total (orbital) angular mo-
mentum, and angular momentum in ẑ direction. Note that [H, L± ] = 0 then En,l,m
must be independent of m (from H(L± |n, l, m⟩) and using [H, L± ] = 0). Hence
there is a (2l + 1)-fold degeneracy from raising/lowering m.
Generally we’d expect energies to depend on l, however, Hamiltonians with more
degeneracy have “hidden” symmetries.

3.5.1 The 3D quantum harmonic oscillator

The Hamiltonian of the 3D QHO is
1 mω 2
H= (P 2 + Py2 + Pz2 ) + (Xx + Xy2 + Xz2 ) = Hx + Hy + Hz
2m x 2
There are three raising/lowering operatros with
1 1
A† = √ (mωX − iP), A = √ (mωX + iP)
2mω 2mω
† †
with [A+
i , Aj ] = δij and [Ai , Aj] = [Ai , Aj ] = 0. Then
 
† 3
H = ℏω A · A +
2
The energy eigenstates are then, for nx , ny , nz ∈ Z,

(A†x )nx (A†y )ny (A†z )nz

|n⟩ = |nx , ny , nz ⟩ = p |0⟩
nx !ny !nz !

32
 3


with energy E = ℏω N + 2 where N = nx + ny + nz . The degeneracy is then
−nx
N NX
X 1
1= (N + 2)(N + 1)
nx =0 ny =0
2

But why is it so big? Indeed, H has more symmetries.

H is invariant under the transformation Ai 7→ uij Aj (so Adi ag 7→ A†k u†ki ) provided
that u†ki uij = δkj . uij mixes X ∼ Re(A) with P ∼ Im(A). The 3 × 3 matrix uij
has 9 real parameters so we’d expect 9 conserved generators. Infinitesimally,

uij = δij − iεik tkj + O(ε2 ), t†ij = tij

Requiring that

U † (u)Ai U (u) = uij Aj , U (u) = 1H − iεik Tkj + · · ·

We find that
Tij = A†i Aj
Tij is Hermitian and conserved:
1
[Tij , H] = [A†i Aj , A†k Ak ] = A†i [Aj , A†k ]Ak + A†k [A†i , Ak ]Aj
ℏω
= A†i δjk Ak − A†k δjk Aj = 0

Decompose Tij into trace, anti-symmetric part and symmetric part:

 
1 † 1 † † 1 † † 1 †
Tij = δij Ak Ak + (Ai Aj − Aj Ai ) + (Ai Aj + Aj Ai ) − δij Ak Ak
3 2 2 3
= Trace + A†[i Aj] + A†<i Aj>

i.e. Tr(Tij ) ∝ H + const. The only symmetric part generates rotations in SO(3):
1
εijk A†j Ak = (mωXj − iPj )(mωXk + iPk )
2mℏω
1
= εijk (Xj Pk − Pj Xk )
2ℏ
i
= Li
ℏ
So
L = −iℏA† × A
[Lecture 14 finish]

33
T⟨i j⟩ maps Span(|n, l, ·⟩) to Span(|n, l, ·⟩ , |n, l − 2, ·⟩). Also m is even (odd) if l is
even (odd). Hence, the degeneracy of an energy eigenstate |n, ·⟩ is (for n even):
l
X X X 1
= (2l + 1) = (n + 2)(n + 1)
2
l even m=−l l even

and we can just replace even with odd in the sum for odd n.

3.5.2 Hydrogen atom

With a Coulomb potential, the Hamiltonian is
P2 κ
H= −
2µ |X|
where κ = e2 /4πε0 . We have [H, L] = 0, so we can find a complete orthonormal
basis H = Span(|n, l, m⟩) of eigenstates of H, L2 , Lz . Energy levels are En,l =
−R/n2 where Rydberg’s constant is
e4
R=µ = 13.6eV
32π 2 ε20 ℏ2
By fixing n, we can find (as in IB QM) l ∈ (0, 1, 2, · · · , n−1). So the nth eigenstate
has degeneracy
n−1 l m−1
X X X n(n − 1)
= (2l + 1) = 2 + n = n2
2
l=0 m=−l l=0

Why are energies independent of l? Keplerion orbits are closed because the Runge-
Lenz vector
1 X
r= P×L−κ
µ |X|
is conserved:
2E 2
ṙ = 0, |r|2 = κ2 + |l|
µ
Definition. The Runge-Lenz operator R† = R is defined to be
1 X
R= (P × L − L × P) − κ
2µ |X|
Then R is conserved for bound orbits with [H, R] = 0 and also R · L = L · R = 0.
Moreover,
2H 2
R2 = κ2 + (L + ℏ2 )
µ
so eigenvalues of H are fixed by those of R and L. Note [Lz , Rj ] = iℏεijk Rk and
[Ri , Rj ] = − 2iℏ
µ Hεijk Rk . Hence the subspace H ⊃ HE = (|ψ⟩ : H |ψ⟩ = E |ψ⟩)
represents the algebra (L, R).

34
Let’s introduce the operators:
 r 
1 µ
A± = L± R
2 −2E

Note that A± is Hermitian for E < 0. The algebra (A+ , A− ) is the SO(2) × SO(2)
algebra:
[A±i , A±j ] = iℏεijk A±k , [A+i , A−j ] = 0
So these are two copies of the algebra of angular momentum. Eigenvalues of |A± |2
are a± (a± + 1)ℏ2 with a± ∈ (0, 1/2, 1, 3/2, · · · ). Given an a± there are 2a± + 1
possible eigenvalues of A±z , m± .
We have an extra constant. Since L · R = R · L = 0, we have

µκ2 ℏ2
 
1 µ
ℏ2 a(a + 1)A2+ = A2− = L2 + R2 = −
4 2|E| 8|E| 4

Then a+ − = a− = a (but m+ and m− are still independent). So states with energy

E are |a, m+ , m− ⟩. Energy levels are determined by 2a ∈ N:

µκ2 1 µκ2 1 µκ2 1

E=− 2
=− 2
=−
8 a(a + 1)ℏ + ℏ/4 2 (2a + 1) 2 n2

where n ≡= 2a + 1 ∈ (1, 2, 3, · · · ). Degeneracy of the nth level is

m+ m− = (2a + 1)(2a + 1) = n2

35
4 Addition of angular momentum
How does angular momentum of subsystems combine into total angular momen-
tum? Classically, for J1 ⩾ J2 we have

Jtot = J1 + J2 ⇒ |J1 | + |J2 | ⩾ |J|tot ⩾ |J1 | − |J2 |

with equality only when J1 , J2 are (anti)aligned.

4.1 Combining two states

Consider two independent systems with J1 , J2 . Let (|J1 , m1 ⟩) be the eigenstates
of (L21 , L1z ) and similarly for 2. Then (|J1 , m1 ⟩ ⊗ |J2 , m2 ⟩) is a basis of the total
system H = HJ1 ⊗ HJ2 :
X
H ∋ |ψ⟩ = am1 m2 |J1 , m1 ⟩ ⊗ |J2 , m2 ⟩
m1 ,m2

We will find a way to express these states in terms of total angular momentum
eigenvalues of the combined system (J2tot , Jtotz ) ∼ (J, m).

[Lecture 15 finish]

How can we write this as |J, m⟩, eigenstates of J2 = (J1 + J2 )2 and Jz = J1z + J2z ?
We’ll find that we need different j’s so that
M
H = Hj1 ⊗ H j2 = Hj
J

But with what J’s?

The total angular momentum obeys

J2 = J21 + J22 + 2J1 · J2

1
How does J1 · J2 act on states of H? Since Jx = 12 (J+ + J− ) and Jy = 2i (J+ − J− ),
we have
 
J1+ + J1− J2+ + J2− J1+ − J1− J2+ − J2−
2J1 · J2 = 2 − + + + J1z J2z
2 2 2i 2
= J1+ J2− + J1− J2+ + 2J1z J2z

and we know how these act on |J1 , m1 ⟩ |J2 , m2 ⟩.

36
Consider the maximally aligned states |J1 , J1 ⟩ |J2 , J2 ⟩, we have

Jz (|J1 , J1 ⟩ |J2 , J2 ⟩) = (J1z + J2z )(|J1 , J1 ⟩ |J2 , J2 ⟩) = (j1 + j2 )ℏ |J1 , J1 ⟩ |J2 , J2 ⟩

and

J2 (|j1 , j2 ⟩ |j2 , j2 ⟩) = J12 + J22 + J1+ J2− + J1− J2+ + 2J1z J2z |j1 , j2 ⟩

= ℏ2 [j1 (j1 + 1) + j2 (j2 + 1) + 2j1 j2 ] |j1 , j1 ⟩ |j2 , j2 ⟩

= (j1 + j2 )(j1 + j2 + 1)ℏ2 |j1 , j1 ⟩ |j2 , j2 ⟩

where the second steps follows since J1+ |j1 , j1 ⟩ = 0 and similarly for j2 . So we
identify |j1 , j1 ⟩ |j2 , j2 ⟩ as the state |j, j⟩ of the combined system where j = j1 + j2 .
It has maximum Jz . More states are determined by acting with J− = J1− + J2−
on |j, j⟩. Recall that ⟨j, m|j, m⟩ = 1, then
p
J− |j, m⟩ = j(j + 1) − m(m − 1)ℏ |j, m − 1⟩

Consequently p
J− |j, j⟩ = 2jℏ |j, j − 1⟩
where
p p
J1− |j1 , j1 ⟩ = 2j1 ℏ |j1 , j1 − 1⟩ , J2− |j2 , j2 ⟩ = 2j2 ℏ |j2 , j2 − 1⟩

Compounding these, we have

s s
j1 j2
|j, j − 1⟩ = |j1 , j1 − 1⟩ |j2 , j2 ⟩ + |j1 , j1 ⟩ |j2 , j2 − 1⟩
j j

Repeating this builds all |j, m⟩ for j = j1 + j2 and −j ⩽ m < j. But there are also
‘imperfectly’ aligned states, such as |j − 1, j − 1⟩. Since

Jz |j − 1, j − 1⟩ = (j1 + j2 − 1)ℏ |j − 1, j − 1⟩

we must have

|j − 1, j − 1⟩ = a |j1 , j1 − 1⟩ |j2 , j2 ⟩ + b |j1 , j1 ⟩ |j2 , j2 − 1⟩

for some a, b ∈ C. We compute a, b demanding that ⟨j, j − 1|j − 1, j − 1⟩ = 0 and

|a|2 + |b|2 = 1. Then we must have
s s
j2 j1
|j − 1, j − 1⟩ = |j1 , j1 − 1⟩ |j2 , j2 ⟩ − |j1 , j1 ⟩ |j2 , j2 − 1⟩
j j

We can obtain the full (j1 + j2 − 1) “multiplet” |j1 + j2 − 1, m⟩ acting with J− =

J1− + J2− .

37
Proposition. The lowest total angular momentum is |j1 − j2 | (anti-alignment).
Indeed, this accounts for all (2j1 + 1)(2j2 + 1). Wlog j1 ⩾ j2 , then
jX
1 +j2 j jX
1 +j2
X 2j1 (2j2 + 1)
= (2j + 1) = 2 + 2j2 + 1 = (2j1 + 1)(2j2 + 1)
j=j1 −j2 m=−j j=j1 −j2
2

i.e. we’ve covered all states.

Example. Consider j1 = j2 = 21 , then
1
M
H1/2 ⊗ H1/2 = H = H ⊕ H0
j=0

For example, for the Hydrogen atom, the spin of electron and proton are
1 1
Se = , Sp =
2 2
Using our previous result:
1 1 1 1
|1, 1⟩H = | , ⟩ ⊗ | , ⟩ = |↑⟩p |↑⟩e
2 2 e 2 2 p
Applying J−H , this is
1  
|1, 0⟩H = √ |↓⟩e |↑⟩p + |↑⟩e |↓⟩p
2
applying J−H again gives
|1, −1⟩H = |↓⟩e |↑⟩p
All spin 1 states of H are symmetric under the exchange 1 ↔ 2, e ↔ p. The
remaining state |0, 0⟩H must be perpendicular to |1, 0⟩H , so we need

1  
|0, 0⟩H = √ |↓⟩e |↑⟩p − |↑⟩e |↓⟩p
2
This is anti-symmetric under e ↔ p. Indeed this is annihilated by Jz = Jez + Jpz
and also by JHx and JHy :

|↑⟩e |↓⟩p − |↓⟩e |↑⟩p

  
J+e + J−e + J+p + J−p
JHy |0, 0⟩H = √ =0
2 2
So we’ve written H1/2 ⊗ H1/2 = H1 ⊕ H0 and found the triplet |1, 1⟩ , |1, 0⟩ , |1, −1⟩
and the singlet |0, 0⟩.

38
 [Lecture 16 finish]

Example. Recall
1 1
|1, 0⟩ = √ |↓⟩ |↑⟩ + √ |↑⟩ |↓⟩
2 2
1 1
|0, 0⟩ = √ |↓⟩ |↑⟩ − √ |↑⟩ |↓⟩
2 2
|1, 1⟩ = 1 |↑⟩ |↑⟩ , |1, −1⟩ = 1 |↓⟩ |↓⟩

These are called Clebsch-Gordon coefficients.

Example. We claim that
1 3 1
1⊗ = ⊕
2 2 2
where 32 = 1 + 12 and 21 = 1 − 12 . Let’s check their dimensions: spin-p has dimension
2p + 1, so LHS has dimension 3 × 2 while RHS has 4 + 2. So they agree.
3
We have the j = 2 multiplet:

3 3
| , ⟩ = 1 |↑, ↑⟩ |↑⟩
2 2 r r
3 1 2 1
| , ⟩= |1, 0⟩ |↑⟩ + |1, 1⟩ |↓⟩
2 2 3 3
r r
3 1 1 2
| ,− ⟩ = |1, −1⟩ |↑⟩ + |1, 0⟩ |↓⟩
2 2 3 3
3 3
| , − ⟩ = 1 |1, −1⟩ |↓⟩
2 2
1
and the remaining j = 2 multiplet is:
r r
1 1 1 2
| , ⟩= |1, 0⟩ |↑⟩ − |1, 1⟩ |↓⟩
2 2 3 3
r r
1 1 2 1
| ,− ⟩ = |1, −1⟩ |↑⟩ − |1, 0⟩ |↓⟩
2 2 3 3

4.2 Identical particles

All electrons (protons, quarks, . . . ) are indistinguishable. Some Hilbert space
L2 (R3 ) ⊗ C2 has the same set of quantum numbers (eigenvalues of the set of com-
muting Hermitian operators).
For two particles, let |ψ⟩ = H1 ⊗H2 with basis {|α1 , α2 ⟩ = |α1 ⟩⊗|α2 ⟩} where αi are

39
quantum numbers of the i-th particle. For indistinguishable particles exchanging
1 ↔ 2 maps
|α1 , α2 ⟩ 7→ |α2 , α1 ⟩ = λ |α1 , α2 ⟩ for some λ ∈ C
Exchanging twice gives |α1 , α2 ⟩ = λ2 |α1 , α2 ⟩, so λ = ±1. If λ = +1 then particles
are called bosons (e.g. photons, Hydrogon atoms, Higgs, . . . ) and if λ = −1 then
fermions (electrons, neutrinos, etc.). For two indistinguishable bosons,
|α1 , α2 ⟩ + |α2 , α1 ⟩
|ϕ⟩ = √
2
where √1 is a normalisation constant. For fermions,
2

|α1 , α2 ⟩ − |α2 , α1 ⟩
|ψ⟩ = √
2
If α1 = α2 then |ψ⟩ = 0, so no two fermions can be in the same state. This is
known as Pauli’s exclusion principle. It is important to exchange all quantum
numbers, for example: the transformation

ψσ1 ,σ2 (x1 , x2 ) 7→ −ψσ2 ,σ1 (x2 , x1 )

has no relation to ψσ1 ,σ2 (x2 , x1 ). For N identical particles let σ be a permutation

(1, 2, 3, · · · , N ) 7→ (σ1 , σ2 , σ3 , · · · , σN )

then
Wσ |α1 , α2 , · · · , αN ⟩ = ησ |ασ1 , ασ2 , · · · , ασN ⟩
where Wσ is the operator for this permutation and ησ = ±1 since every σ is a
composition of pairwise exchanges. For bosons we have ησ = +1 ∀r, while for
fermions ησ = sign(r) = (−1)number of pairwise exchanges . Hence the Hilbert space is
HN b/f for N indistinguishable particles. So HN b is the symmetrised tensor product
of N H1b ’s while HN f is the alternating tensor product of N H1f ’s.
It turns out that particles with spin S ∈ (0, 1, 2, · · · ) are bosons, whereas particles
with S ∈ ( 12 , 32 , · · · ) are fermions. This follows from spin-statistic theorem and
requires Lorentz invariant and quantum field theory. This applies also to composite
systems as well as fundamental particles.

4.3 Ideal quantum gas

Consider N indistinguishable free fermions in a box of side L:
N
X P2a
J=
a=1
2m

40
 ℏ2 k 2
A single particle with energy eigenstate |k⟩ has E = 2m with P(k) = ℏk |k⟩. Here
k = 2πL
( n1 , n2 , n3 ) for ni ∈ Z and wavefunction

e−ik·X
⟨X|k⟩ =
L3/2
Pauli’s exclusion principle gives only one fermion per state. At low energy, tem-
perature is zero, so only the first energy levels are filled. Highest fillled state has
ℏ2 k 2
|k| = kF , the Fermi momentum, and Fermi energy EF = 2mF . For N >> 1 we go
to continuum. Each fermion occupies a k-space volume of (2π/L)3 , so we have
Z |k|=kF  3
3 4π 2 2π
d k= kF = N
0 3 L
The total energy is then
Z kF
1 ℏ3 k 2 ℏ2 kF5 L3 ℏ2 1
Etot = d3 k 3
= 2
= 2m
(6π 2 N )5/3 2/3
0 (2π/L) 2m 20π m 20π V
Any reduction of size of the box is opposed by a degeneracy pressure:
∂Etot ℏ2 (6π 2 N/V )5/3
Pdeg = − =
∂V 30π 2 m
Brown drawf and neutron stars are supported by such a pressure.
[Lecture 17 finish]

4.4 Exchange and parity

Exchanging two indistringuishable particles is related to parity. Let
X1 + X2
XCoM = , PCoM = P1 + P2
2
P1 − P2
Xrel = X1 − X2 , Prel =
2
Exchanging 1 ↔ 2 gives
(XCoM , PCoM ) → (XCoM , PCoM ), (Xrel , Prel ) → (−Xrel , −Prel )

4.4.1 Parity of pion

Consider a pion π − orbiting a Deuterium nucleus. It has the same energy levels as
the Hydrogen atom, with ground state |n, l, m⟩ = |1, 0, 0⟩. Sπ = 0, SD = 1, so this
atom has J = 1. This atom decays into two neutrons:
π− + D+ → n + n
Since Sn = 21 as neutron is a fermion, so the final state be anti-symmetric under
exchange. The total angular momentum is conserved, so J = 1 in final state.

41
 1 1
Spin: Recall that 2 ⊗ 2 = 1 ⊕ 0.
• If Snn = 0 (anti-symmetric), we require l to be even (symmetric). but then
l = 1 is forbidden.
• If Snn = 1 (symmetric), we require l to be odd (antisymmetric). l = 1 is
allowed. Since J = 1 ∈ Sn,n ⊗ odd, the only possibility is l = 1. Assuming
that parity is conserved, we can derive the parity ηπ of the pion:

Π |X, π⟩ = ηπ |−X, π⟩

where Π is the parity operator.

We have

π+D →n+n
ηπ · πD = (−1)l ηn2 = −1

From other experiments we know ηD = 1 so we conclude that pion has ηπ = −1.

The pion is called a pseudo scalar.

42
5 Perturbation theories
5.1 Time-independent perturbation theory
For realistic systems we can’t solve dynamics exactly. Let’s develop some approx-
imation. Suppose H = H0 + ∆H and we can solve H0 . The idea is to consider
H = H0 + λ∆H and Taylor expand around λ = 0 to approximate for the case
where λ = 1. We seed energy eigenstates |Eλ ⟩ of Hλ . If the dependence on λ is
analytic around λ = 0:

|Eλ ⟩ = |α⟩ + λ |β⟩ + λ2 |γ⟩ + · · ·

then if Hλ |Eλ ⟩ = E(λ) |Eλ ⟩, we have

E(λ) = E (0) + λE (1) + λ2 E (2) + · · ·

Plugging these expansion in Hλ |Eλ ⟩ = E(λ) |Eλ ⟩ gives

(H0 + λ∆H)(|α⟩ + λ |β⟩ + λ2 |γ⟩ + · · · )

= (E (0) + λE (1) + λ2 E (2) + · · · )(|α⟩ + λ |β⟩ + λ2 |γ⟩ + · · · )

Equality must be reached on each order of λ, so we can solve to get

• λ0 : H0 |α⟩ = E (0) |α⟩.
• λ1 : H0 |β⟩ + ∆H |α⟩ = E (0) |β⟩ + E (1) |α⟩.
• λ2 : H0 |γ⟩ + ∆H |β⟩ = E (0) |γ⟩ + E (1) |β⟩ + E (2) |α⟩.
For the zeroth order we have H0 |α⟩ = E (0) |α⟩, which is known by assumption.
Now rename |a⟩ = |n⟩ and E (0) = En . For each n, we can contract both sides of
the above equation of coefficients with ⟨n| to find

⟨n|∆H|n⟩ = ⟨n|E (1) |n⟩ ⇒ E (1) = ⟨n|∆H|n⟩

Contracting with ⟨m| =

̸ ⟨n| gives

⟨m|∆H|n⟩ = (En − Em ) ⟨m|β⟩

P
Now expand |βn ⟩ = k bk |k⟩ in the basis H0 |k⟩ = Ek |k⟩ with bm = ⟨m|β⟩. If
En ̸= Em (n-th is not degenerate) we have
⟨m|∆H|n⟩
bm =
En − Em
so
X ⟨m|∆H|n⟩
|βn ⟩ = |m⟩
En − Em
m̸=n

43
From the Examples Sheet we have ⟨Eλ |Eλ ⟩ = 1 at order λ1 so bn = ⟨n|β⟩ = 0. At
the second order λ2 we contract with ⟨n| to find
X | ⟨n|∆H|m⟩ |2
E (2) = ⟨n|∆H|β⟩ =
En − Em
m̸=n

For similar ⟨n|∆H|m⟩ nearby states contribute the most.

[Lecture 18 finish]

Example. Consider the 1D translated quantum harmonic oscillator:

P2 1 P2 1 1
H= + mω 2 X 2 − λmω 2 x0 X = + mω 2 (X − λx0 )2 − λ2 mω 2 x20
|2m 2{z
| {z } 2m 2 2
} ∆H
H0

1
+ n − 12 λ2 mω 2 x20 . Treating ∆H =


The exact energy levels are En (λ) = ℏω 2
−λmω 2 x0 X as a perturbation gives
X | ⟨k|X|n⟩ |2
En (λ) = Dn + λ ⟨n|∆H|n⟩ + λ2 m2 ω 4 x20 + O(λ3 )
En − Ek
k̸=n
 
1 1
= ℏω n + + 0 − λ2 mω 2 x20 + O(λ3 )
2 2
which happens to be exact.
√
Example. Consider a quantum harmonic √ oscillator with frequency ω(λ) = ω 1 + λ.
The exact energy levels are En (λ) = ℏω 1 + λ n + 12 . Treating ∆H = 12 mω 2 λX 2


P2
with H0 = 2m + 12 ω 2 mX 2 . From the perturbation theory we have

λ λ2 X | ⟨k|X 2 |n⟩ |2
En (λ) = En + mω 2 ⟨n|X 2 |n⟩ + m2 ω 4 + O(λ3 )
2 4 (n − k)ℏω
k̸=n
2 √
   
1 λ λ 3
= n+ ℏω 1 + − + O(λ ) by Taylor expanding 1 + λ
2 2 8

Example. Let Hλ = HQHO + λX 4 . Perturbatively we have

∞
1 X
E0 (λ) = ℏω + λk ak
2
k=1

where √ 
(−1)k+1 3k 6
  
1 95 1 1
ak = 3/2
Γ k+ 1− +O
π 2 72 k k

44

k
For k → ∞ we have Γ k + 12 ∼ k k e−k = (k/e)k then λk ak ≈ λk


e . This has
zero radius of convergence i.e. diverges for all λ > 0. Physically this is as expected
because λ < 0 is unbounded potential. Perturbation theory does not converge:
N
X
lim λk ak = ∞
N →∞
k=0

but is is asymptotic:
N
1 X
lim N
E(λ) − λk ak = 0 for fixed N
λ→0 λ
k=1

5.2 Degenerate perturbation theory

If H0 has degenerate states En,α = En,α′ for different α ̸= α′ , even tiny perturba-
tion can have a large effect. We already saw the problem in
X ⟨m|∆H|n⟩
|nλ ⟩ = |n⟩ + λ |m⟩ + · · · (∗)
En − Em
m̸=n

Assume the degenerate subspace V ⊂ H with H0 |ψ⟩ = EV |ψ⟩ for all |ψ⟩ ∈ V . Let
⟨r|∆H|r ′ ⟩ ′
{|r⟩}N
r=1 be an orthonormal basis. Divergence of EV −EV is avoided if ⟨r|∆H|r ⟩ =
0 for r ̸= r i.e. if ∆H is diagonalised in V . We can always diagonalise ⟨r|∆H|r′ ⟩
′

because it’s Hermitian. Let {|n⟩} be the new diagonal basis then ⟨m|∆H|n⟩ = 0
for m ̸= n. Then (∗) still applies.

5.2.1 The Stark effect

Suppose a Hydrogen atom is in a constant electric field E = |E|x̂e . The new inter-
action is then ∆H = l|E|X3 (neglecting spin). To compute ⟨n, l, m|∆H|n′ , l′ , m′ ⟩
we prove some general results. Parity P 2 = 1H implies
⟨n, l′ , m′ |X3 |n, l, m⟩ = ⟨n, l′ , m′ |P P X3 P P |n, l, m⟩
′
= (−1)l +l+1 ⟨n, l′ , m′ |X3 |n, l, m⟩
(1)
These are called selection rules. Hence ⟨1, 0, 0|X3 |1, 0, 0⟩ = 0 so E0 = 0. The first
excited state n = 2 has 4-degeneracy:
V = span (|2, 0, 0⟩ , |2, 1, 1⟩ , |2, 1, 0⟩ , |2, 1, −1⟩)
We’ll need
Lz X3 |2, 1, ±1⟩ = ( z , X3 ] + X3 Lz ) |2, 1, ±1⟩ = ±ℏ (X3 |2, 1, ±1⟩)
[L 

Hence ⟨2, 0, 0|X3 , 2, 1, ±1⟩ = 0.

[Lecture 19 finish]

45
It follows that  
0 0 a 0
0 0 0 0
∆H = e|E| 
a 0

0 0
0 0 0 0
where a = ⟨2, 0, 0|X3 |2, 1, 0⟩ = −3a0 where a0 is the Bohr radius. ∆H is diago-
nalised in the new basis
|2, 0, 0⟩ − |2, 1, 0⟩ |2, 0, 0⟩ + |2, 1, 0⟩
, |2, 1, 1⟩ , |2, 1, −1⟩ , √
2 2
with eigenvalues −3a0 , 0, 0, −3a0 repsectively. Degeneracy between |2, 1, ±1⟩ is not
lifted. Classically, the energy −3ea0 |E| suggest the n = 2 states have an electric
dipole moment of magnitude 3ea0 .

5.3 Time-dependent perturbation theory

Consider time-independent H0 (which we can solve) and time-dependent pertur-
bation ∆(t). In the Schrödinger picture we have

iℏ∂t |ψ(t)⟩S = (H0 + ∆(t))S |ψ(t)⟩S

In the interaction picture we move part of the time evolution to the operators:

|ψ(t)⟩I = eiH0 t/ℏ |ψ(t)⟩S , AI (t) = eiH0 t/ℏ AS (t)e−iH0 t/ℏ

The table below shows the existence of evolution for states and operators in each
picture:

Schrödinger Interaction Heisenberg

States Yes Yes No
Operators No Yes Yes

Substituting into the TDSE we have

iℏ∂t |ψ(t)⟩I = ∆I (t) |ψ(t)⟩I (†)

To solve this, let {|n⟩} be an orthonormal basis of H with H0 |n⟩ = En |n⟩. Then
X
|ψ(t)⟩I = an (t) |n⟩
n

Contracting (†) with ⟨k| gives

X
iℏȧk = λ an (t)ei(Ek −En )t/ℏ ⟨k|∆S (t)|n⟩
n

46
or equivalently
Z tX
λ ′
ak (t) = ak (t0 ) + an (t′ )ei(Ek −En )t /ℏ ⟨k|∆S (t′ )|n⟩ dt′
iℏ t0 n

To solve this in perturbation theory let Hλ = H0 + λ∆(t) and

an (t) = a(0) (1) 2

n (t) + λan (t) + O(λ )

Solving, we get
(0)
• λ0 : an (t) = an (t0 ) = const.
• λ1 : Z tX
(1) λ ′
ak = a(0)
n e
iωkm t
⟨k|∆S (t′ )|n⟩ dt′
iℏ t0 n

where ωkm = (Ek −Em )/ℏ. Assuming we start with |ψ(t0 )⟩ = |m⟩ so an (t0 ) =
δnm , we get
1 t iωkm t′
Z
ak (t) = δkm + e ⟨k|∆S (t)|m⟩ dt′
iℏ t0

Example (Forced oscillator). Consider pushing a quantum harmonic oscillator

2 2
with ∆(t) = F0 Xe−t /tau , so

P2 1 2 2
H= + ωm2 X 2 −F0 X −t /tau
2m 2 | {z }
∆(t)
| {z }
H0

Suppose |ψ(−∞)⟩ = |0⟩. For |k⟩ ≠ |0⟩ we have

Z
F0 ′ ′2 2
lim ak (t) = − eikωt e−t /τ ⟨k|X|0⟩ dt′
t→∞ iℏ −∞
r
πℏ 2 2 2
= λF0 τ e−k τ ω /4 ⟨k|A† |0⟩
2mω
r
πℏ 2 2
= iδk1 F0 τ e−ω τ /4
2ωm
The probability that the oscillator is in |1⟩ after a long time is then

P(|0⟩ → |k⟩) = | ⟨k|0, t⟩S |2

= | ⟨k|0, t⟩I |2 = |ak (t)|2
(
0 k≠ 1
= πℏF 2 2 −ω2 τ 2 /2
2mω τ e k=1
0

47
Remark.
√
• Maximum probability occurs at τ = 2/ω (resonant).

• There is no transition as τ → ∞ (special case of Adiabatic theorem).

• Higher |k⟩ only excited at higher orders in perturbation theory.
Example (Monochromatic perturbations). Consider a monochromatic perturba-
tion (
0 t<0
∆S (t) =
∆e−iωt + ∆† eiωt t > 0
Here ∆ is time-independent. If the state was |m⟩ at t < 0 then for t > 0 and
|k⟩ =
̸ |m⟩ we have

⟨k|∆|m⟩ h i(ωkm −ω)t i ⟨k|∆† |m⟩ h i(ωkm +ω)t i

ak (t) = e −1 + e −1
ℏ(ωkm − ω) (ωkm + ω)ℏ

When Ek ≃ Em + ℏω then the first term dominates (absorption). When Ek ≃

Em − ℏω then the second term dominates (stimulated emission).

[Lecture 20 finish]

We claim that as t → ∞ we can write

1 sin(ωt/2) π
= δ(ω) + O(t0 )
t ω2 2
Definition. The transition rate from |m⟩ to |k⟩ is defined by

∂
Γ(|m⟩ → |k⟩) = lim |ak (t)|2
t→∞ ∂t
At late times,
(
2π 2
Γ= ℏ | ⟨k|∆|m⟩ | δ(Ek − Em − ℏω) absorption
2π † 2
ℏ | ⟨k|∆ |m⟩ | δ(Ek − Em + ℏω) emission

These are known as Fermi’s golden rules.

5.4 Ionisation
Strong radiation can ionise an atom, liberating e− to a state with positive energy.
−r/a
Assume Hydrogen is in its ground state ⟨X|1, 0, 0⟩ = eπa3 where a is the Bohr

48
radius, and compute the transition to the plane wave ⟨X|k⟩ = eik·x /(2πℏ)3/2 . In
the dipole approximation the perturbation is

∆(t) = e(E · Xe−iωt + (E · X)† eiωt ) = 2eE · X cos(ωt)

For radiation of frequency ω we choose E = |E|ẑ. The matrix element ⟨k|X3 |1, 0, 0⟩
is then
−ik·x −r/a
√
8a5
Z
3 e ze 4πkz 8 2 cos θ
d x √ = √ ≈
(2πℏ)3/2 πa3 i(2πℏ)3/2 πa3 (1 + |k|2 a2 )3 πℏ3/2 |k|5 a5/2

where kz = |k| cos θ and cos θ = Ê · k̂ and in the last equality we used the dipole
approximation with |k|a >> 1. Transition necessarily absorbs energy
2π 2 2
Γ|1,0,0⟩→|k⟩ = l |E| | ⟨k|X3 |1, 0, 0⟩ |2 δ(Ek − Em − ℏω)
ℏ
Definition. The differential ionisation rate is defined to be

dΓ|1,0,0⟩→|k⟩ = Γ|1,0,0⟩→|k⟩ ℏ3 k 2 dk sin θdθdϕ

| {z }
dΩS 2

which describes the ionisation rate to momenta in the range (k, k + dk).
Hence we can integrate over k = |k| but keep the angular information:

dΓ|1,0,0⟩→|k⟩ 256 e2 E 2 mc cos2 θ

=
dΩ πℏ3 |k|3 a5
p √
where ℏ|k| = 2m(ℏω + E1,0,0 ) ≈ 2mℏω.

49
6 Interpreting quantum mechanics
6.1 The density operator
We currently have some issues with quantum mechanics:
• Often we don’t know the exact quantum state, and
• to discuss measurement we’d need to quantise the environment.

To make progress, we make the following definition:

Definition. The density operator is defined to be
X
ρ= Pα |ψα ⟩ ⟨ψα |
α

where ¶α is the classical probability (Pα ∈ R) to be in state |ψα ⟩ ∈ H. Note that

|ψa ⟩’s need not be orthogonal.

Definition. If there exists |x⟩ ∈ H such that ρ = |x⟩ ⟨x| the system is said to be
pure, otherwise it is mixed. Any ρ : H → H with the properties:
• ρ = ρ† Hermitian (since probabilities real).
• ⟨ϕ|ρ|ϕ⟩ ⩾ 0 for all |ϕ⟩ ∈ H (since probabilities non-negative).

• TrH (ρ) = n ⟨n|ρ|n⟩ = 1 (since probabilities sum to one).

P

is the density operator of the same system.

Indeed, for dimH < ∞, let {|n⟩} be the P eigenstates of ρ with ρ |n⟩ = Pn |n⟩ with
Pn ∈ P R since ρ = ρ† . In this basis ρ = n Pn |n⟩ ⟨n| and we can normalise such
that n Pn = 1. All components of ρ are bounded and the diagonal elements
satisfy 0 ⩽ ρn,n ⩽ 1 since Trρ = 1. For off-diagonal elements, wlog let ρ = A† A
then
√ 1
q
|ρn,m | = | ⟨n|A† A|m⟩ | ⩽ ∥A |m⟩∥ · ∥A† |n⟩∥ = ρm,m ρn,n ⩽
2
The system is pure if and only if ρ2 = ρ. To see this, note that if ρ = |ψ⟩ ⟨ψ| then
ρ2 = |ψ⟩ ⟨ψ|ψ⟩ ⟨ψ| = |ψ⟩ ⟨ψ| = ρ. Conversely, if (ρ − 1)ρ = 0 then the eigenvalues
of ρ are either 0 or 1. Since Trρ = 1 it would only have one eigenvalue 1. Then
there exists some normalised |x⟩ ∈ H such that ρ = |x⟩ ⟨x|.

[Lecture 21 finish]

50
For dim H = N < ∞, ρ is specified by the number of real parameters

(N 2 − N )2
N + − 1 = N2 − 1
diagonal 2 trace
off-diagonal

For a system described by ρ, the expectation value of Q = Q† is

X X
⟨Q⟩ = TrH (ρQ) = Pα Tr(|ψα ⟩ ⟨ψα | Q) Pα ⟨ψα |Q|ψα ⟩
α α

In the Schrödinger picture, |ψ(t)⟩ = U (t) |ψ(0)⟩ with U (t) = e−iH(t−t0 ) (for ∂t H =
0). Hence
dρ(t)
ρ(t) = U (t)ρ(0)U † (t) → iℏ = [H, ρ(t)]
dt
In the Heisenberg picture, ρH (t) = ρH (0) (does not evolve). Of course, ⟨Q⟩ is the
same in all pictures.
Example. Consider a 2-state system H ∼
= C2 with basis {|↑⟩ , |↓⟩} (qubit).
• If state is |↑⟩ then ρ = |↑⟩ ⟨↑| (pure).
• If probability to be either |↑⟩ or |↓⟩ is 1
2 then
1 1 1
ρ= |↑⟩ ⟨↑| + |↓⟩ ⟨↓| = 1H
2 2 2
then ρ2 ̸= ρ so mixed. This gives us the least information about a state.
• If probability is 1
2 to be either |↑⟩ or |↑X ⟩. Then
1 1
ρ= |↑⟩ ⟨↑| + |↑X ⟩ ⟨↑|
2 2  
1 1 |↑⟩ + |↓⟩ ⟨↑| + ⟨↓|
= |↑⟩ ⟨↑| +
2 2 2 2
1 1 1
= 1H + |↑⟩ ⟨↑| + (|↑⟩ ⟨↓| + |↓⟩ ⟨↑|)
4 2 4

6.1.1 The Bloch ball

Any ρ in C2 can be written as
 
1 1 1 + bz bx − iby
ρ= (1 + b · σ) =
2 2 bx + iby 1 − bz

where σ is the Pauli matrices. This has N 2 − 1 real parameters. In C2 , this is 3.

Since Trσ = 0, coefficient of 1 is fixed.

51
For ρ ⩾ 0, we require
1
det(ρ) = (1 − b · b) ⩾ 0 ⇒ |b| ⩽ 1
4
Thus for the Bloch ball, if |b| = 1 the state is pure with spin along b-direction. If
|b| < 1, it is mixed. If b = 0, we are maximally ignorant about the state.

6.1.2 No-cloning theorem (intermezzo)

We cannot make a copy of a quantum system. Let’s try to copy a normalised
|ψ⟩ ∈ H1 to H1 ∼= H1 initially in some blank (normalised) state |e⟩ ∈ H2 . So we
seek unitary C such that
C : |ψ⟩ ⊗ |e⟩ 7→ e−iα(ψ,e) |ψ⟩ ⊗ |ψ⟩
where α(ψ, e) is some probability phase. Since C † C = 1 we have for all |ϕ⟩ ∈ H1 :
⟨ϕ|ψ⟩ = (⟨ϕ|1 ⟨e|2 ) (|ψ⟩1 |e⟩2 )
= (⟨ϕ|1 ⟨e|2 ) C † C (|ψ⟩1 |e⟩2 )
2
= ei(α(ϕe )−α(ψ,e)) ⟨ϕ|ψ⟩
So | ⟨ϕ|ψ⟩ | = | ⟨ϕ|ψ⟩ |2 which implies | ⟨ϕ|ψ⟩ | = 0 or 1. But this cannot be true so
there is a contradiction. So C cannot exist.

6.2 Entropy
The von Neumann entropy
XH
dim
S(ρ) = −TrH (ρ · log ρ) = − Pn log Pn
n

quantifies how much information ρ gives us about the system. Here are some
properties:
• Eigenvalues of ρ are in [0, 1]. So S ⩾ 0. Also S(ρ) = 0 if and only if ρ is pure.
• To maximise S subject to a constant Trρ = 1, we use the Lagrange multiplier
to extremise S(ρ) − λ(1 − Trρ):
(
0 = −TrH (δρ log ρ + ρρ−1 δρ − λδρ) 1
⇒ ρmax = 1H
0 = δλ(Trρ − 1) dimH

If dim H < ∞, the maximal entropy is Smax = S(ρmax ) = log(dim H).

[Lecture 22 finish]

52
6.3 Entanglement
Consider bipartite system H = HA ⊗ HB . We are interested in the “subsystem” A
while B is the “environment”.
Definition. Any |ψ⟩ ∈ H such that |ψ⟩ = |ϕ⟩ ⊗ |χ⟩ for |ϕ⟩ ∈ HA and |χ⟩ ∈ HB is
called a product state. Otherwise it is an entangled state.
 
Example. If HA,B are qubits, |ϕ⟩ = |↑⟩ ⊗ |↓⟩ and |ψ⟩ = |↑⟩+|↓⟩√
2
⊗ |↓⟩, where as

|↑⟩ |↓⟩ − |↓⟩ |↑⟩

|EPR⟩ = √
2
is entangled.
Definition. The reduced density operator for subsystem A is
ρA = hHB (ρAB )
ρA is useful if we only observe A. For example, consider the observable Q =
QA ⊗ 1B , then
⟨Q⟩ = TrH (ρAB Q) = TrHA ⊗HB [ρAB QA ⊗ 1B ] = TrHB [ρA QA ]
For another example, the product state |ψ⟩ = √1 (|↑⟩ + |↓⟩)A ⊗ |↓⟩B has
2

ρA = TrHB (ρAB ) = TrHB (|ψ⟩ ⟨ψ|)

 
1
= TrHB ((|↑⟩ + |↓⟩)A ⊗ |↓⟩B ) ((⟨↑| + ⟨↓|)A ⊗ ⟨↓|B )
2
 
1 1 1
= = ρ2A ⇒ A is in a pure state
2 1 1
whereas for ρAB = |EPR⟩ ⟨EPR|,
 
1 1 0
ρA = TrHB (ρAB ) = ̸= ρ2A
2 0 1
so A is in a mixed state.
Definition. The entanglement entropy quantifies the entanglement between A and
B by
SA = −TrHA (ρA (log ρA ))
namely just the von Neumann entropy of ρA .
It can be shown that SA = SB = S(ρB ) and ρA is pure (and SA = 0) if and only
if ρAB is separable. Conversely, ρA is mixed (so SA > 0) if and only if ρAB is
entangled. Moreover, sub-additivity gives SA∪B ⩽ SA + SB .

53
6.4 Bell’s inequalities
Is QM probabilistic because we miss some hidden “aspect of reality”? Einstein,
Podolsky and Rosen (EPR) proposed a gedankenexperiment (thought experiment)
in the 1930s. Bohm refined this into the following toy model:
Consider e− and e+ = p in spin-0 state (pure), and consider |EPR⟩.
e− e+
Alice Bob
Alice measures spin along â. If Alice measures +ℏ/2 then
|EPR⟩ → |↑a ⟩ |↓a ⟩
Bob measures the spin of e+ along b̂. Using
   
θ θ
|↑b ⟩ = cos e−iϕ/2 |↑a ⟩ + sin eiϕ/2 |↓a ⟩
2 2
where θ, ϕ are spherical coordinates of b w.r.t. a. If Alice founds e− in |↑a ⟩ then
 
θ
P(B, |↑b ⟩) = | ⟨↑b | ↓a ⟩ |2 = sin2
2
Bob and Alice may be spacelike but there is no violation of causality. Probability
is symmetric in A and B so it’s the same whether tA < tB or tA > tB . Can we
explain this phenomenom by saying that e± already had definite spin? Suppose
there exists hidden variables v ∈ Rn that completely determines the spins. Let
P (v)dn v be the probability for v to be in (vv, v + dvv). We want
Z
⟨Se (a)Sp (b)⟩hidden variable average = dn vP (vv)Se (a, v )Sp (b, v )

Conservation of angular momentum gives Se (a, v )+Sp (a, v ) = 0. So ⟨Se (a)Sp (b)⟩ =
− ⟨Sp (a)Sp (b)⟩. Imagine Bob measures spin along b′ . Since Sp (a, v )2 = ℏ2 /4 we
have
| ⟨Se (a)Sp (b)⟩ − ⟨Se (a)Sp (b)⟩ |
Z
= Se (a, v ) [Sp (b, v ) − Sp (b′ , v )] ρ(v)dn v
Rn
Z  
4
= Se (a, v )Sp (b, v ) 1 − 2 Sp (b, v )Sp (b′ , v ) ρ(v)dn v
Rn ℏ
Z  2 
ℏ
⩽ − Sp (b, v )Sp (b′ , v ) ρ(v)dn v
4
ℏ2
= − ⟨Sp (b)Sp (b′ )⟩
4

54
These are Bell’s inequalities (for hidden variable theories).
Let’s see what quantum mechanics predicts. Since (Se ⊗ 1p + 1e ⊗ Sp ) |EPR⟩ = 0,
we have

(a · Se ⊗ 1p ) (1 ⊗ b · Sp ) |EPR⟩ = −1e ⊗ (a · Sp b · Sp ) |EPR⟩

ℏ2 iℏ
Then we use a · S b · S = 4 a · b1 + 2 (a × b) · S. But ⟨EPR|Sp |EPR⟩ = 0, so

ℏ2
⟨a · Sp b · Sp ⟩EPR = a·b
4
In quantum mechanics we can re-write Bell’s inequalities as

ℏ2
| ⟨a · Sp b · Se− ⟩EPR − ⟨a · Se+ b′ · Se+ ⟩EPR | = |a · (b − b′ )|
4
2
But RHS = ℏ4 (1 − b · · · b′ ). This inequality is violated e.g. when a = x̂, b = ŷ
and b′ = √12 (x̂ + ŷ). In 1982, Aspect and Roger totted this (a slightly differed
version of the CHSH).

[Lecture 23 finish]

NB. Only 23 lectures were given in Michaelmas 2024.

55