Quantum Mechanics B

SPA 6413

Prof Andreas Brandhuber

C ENTRE FOR R ESEARCH IN S TRING T HEORY

S CHOOL OF P HYSICS AND A STRONOMY

Q UEEN M ARY U NIVERSITY OF L ONDON

Version 1, August 2020

 Contents

1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
1.1 Classical vs. Quantum 7
1.2 Fundamental Postulates of Quantum Mechanics 10
1.2.1 State of a system . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
1.3 Time Dependent & Independent Schrödinger Equations (TDSE & TISE) 13
1.4 The Probability (or Born) Interpretation of the Wave Function 15
1.4.1 Conservation of Probability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
1.4.2 More on the Probability Interpretation of Ψ . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
1.5 Superposition Principle 20
1.5.1 Double Slit Experiment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
1.6 Optional: Even More on the Born Interpretation of Ψ 22

2 Operators and Commutators . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27

2.1 Quantum Mechanical Observables 27
2.2 Commutators 29
2.2.1 Examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
2.3 Heisenberg’s Generalised Uncertainty Principle (HGUP) 32
2.4 Proof of the uncertainty principle for position and momentum 34
3 The Expansion Theorem (Exp Thm I) . . . . . . . . . . . . . . . . . . . . . . . . 37
3.1 Energy Eigenstates are Stationary States 37
3.2 Orthornormality of Energy Eigenstates and Reality of Eigenvalues 40
3.3 The Expansion Theorem 43
3.3.1 Interpretation of expansion coefficients . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
3.4 Measurement Theorem 46
3.5 Examples of Expansion Theorem 47
3.6 Optional: simple example of a non-stationary state 50

4 Matrix Q.M. and Bra-Ket Notation . . . . . . . . . . . . . . . . . . . . . . . . . . 53

4.1 Motivation 53
4.2 Basics of Matrix Quantum Mechanics 54
4.2.1 States in Matrix Quantum Mechanics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
4.2.2 Operators in Matrix Quantum Mechanics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
4.2.3 Multiplication of operators in matrix form . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
4.2.4 Matrix form of TISE and TDSE . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
4.3 2-State System in Matrix Q.M. 61
4.3.1 Neutrino Oscillations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
4.4 Bra-Ket Notation 66

5 Hermitian Operators & Exp Thm II . . . . . . . . . . . . . . . . . . . . . . . . . . 67

5.1 Hermitian Matrices 67
5.2 Hermitian Operators 68
5.3 General Properties of General Linear, Hermitian Operators 71
5.3.1 Eigenvalues and Eigenstates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
5.3.2 Generalised Expansion Theorem and Generalised Measurement Theorem 72

6 The Simple Harmonic Oscillator . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75

6.1 Intro 75
6.2 SHO in Operator Formalism 78
6.3 Applications of SHO in operator formalism 82
6.3.1 Eigenfunctions reloaded . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82
6.3.2 Raising/lowering operators and normalisation . . . . . . . . . . . . . . . . . . . . . . . . . 83
6.4 Examples 83
7 Operators and Symmetries . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89
7.1 Compatible Operators and Simultaneous Eigenstates – A fresh look at
commuting operators 89
7.2 Commutators and Symmetries 92
7.2.1 Parity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92
7.2.2 Translations and Rotations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95
7.3 Comments on degeneracy 97
7.4 Heisenberg’s equations of motion 98
7.5 Ehrenfest’s Theorem 100

8 Angular Momentum in Q.M. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103

8.1 Intro 103

8.2 Orbital angular momentum ~L

b and its commutators 105
8.3 Angular momentum operators in spherical polar coordinates 109
8.4 The 3D Schrödinger equation in spherical polar coordinates 112
8.5 The radial equation for central potentials 115
8.6 Abstract angular momentum operators 119
8.7 Using raising/lowering operators to find spherical harmonics 127
1
8.8 Why ` = 2 does not work 130
8.9 More on particles in a central potential 131
8.10 Properties of the spherical harmonics 133
8.10.1 Explicit form of spherical harmonics up to ` = 3 . . . . . . . . . . . . . . . . . . . . . . . 135
8.11 More on (Hydrogen) Atoms 136

9 Spin Angular Momentum . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 141

9.1 Periodic table and the missing quantum number 141
9.2 Atoms in a magnetic field – Zeeman effect 144
9.3 Stern Gerlach experiment 147
9.3.1 Why Spin cannot be understood classically . . . . . . . . . . . . . . . . . . . . . . . . . . 151
9.4 Matrix mechanics and spin 153
9.4.1 Working out the matrix form of Sx , Sy , Sz . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 155
9.5 Pauli matrices and spin wave functions for j ≡ s = 1/2. 157
9.6 Spin quantisation along the x-axis: eigenstates of Sx 162
6

9.7 Rotated spin operators and their eigenfunctions 164

9.8 The Triple Stern Gerlach Experiment 167
9.9 Spin precession in a magnetic field 172
 1. Introduction

In this Chapter we will briefly compare some aspects of classical and quantum physics, and
set the stage by reviewing and extending the basic theorems of Quantum Mechanics.

1.1 Classical vs. Quantum

Classical Physics: in classical physics observables are function of the position ~x and velocity,
or equivalently the momentum ~p = m~x. ˙

The State of the System (particle) is described by ~x(t) which describes the time dependent
history of the particle = Trajectory. If this is given by reasonably smooth functions we can
˙
obtain the momentum (velocity) by differentiation: ~p(t) = m~x(t).

The Dynamical Equation is Newton’s 2nd law, which is a 2nd order differential equation:

m~x¨ = ~F(~x(t)) (1.1)

A crucial assumption here is that we can measure ~x and ~p simultaneously to arbitrary

precision at some time t0 :
Note, the Dynamical Equation being 2nd order requires us to fix two integration constants to
obtain a unique solution. Most naturally these are the position AND velocity at time t0 . BUT
this is only useful if, at least in principle, we can measure them simultaneously to arbitrary
precision!
Quantum Mechanics shatters these intuitive expectations:
• At the microscopic level position and momentum CANNOT be measured to arbi-
trary precision (Heisenberg uncertainty principle ∆x∆px ≥ h̄/2), this has nothing to do
with shortcomings of our experiments.
8 Chapter 1. Introduction

• More deeply, this tells us that there are fundamental limitations to what we can know
about a system. Many people do not like this, but in some sense it is a dumb question
to ask where AND how fast a particle is moving!
• At a microscopic level for example electrons around a nucleus do not follow trajecto-
ries. Atoms are not mini solar systems, although they are often drawn like that, it is a
completely wrong picture.
But we need to realise that without these fundamental changes of classical physics the world
around us cannot exist.
• The stability of atoms and molecules would be impossible. Consider the Hydrogen
atom.
Classical Picture: The electron orbiting around the nucleus producing the potential

e2
V =− (1.2)
4πε0 r

would constantly radiate off energy (light) until it collapses into the proton and releasing
in the process an infinity amount of energy because the minimum of this potential is
−∞! This is a complete disaster and completely wrong.
Quantum Physics: in this case the uncertainty principle guarantees that there is a finite
lowest energy state, called the groundstate, and the electrons actually occupy stable,
quantised orbits with energy

13.6eV
En ∼ − where we used the Principal Quantum number n = 1, 2, 3, . . . (1.3)
n2

Figure 1.1: Comparing classical and quantum description of hydrogen.

1.1 Classical vs. Quantum 9

• Quantum tunneling explains radioactivity – in classical physics there would not be any
radioactive isotopes because radioactive decay is classically forbidden – if the particle
does not have enough energy to climb over a potential barrier it will remain trapped.

• Double slit experiment, an example of “particle-wave" duality will be discussed later...

 10 Chapter 1. Introduction

1.2 Fundamental Postulates of Quantum Mechanics

The postulates of QM
• have no intuitive explanation.
• but they work, we have many, many experimental checks.
• no one could have made them up! They completely clash with classical expectations.

1.2.1 State of a system

Postulate 1 The State of a System is described by the wave function Ψ.
We write for one particle in 1D Ψ(x,t) and in 3D Ψ(~x,t).
For a system of several particles in 3D Ψ(~x1 , . . . ,~xN ,t).

Comments: Everything we can ever know about the system (more precisely an ensemble of
systems) is contained in Ψ. (No hidden hocus pocus!)
Ensemble of systems = large collection of identical systems in the same state. Ψ only tells
about repeated experiments of identical systems, and very rarely can tell us anything about
a single measurement of a system. A convenient way to realise an ensemble is a beam of
identical particles, for example protons at the LHC, with identical momentum and energy.
They are approximately described by the de Broglie wave function

Ψ ∼ ei(px−Et)/h̄ . (1.4)

Properties of Ψ:
1. Ψ has no direct physical meaning, it is a Complex Function
2. Singlevaluedness: since |Ψ(x,t)|2 is a physical probability density, it must have a
unique value. This is usually taken to imply that Ψ be single-valued, however we will
see that in the case of spin-half particles – electrons, protons, quarks, etc. – the wave
function is double-valued.
3. Boundary Condition All physical systems we study are localized in some finite region
of space, so that the probability of finding the system at spatial infinity, x → ±∞, is zero:

lim ψ(x) = 0 (1.5)

x→±∞

4. Normalisability (see Probability Interpretation): in 1D or 3D the wave function must

be square integrable
Z ∞ Z Z Z ∞
2
In 1D: dx|Ψ(x,t)| < ∞ In 3D: d 3 x|Ψ(~x,t)|2 < ∞ (1.6)
−∞ −∞

Question 1: Is Condition 3. sufficient to guarantee normalisability?

2 √ 2
Question 2: Are e−x , 1/ x or ex normalisable?
1.2 Fundamental Postulates of Quantum Mechanics 11

Figure 1.2: Two examples of non-normalisable wave functions (Conditions 3 & 4 violated):
A) diverges at x → ±∞ and is physically unacceptable, B) is finite at infinity but is physically
acceptable; it describes a (beam of) free particles.

5. Ψ must be a continuous function everywhere (No Jumps). Wave functions that jump
are unphysical, no ifs and buts!

Figure 1.3: This wave function is discontinuous at x0 and is, hence, physically unacceptable.

6. All derivatives of Ψ, ∂∂Ψx , ∂∂Ψy , ∂∂Ψz , must be continuous everywhere (No Kinks). If
this were not required we would allow 2nd derivatives to be infinite. As you know such
second derivatives appear in the Schrödinger equation in which every term is finite! The
only exception is when we allow the potential V to be infinite – this only happens in
idealised situations like the infinite square well.
12 Chapter 1. Introduction

Figure 1.4: Example of groundstate wave functions of finite and infinite square well. In the
case D) Ψ is continuous everywhere and has no kinks (finite potential); in the case E) there
are kinks at two points, these are allowed since at these points we enter a region with infinite
potential where Ψ = 0. Note that Ψ is again continuous everywhere.

Figure 1.5: Another example of a kinky wave function; here the kink is due to a δ function
potential at x0 . For illustration 1st and 2nd derivatives are indicated.
 1.3 Time Dependent & Independent Schrödinger Equations (TDSE & TISE) 13

1.3 Time Dependent & Independent Schrödinger Equations (TDSE &

TISE)
Postulate 2 The time evolution of the wave functions Ψ is described by the Time De-
pendent Schrödinger Equation (TDSE). In 3D this is

h̄2 ∂ 2 Ψ(~x,t) ∂ 2 Ψ(~x,t) ∂ 2 Ψ(~x,t)

 
∂ Ψ(~x,t)
− + + +V (~
x,t)Ψ(~
x,t) = ih̄
2m ∂ x2 ∂ y2 ∂ z2 ∂t

h̄2 2 ∂ Ψ(~x,t)
or, equivalently, − ∇ Ψ(~x,t) +V (~x,t)Ψ(~x,t) = ih̄ ,
2m ∂t
or even more compactly
2
b x,t) = ih̄ ∂ Ψ(~x,t) with H
HΨ(~ b = − h̄ ∇2 +V (~x,t) ,
∂t 2m

where we have introduced the Hamiltonian H.

b
In 1D this simplifies to
2 2
HΨ(x,t)
b = ih̄
∂ Ψ(x,t) b = − h̄ ∂ +V (x,t) .
with H
∂t 2m ∂ x2
The TDSE is the most important dynamical equation in QM. Given the wave function
at a fixed time t0 we can compute it for all later times! This is the analogue of Newtons 2nd
equation (or any classical equation of motion).
Another extremely important consequence is that if the potential happens to be time
independent
∂V
V (~x,t) = V (~x) or in other words =0 (1.7)
∂t
then we can derive the Time Independent Schrödinger Equation (TISE).
In this case we can write the wavefunction using separation of variables as
Ψ(~x,t) = ψ(~x) f (t) , (1.8)
and plugging this into the TDSE we obtain
 
  ∂ f (t)
Hψ(~x) f (t) = ψ(~x) ih̄
b , (1.9)
∂t
and after dividing both sides by ψ(~x) f (t) we arrive at
 
1 b  1 d f (t)
Hψ(~x) = ih̄ =E . (1.10)
ψ(~x) f (t) dt
The right hand side must be a constant since the first term in the last equation only depends on
~x (due to the time-independence of the potential V ) and the second term only depends on t.
14 Chapter 1. Introduction

The first terms, hence, leads to the Time Independent Schrödinger Equation (TISE from
hereon).

2
Hψ b = − h̄ ∇2 +V (~x)
b E (~x) = EψE (~x) with H (1.11)
2m

and the second term gives a trivial differential equation with solution

f (t) = e−iEt/h̄ . (1.12)

Mathematically the TISE has the form of an eigenvalue equation where E are the energy
eigenvalues and ψE (x) are the eigenfunctions. The corresponding solutions of the TDSE,
which are also called stationary states, are given by

ΨE (~x,t) = ψE (~x)e−iEt/h̄ (1.13)

and have very special properties as we will see.

Since, we usually have a large or infinite number of solutions & eigenvalues of the TISE,
we label them by an index n and denote them by ψn & En , respectively

b n (~x) = En ψn (~x) and Ψn (~x,t) = ψn (~x)e−iEnt/h̄

Hψ (1.14)

This is for the case of QM in 3D, in 1D you simply drop the vector symbols.
There are two superimportant consequences of this:
(A) the most general solution of the TDSE is a linear combination of the Ψn (stationary
states), and
(B) the energy eigenvalues En provide important physical predictions, in that they are the
only possible outcomes of measurements of the energy of the system. Again in classical
physics this would be a non-sensical question to ask “What are the allowed values of the
energy?", since we would simply say “Any real number is allowed!"
 1.4 The Probability (or Born) Interpretation of the Wave Function 15

1.4 The Probability (or Born) Interpretation of the Wave Function

We noted that the wave function has no direct physical meaning. So what can we do with Ψ?
The question of the physical significance of the wave function Ψ(x,t) caused considerable
difficulties in the early days of the development of quantum mechanics and even today is
not completely settled. It was Max Born (1926) who came up with the currently accepted
and hugely successful probability interpretation. It is this interpretation which is at the heart
of many of the conceptual difficulties students (and others) have in understanding quantum
mechanics.

Postulate 3 The Probability/Born Interpretation:

|Ψ(x,t)|2 dx is the probability that a measurement at time t of the particle’s position gives
a value lying between x and x + dx - Fig.1a.

or, in 3-dimensions:

|Ψ(r,t)|2 d 3 x (where d 3 x ≡ dxdydz ≡ dV ) is the probability that a measurement at time

t of the particle’s position gives a value lying in the volume element dV = d 3 x at position ~x -
Fig.1b.

This is the Born interpretation or the probability interpretation of the wave function.

dx
-
-x
6 6
x x + dx

Figure 1a: Illustrates the Born interpretation in 1-dimension: the probability of finding the
particle in the infinitesimal region between x and x + dx at time t is |Ψ(x,t)|2 dx.

z
6  Volume element dV

 with sides dx, dy, dz.





r









-y

x
 16 Chapter 1. Introduction

Figure 1b: Illustrates the Born interpretation in 3-dimensions: the probability of finding
the particle in the infinitesimal volume dV at position r at time t is |Ψ(r,t)|2 dV .
But for this we to work we need the wave functions to be normalisable – see earlier –
because then we can demand the wave function to be normalised
Z Z Z
3D |Ψ(~x,t)|2 dxdydz = 1 (1.15)
Z
1D |Ψ(x,t)|2 dx = 1 (1.16)

where we assumed that all unspecified integration domains, by default, run from −∞ to +∞.
Comments:
1. Normalisation to one has the simple interpretation that the particle must be somewhere
and the probability to find the particle somewhere inside the infinite interval/volume
must be equal to one.
2. In 1D we can define the probability to find the particle in a finite region between x = a
and x = b
Z b Z b
2
Pab = |Ψ(x,t)| dx ≡ P(x,t)dx , (1.17)
a a

where we introduced the probability density P(x,t) = |Ψ(x,t)|2 . This can also be
generalised to 3D where integrate over finite subspaces of finite volume to find the
probability to find the particle inside that particular subspace.
3. Notice that wave functions depend on time but the total probability after integrating over
all of space is constant and hence conserved. The interpretation is that the probability
of a particle does not fizzle out or disappear somehow – the wave function can slosh
around in space but the particle will always be somewhere. These are nice words, but is
this actually true given the time dependence of Ψ? We will look at this now in 1D...

1.4.1 Conservation of Probability

In the following we want to show that1

∂ Ψ∗
Z ∞ 
d ∞
Z
2 ∗∂Ψ
0= |Ψ(x,t)| dx = Ψ+Ψ dx (1.18)
dt −∞ −∞ ∂t ∂t
| {z }
W

Now use the TDSE:

(
ih̄ ∂ 2 Ψ
h̄2 ∂ 2 ∂Ψ i
 
− +V (x) Ψ = ih̄
∂Ψ
⇒ ∂t = 2m ∂ x2 − h̄ V Ψ , (1.19)
2m ∂ x2 ∂ Ψ∗ ih̄ ∂ 2 Ψ∗ i ∗
∂t = − 2m ∂ x2 + h̄ V Ψ
∂t
1 In order to save space we will drop the arguments (x,t) of Ψ in various places.
 1.4 The Probability (or Born) Interpretation of the Wave Function 17

to write
∂ Ψ∗
 
∗∂Ψ
W = Ψ+Ψ (1.20)
∂t ∂t
ih̄ ∂ 2 Ψ∗ i ih̄ ∂ 2 Ψ i
   
∗ ∗
= − + VΨ Ψ+Ψ − VΨ
2m ∂ x2 h̄ 2m ∂ x2 h̄
2 ∂ 2 Ψ∗
 
h̄ ∗∂ Ψ
= i Ψ − Ψ (1.21)
2m ∂ x2 ∂ x2
∂ Ψ∗
 
h̄ ∂ ∗∂Ψ ∂
= i Ψ − Ψ ≡ − J(x) , (1.22)
2m ∂ x ∂x ∂x ∂x

where we introduced the probability current

∂ Ψ∗
 
h̄ ∗∂Ψ
J(x) = −i Ψ − Ψ . (1.23)
2m ∂x ∂x

So we find
d ∂J
Z ∞ Z ∞
2
|Ψ(x,t)| dx = − dx = − [J(x)]+∞
−∞ = 0 , (1.24)
dt −∞ −∞ ∂x

IF the wavefunction vanishes Ψ = 0 for x → ±∞. But this is just our minimum requirement for
2
R
any normalisable wavefunction! So we conclude that indeed the total probability |Ψ| dx
is conserved (time independent). The same proof can easily be generalised to 3D.

1.4.2 More on the Probability Interpretation of Ψ

More comments, in order to address some VERY common misconceptions:
• Ψ usually cannot predict the outcome of a single measurement of position, or more
strongly it tells us nothing about the position of a single particle (no trajectories)! In
some sense the particle has not made up its mind where it is, until the act of measurement
forces the particle to make a decision and it will be found at position x at time t with
probability |Ψ(x,t)|2 . The last bit is the only piece of information quantum mechanics
gives us.
• So is there any way we can reconstruct Ψ or |Ψ|2 from experiment? Yes, but we will
have to make a very large number of experiments on identical systems (ensemble of
systems) or repeatedly prepare the system in the same state to rerun the experiment over
and over again. In principle this has to be done an ∞ number of times to get a precise
picture of the wave function. Compare this classical where you just need to watch the
path of a single particle once and you know its state.
• Therefore, it is imprecise/wrong to say a particle/system is in the state Ψ. We will still
often say but what we really mean is that Ψ describes an ensemble of systems. This
ensemble interpretation is an important but also confusing part of the Born/Probability
Interpretation.
18 Chapter 1. Introduction

What else can we predict or know?

Expectation values of Physical Observables A

b (more on their definition in the next chapter):
Z
hAi = Ψ∗ (x,t)AΨ(x,t)dx
b , (1.25)

is the expectation value of the operator (corresponding to some observable) A b for a given
state and gives the quantum mechanical average of many measurements of the same system
(ensemble). Again you would need to do this an ∞ number of times to get the exact result.
Nobody has so much time so there will always be a fundamental error in your result which is
not due to your experimental set-up being impricise (this error comes on top of the quantum
mechanics uncertainty).
There is also some measure how much deviation from the expectation value you can expect
in different measurements. This is called the uncertainty (variance) which is defined as
q
∆A = hA b2 i − hAi
b2, (1.26)

where examples of A
b include

∂ h̄2 2
Position: xb = x , Momentum: pbx = −ih̄ , Energy H = − ∇ +V (~x) .
b (1.27)
∂x 2m

Figure 1.6: Illustrating large/small uncertainty.

Example: Consider QM in 1D and assume the particle is in the following idealised wave
function

N for 0 ≤ x ≤ L ,
Ψ(x, 0) = , (1.28)
0 for x < 0 or x > L .
Now, find the correct normalisation constant N assuming it is real, positive, and compute the
expectation values hxn i and the uncertainty ∆x.
To normalise consider the integral
Z Z L
2
1= |Ψ(x, 0)| dx = |N|2 dx = L|N|2 , (1.29)
0
1.4 The Probability (or Born) Interpretation of the Wave Function 19
√
hence we find (using the assumptions) that N = 1/ L. Also plot the wave function and state
what is wrong with it in relation to our discussion on general properties of wave functions?
Next, we compute the expectation values
x=L
1 xn+1 Ln
Z Z L 
n ∗ n 2 n
hx i = Ψ x Ψdx = |N| x dx = = , (1.30)
0 L n + 1 x=0 n + 1

therefore hxi = L/2, hx2 i = L2 /3 and

r
1 1 L
q
∆x = hx2 i − hxi2 = L2 ( − ) = √ (1.31)
3 4 12
 20 Chapter 1. Introduction

1.5 Superposition Principle

The reason that QM has a lot to do with waves is related to the fact that the Hamiltonian H
b is
a Linear Operator and hence the TDSE is a linear equation.
b 1 Ψ1 + c2 Ψ2 ) = c1 HΨ
H(c b 1 + c2 HΨ
b 2 due to linearity (1.32)

= c1 ih̄
∂ Ψ1
+ c2 ih̄
∂ Ψ2 b 1,2 = ih̄ ∂ Ψ1,2
using HΨ (1.33)
∂t ∂t ∂t
∂
= ih̄ (c1 Ψ1 + c2 Ψ2 ) (1.34)
∂t
In words: IF Ψ1 and Ψ2 are solutions of the TDSE, for example they could be the Stationary
States reviewed earlier, then an arbitrary linear combination
Ψ = c1 Ψ1 + c2 Ψ2 (1.35)
with complex constants c1,2 is guaranteed to be a solution of the TDSE as well.
This is the essence of the Superposition Principle and typical of wave behaviour.
This basic mathematical property leads to profound and mysterious physical consequences:
• Double slit experiment
• Schrödinger cat
• Tunneling through forbidden barriers
• Ghostly instantaneous interaction over large distances
• Brexit (actually it seems the wave function has collapsed by now)
• ...

1.5.1 Double Slit Experiment

Comments and Interpretation

1.5 Superposition Principle 21

• The result contradicts classical intuition, which would assume that we always know
where the particles are (trajectories). In this picture the individual probabilities for the
two paths would add up P = P1 + P2 6= 12 |Ψ1 + Ψ2 |2 , and NOT the wave functions as in
QM Ψ = √12 (Ψ1 + Ψ2 ) (interference).
• Born/Probability Interpretation:
(A) Ψ does not predict through which slit a particle goes. Need to check by measurement.
(B) So are particles big, fluffy, wavy objects? No the detector observes very sharp dots.
(C) Do particles interfere with each other? No, we can work with very low intensity
beams so that only a single particle at a time passed through the apparatus. Still after
a while interference patterns emerge. (See this webpage explaining the experimental
set-up for electrons and containing a nice illustrating this point. )
(D) Crucially this only works if particles are not disturbed! If we actually put detectors
at slit 1 or 2 to see where the particles went the interference patter on the screen is ruined!
Why? The Uncertainty Principle ∆x∆px ≥ h̄/2 implies that if we measure position then
∆x is small but at the same the momentum is changed in an unpredictable way (∆px is
increased). But for coherence we need a monochromatic beam of particles (all particles
have same momentum) and this property is lost if we try to find the particle’s position.
• Wave-particle duality: depending of what measurement we perform we make the
wave-like property (undisturbed propagation through interferometer, interference) or
the particle property manifest (checking if particle went through slit 1 or 2, particle
detection on screen).
• This important experiment was originally performed for light (at that time thought to
be a wave and not made of photons), and later for electrons, protons and recently for
huge molecules made of 60 carbon atoms (C60 – so called buckyballs or Buckminister-
fullerenes). Check out this link for the experiment with buckyballs.
 22 Chapter 1. Introduction

1.6 Optional: Even More on the Born Interpretation of Ψ

This appendix contains a more detailed discussion of the probability interpretation which is
not covered in class; nevertheless I highly recommend to read it!
We expressed the Born interpretation of the wave function in terms of an infinitesimal
portion dx of the x-axis; the probability of finding the particle at time t somewhere between
two points a and b separated by a finite distance is simply the ‘sum’ of the probabilities,
Z b
Pab (t) = |Ψ(x,t)|2 dx (1.36)
a

If a wave function Ψ(x,t) describes the state of a single particle, then we are certain to
find the particle somewhere, i.e. the probability for finding it anywhere on the x-axis is 1:
Z ∞
|Ψ(x,t)|2 dx = 1 , (1.37)
−∞

and in three dimensions this becomes,

Z ∞Z ∞Z ∞
|Ψ(r,t)|2 d 3 x = 1 (1.38)
−∞ −∞ −∞

This is known as the normalisation of the wave function: any proper single-particle wave
function must be normalised (to 1).
The Born interpretation implies a profound change in our concept of measurement. The
wave function is meant to encode all that we can possibly know about the system in quantum
mechanics; the Born interpretation tells us that a measurement of position does not lead to a
unique or definite result, but only to one of an infinite set of possible results. What is definite
is the probability of obtaining a particular result; it’s just like throwing a dice: each throw of
a dice yields a definite result, although we only know after the throw what that result is. A
particular result (say a 6) has a probability of 1/6, and this is all we can say before the throw
is made. Similarly, if we prepare the system in the quantum state Ψ(x,t) then we can only
predict the probability of a given result for a position measurement (|Ψ(x,t)|2 dx for the result
to lie between x and x + dx); but once the measurement has been made then we know where
the particle is at time t of the measurement. Suppose we had computed the wave function
and wished to check our calculation against experiment. The Born interpretation tells us
that we can only use our wave function to predict probabilities. These probabilities can only
be determined experimentally by doing very many (in principle infinitely many) repetitions
of the measurement, each time preparing the system anew in the same state Ψ(x,t) before
repeating the measurement. This corresponds to the process of checking whether a dice is
biassed, i.e. what the probabilities are for the various outcomes and whether they are equal:
we would have to throw the dice infinitely many times and compare the relative frequencies of
the different outcomes; these are proportional to the probabilities of the different outcomes
and if they are equal, the dice is unbiassed. Preparation of the system is like throwing the dice;
measuring position is like looking at the dice when it’s landed and noting the outcome. Just as
with any repeated measurement, we can compute the average of the individual measurements:
1.6 Optional: Even More on the Born Interpretation of Ψ 23

the weight of each value of x is its probability, |Ψ(x,t)|2 dx, and the average is obtained by
summing over all the possibilities,
Z ∞
hxi ≡ x|Ψ(x,t)|2 dx (1.39)
−∞

Note that we did not divide by the integral of |Ψ|2 because the wave function is normalised.
hxi is called the expectation value for x in quantum mechanics; it is usually written in an
equivalent form resembling that for other quantum mechanical variables which are repre-
sented by operators where the order of factors is important (in other words “sandwich" the
observable/operator between Ψ∗ and Ψ in the integrand):
Z ∞
hxi ≡ Ψ ∗ (x,t) x Ψ(x,t) dx (1.40)
−∞

The measurement process can either be performed as described above or by preparing an

infinite number of identical replicas of the system, each in the same state Ψ, and measuring
the position of every one; this collection of systems is known as an ensemble, and the
interpretation of the wave function I have described is known as the ensemble interpretation
of quantum mechanics. (A practical example of an ensemble is a beam of identical particles,
all in the same state.) This is usually called the Copenhagen interpretation because it was
the one developed by Niels Bohr and his many collaborators, especially Heisenberg, in
Copenhagen during the 1920’s. You should particularly notice that this interpretation is
rather careful not to identify Ψ with the state of a single system, but with an ensemble; Ψ
is the outcome of a preparation process and is our tool for computing the probabilities of
various results for individual measurements, and the expectation values obtained in ensemble
measurements.
In an ensemble measurement of the particle’s position we measure many different values
of x and compute their average hxi. |Ψ(x,t)|2 dx gives the probability for the result of a single
one of these measurements lying between x and x + dx; the expectation value
hxi is then obtained by averaging all the individual results. Since individual results differ
from each other, we would like to find a measure of the spread of these values about the mean.
This is given by the variance or uncertainty ∆x: 2

q
∆x = hx2 i − hxi2 (1.41)
2 From laboratory measurements you may be familiar with the definition of the squared ‘error’ as the result of
taking the difference between each measurement and the mean, squaring, and then averaging:

(∆x)2 = h(hxi − x)2 i

= h(hxi2 − 2xhxi + x2 )i
= hxi2 − 2hxihxi + hx2 i
= hx2 i − hxi2

This is therefore equivalent to the definition given in the text.

24 Chapter 1. Introduction

where
Z +∞
hx2 i = Ψ ∗ x2 Ψ dx (1.42)
−∞

We can interpret this as saying that ‘the position of the particle is’
hxi± ∆x. I put this statement in quotation marks because it is not really the position of a
single particle, but the average position of an ensemble of particles, each prepared in the state
Ψ. We can usefully think of our particle in state Ψ as ‘smeared out’ over this region
hxi± ∆x, but you should be careful not to take this interpretation too literally.
Physicists usually loosely identify Ψ as the state of an individual system; this works well
enough in most situations, but is also a source of misconceptions in others.
Let me illustrate the trouble caused by saying the particle ‘is’ literally in the state Ψ. Take
the example of Young’s two-slit experiment using an electron beam. Let Ψ1 be the wave
function for the electron arriving at position x on the screen after going through slit 1; Ψ2 for
going through slit 2. To explain why an electron beam √ produces interference at the screen we
need to write the wave function as Ψ = (Ψ1 + Ψ2 )/ 2, representing the fact that it is equally
likely to go through either slit, but we don’t know which (think of tossing a coin). At the
screen we measure the number of electrons arriving at each point, which is proportional to the
probability of an electron arriving there; 3 this probability is
1
|Ψ|2 = |Ψ1 |2 + |Ψ2 |2 | + 2ℜe (Ψ1∗ Ψ2 ) . (1.43)
2
It is the cross-term - the real part, ℜe - which gives the constructive and destructive
interference observed in the experiment; without the linear combination of both Ψ1 and Ψ2
there would be no interference. We are tempted to conclude that the state of a particle just
before it arrives at the screen ‘is’ in some sense both Ψ1 and Ψ2 ; has the particle gone through
both slits, half through slit 1; the other half through slit 2? Surely not! The electron is
not a particle which can be broken up in that way: we have never seen such a division in
any experiment and have good reasons to believe that this is impossible - lepton number
conservation, quantisation of electric charge, no known particles with mass anywhere near
one-half the electron mass. 4
What then should we say? In the Copenhagen version of quantum mechanics we refuse
to identify Ψ as the actual state of a particular particle and so we avoid saying that it went
through both slits or either slit; instead we take Ψ to represent the result of a preparation
process (sending the particle into the apparatus) which tells us the possible outcomes and
3 Notice that the use of a beam amounts to an ensemble measurement of position on the screen; if we had
sent just one electron through the apparatus it would have arrived somewhere on the screen, but on its own that
would have told us virtually nothing about the probability distribution |Ψ|2 , and we would not have noticed the
interference. Only by passing many electrons through the apparatus, either in an intense beam or one at a time,
can we get to see the interference pattern.
4 Also, if we tried to check whether the electron went through one or both slits by putting counters at the

slits to detect the electron as it passed through on its way to the screen, quantum mechanics - in the form of
Heisenberg’s uncertainty relation - would show that the pattern on the screen is so smeared out as to remove any
trace of interference.
1.6 Optional: Even More on the Born Interpretation of Ψ 25

probabilites of a measurement process (detection at the screen); we cannot say which slit the
particle went through because we did not carry out an appropriate measurement to investigate
this; indeed we avoided doing so in order to measure an interference pattern on the screen. To
experimentally determine detailed information about the wave function we must carry out an
ensemble measurement, and so Ψ represents the entire ensemble rather than a single member
of the ensemble. In practice we all loosely think of Ψ as representing the state of a single
particle - its actuality - although we have to be careful when doing so. You should be aware that
this is a controversial aspect of quantum mechanics: there is so far no experiment contradicting
the Copenhagen interpretation - and I believe it derives its power from the fact that it is a
rather conservative interpretation - but it makes many people uneasy. Both experimental and
theoretical research continues to explore this crucial area of quantum mechanics.
 2. Operators and Commutators

2.1 Quantum Mechanical Observables

In quantum mechanics everything is different and things that had a clear meaning in classical
mechanics, like position and velocity (and hence momentum) suddenly are replaced by abstract
objects that in general have no direct meaning, namely Quantum Mechanical Operators. We
do not have any intuition why this is and there is no way that this can be derived, hard
experimental facts forced upon us to formulate our fundamental laws of nature in terms of
operators, wave functions etc. A very important postulate is:
Postulate 4 In quantum mechanics every physical observable corresponds to a linear
and hermitian operator. Such operators can be a function F(~x), a differential operator
or a matrix, but in any case they must have these two properties in order to be a potential
observable, something we can measure in an experiment.

Recall linearity means

b 1 ψ1 + c2 ψ2 ) = c1 Aψ
A(c b 1 + c2 Aψ
b 2 (2.1)

Note that for example O b1 Ψ = Ψ2 or O b2 Ψ = sin(Ψ) are not linear operators.

The concept of hermitian operator you should know from previous mathematics courses
but we will review this a bit later in the course.
Examples of linear and hermitian operators (observables):
• The Hamiltonian H b is a linear operator ⇒ TDSE is a linear equation ⇒ Superposition
Principle
• Position: xbψ(x,t) = xψ(x,t)
• Momentum: pbx ψ(x,t) = −ih̄ ∂∂x ψ(x,t)
 
x =~x, ~bp = −ih̄ ∂ , ∂ , ∂ = −ih̄~∇
• 3D: ~b ∂x ∂y ∂z
28 Chapter 2. Operators and Commutators

For classical observables there exists a cute recipe to convert them into a Quantum
Mechanical operator
Postulate 5 Operator Postulate: Every classical observable A(~x,~p) is a function of posi-
tion and momentum. In order to obtain the corresponding Quantum Mechanical operator
simply replace position and momentum by the corresponding operators as:

CLASSICAL QUANTUM MECHANICS

A(~x,~p) −→ A(~ b x, −ih̄~∇)
b x, ~bp) = A(~ (2.2)

However, there is an important caveat: we will see later that there exist quantum mechanical
operators that have no classical counterpart, for example Spin. But let us ignore this for now.

Examples
• In classical mechanics the total energy or Hamiltonian of a particle is the sum of kinetic
and potential energy and we can instantly derive the quantum mechanical Hamiltonian
that appears in the TDSE and TISE
2
~p2 b = − h̄ ~∇2 +V (~x)
E = K +V = H(~x,~p) = +V (~x) −→ H (2.3)
2m 2m
• Angular Momentum (AM): classical ~L =~x ×~p becomes in Q.M. ~Lb =~x × ~bp = −ih̄~x × ~∇,
or in components:  
∂ ∂
Lx = ypz − zpy −→ Lx = y pbz − z pby = −ih̄ y − z
b (2.4)
∂z ∂y
 
∂ ∂
Ly = zpx − xpz −→ Ly = z pbx − x pbz = −ih̄ z − x
b (2.5)
∂x ∂z
 
∂ ∂
Lz = xpy − ypx −→ L
bz = x pby − y pbx = −ih̄ x − y (2.6)
∂y ∂x

Now in general operators have no direct meaning, so what do operators do for a living?
• They act on (wave)functions and in general produce new ones,
Ψ(x,t) −→ AΨ(x,t)
b = χ(x,t)
where A b is a differential operator.
• But sometimes you get back the same function – this is then an eigenvalue equation
like the TISE
b n = En ψn
Hψ
where En are the energy eigenvalues and ψn are the energy eigenstates.
• Sometimes operators can be Hermitian matrices (see later, for example Spin)
• Sometimes they commute, sometimes they do not commute. So we need to learn to
check!
 2.2 Commutators 29

2.2 Commutators
Definition 2.2.1 Commutator: Similar to matrices, in the case of operators it matters in
which order they are applied to a wave function. Hence, we define the commutator of two
operators as

[A, b ≡A
b B] bBb − BbA
b (2.7)

If [A, b = 0 we say that the two operators commute,

b B]
if [A, b 6= 0 we say that the two operators do not commute.
b B]

Properties of the commutators:

1.

[cA, b = c[A,
b B] b B]
b

2.

[A
b + B, b = [A,
b C] b + [B,
b C] b C]
b

3. Leibniz rule for composite operators A

bBb

[A
bB, b = A[
b C] b B, b + [A,
b C] b C]
b Bb

and similarly

[C,
bA b = [C,
bB] b Bb + A[
b A] b C,
b B]
b

Note, the ordering of the expression matters here and that on the RHS of both equations
b always appears before B,
A b because it appears in the same order on the LHS!

There are two methods to compute commutators of specific operators; (A) in the traditional
fashion you act on some arbitrary wave function, but we will also introduce a niftier method (B)
in which we reuse/recycle known commutation relations of position & momentum operators
using the Leibniz rule to find commutation relations of more complicated, composite operators.

Most importantly let us rederive (you should remember the result) [x, pbx ] with method (A),
[x, pbx ]Ψ(~x) = x( pbx Ψ) − pbx (xΨ) (2.8)
∂Ψ ∂
= −ih̄x − (−ih̄) (xΨ) (2.9)
∂x ∂x
∂Ψ ∂Ψ
= ih̄(−x +Ψ+x ) (2.10)
∂x ∂x
= ih̄Ψ(~x) (2.11)
 30 Chapter 2. Operators and Commutators

Only now we can drop Ψ because this is true for any possible wave function Ψ, and we find
the operator identity (Warning, if you perform such derivations without the wave function all
marks are subtracted)

[x, pbx ] = ih̄ , (2.12)

and we can immediately generalise the proof to the other position and momentum operators

[x, pbx ] = [y, pby ] = [z, pbz ] = ih̄ (2.13)

These are so important you need to remember these commutation relations. What about all the
other possible commutation relations between all position and momentum operators. First of
all, position operators commute with each other since they are only numbers and momentum
operators all commute with each other since derivatives commute

[x, x] = [x, y] = . . . = [ pbx , pbx ] = [ pbx , pby ] = . . . = 0 (2.14)

and finally the position operator in one direction always commutes with the momentum
operator in a different directions because for example ∂∂ xy = 0 so we also have

[x, pby ] = [x, pbz ] = [y, pbx ] = . . . = 0 (2.15)

In summary: the only non-vanishing commutation relations between position and momentum
operators are (2.13). And these are the only ones you ever need to remember since more
general operators are built out of position and momentum operators after all!!

2.2.1 Examples
First we use the nifty method. Main idea: reduce, reuse, recycle!!
1. Using the Leibniz rule:

[x2 , pbx ] = [|{z}

xx , pbx ] = x[x, pbx ] + [x, pbx ]x = x(ih̄) + (ih̄)x = 2ih̄x
|{z} | {z } | {z }
=A
bBb Cb A[
b B,
b C]
b [A,
b C]
b Bb

2. Now we will recycle the result from 1. to compute

[x2 , pb2x ] = [x2 , pbx ] pbx + pbx [x2 , pbx ] and now using the result from 1.
= 2ih̄x pbx + 2ih̄ pbx x = 2ih̄(x pbx + pbx x) (2.16)
Warning! many now try to simplify this further by swapping pbx and x in the 2nd term,
but why is this not allowed?
3. Using the Operator Postulate we found L bz = x pby − y pbx . Let us now compute its commu-
tator with x:
bz ] = [x, x pby ] − [x, y pbx ] = [x, x] pby + x [x, pby ] − [x, y] pbx − y [x, pbx ] = −ih̄y (2.17)
[x, L
|{z} | {z } |{z} | {z }
=0 =0 =0 =ih̄
2.2 Commutators 31

4. I leave it as an exercise to show that, for any function f (x),

∂ f (x)
[ f (x), pbx ] = ih̄ . (2.18)
∂x
Here you will need to use the old-fashioned method!
 32 Chapter 2. Operators and Commutators

2.3 Heisenberg’s Generalised Uncertainty Principle (HGUP)

An important application of commutators is Heisenberg’s Generalised Uncertainty Princi-
ple: If we measure two observables A
b and B,
b then the product of the uncertainties of these two
measurements
q q
∆A = hA i − hAi and ∆B = hBb2 i − hBi
b 2 b 2 b2 (2.19)

obeys the inequality

1 b b
∆A∆B ≥ h[A, B]i (2.20)
2
b = Ψ∗ [A,
R
where h[A,
b B]i b B]Ψdx
b in general depends on time and the state Ψ of the system.

Example: Now of courseRyou are familiar with the special case A b = x and Bb = pbx . In this
case h[A, ∗
b = h[x, pbx ]i = ih̄Ψ Ψdx = ih̄ if Ψ is normalised. Furthermore, |ih̄| = h̄ so that
b B]i
we find
h̄
∆x∆px ≥
2
and similarly

h̄ h̄
∆y∆py ≥ and ∆z∆pz ≥
2 2
But on the other hand, since for example [x, pby ] = 0, then whenever the position and momentum
are in different direction then the RHS will vanish and

∆x∆py ≥ 0 .

Interpretation of the HGUP:

1. If two operators do not commute [A, b 6= 0 then the corresponding observables
b B]
cannot be measured simultaneously to arbitrary precision. This is a fundamental
quantum mechanical property – there is nothing wrong with the experiment.
2. Because of that we also say that x and pbx are complementary properties, because a
particle does not have both a sharply defined position AND momentum.
In principle we can measure the position to arbitraryp precision, ∆x = 0, but then we
know nothing about its momentum ∆px → ∞. ψ(x) ∝ δ (x).
On the other hand there are particle waves with a sharply defined momentum, for
example a de Broglie wave function ∼ eixpx /h̄ obeys the momentum eigenvalue equation

pbx eixpx /h̄ = px eixpx /h̄

and, therefore, it momentum uncertainty ∆px = 0. But then we know nothing about its
position ∆x → ∞.
2.3 Heisenberg’s Generalised Uncertainty Principle (HGUP) 33

3. The limit to the combined magnitude of the uncertainties is set by the righthand side
(RHS) of the HGUP (2.20). For position and momentum this limit is h̄/2.

Commutators have many more applications and implications (see later). For example if
[A, b = 0 then we cannot only measure (in principle) both observables simultaneously to
b B]
arbitrary precision, but there also exist simultaneous eigenstates:
b = aψ and Bψ
Aψ b = bψ .

From the uncertainty principle we know that we cannot have perfect knowledge of the
position and momentum (velocity) of a particle. An interesting question is what is the optimal
wave-function that minimises the uncertainty product
h̄
∆x∆px = .
2
It turn out this minimum uncertainty wavefunction is given by a Gaussian wave-packet:
 1
2 4 2 /L2
ψ(x) = e−x (2.21)
πL2
2 /a2 2 2
Using e−2x dx = a π/2 and x2 e−2x /a dx = a3 /4 π/2 one can show
R p R p

L2 L
hxi = 0 , hx2 i = ⇒ ∆x = (2.22)
4 2
2
h̄ h̄
h pbx i = 0 , h pb2x i = 2 ⇒ ∆x = (2.23)
L L
and hence ∆x∆px = h̄/2. Notice that this is a one-parameter (L) of minimal uncertainty wave
functions; if we make L small this becomes a very narrow, sharply peaked wave function while
if L is large it becomes very “smeared out".
 34 Chapter 2. Operators and Commutators

2.4 Proof of the uncertainty principle for position and momentum

Assume a particle is in the normalised state Ψ(~x,t) — often we will just call it Ψ. To make
things a bit simpler we also assume that hxi = 0 and h p̂x i = 0. Hence we have in this case
q q
∆x = hx i − hxi = hx2 i ,
2 2

q q
∆px = h p̂2x i − h p̂x i2 = h p̂2x i .
Now we introduce the following function
∂Ψ
ϕ = x̂Ψ + iλ p̂x Ψ = xΨ + λ h̄
∂x
and the proof starts with the simple observation that the following integral must be non-
negative!
Z +∞ Z +∞
2
|ϕ| dx = ϕ ? ϕdx ≥ 0 .
−∞ −∞

We now need to massage the integral

∂ Ψ?
Z +∞ Z +∞   
? ? ∂Ψ
ϕ ϕdx = dx xΨ + λ h̄ xΨ + λ h̄
−∞ −∞ ∂x ∂x

∂ Ψ?
Z +∞   
2 2 ? ∂Ψ ∂Ψ
= dx x |Ψ| + λ h̄Ψ x + λ h̄ xΨ + λ h̄ .
−∞ ∂x ∂x ∂x
Now we need to use integration by parts (IBP) on the last term. Recall IBP means
Z +∞ Z +∞
∂ f (x) x=+∞ ∂ g(x)
dx g(x) = f (x)g(x) − dx f (x) .
−∞ ∂x x=−∞ −∞ ∂x
Now in QM we always deal with nice (wave) functions that are normalisable which in particular
means all functions we integrate vanish when x → ±∞ so we can drop the boundary term and
arrive at
Z +∞ Z +∞
∂ f (x) ∂ g(x)
dx g(x) = − dx f (x) .
−∞ ∂x −∞ ∂x
This is a standard trick we use over and over again: it allows us "to move around derivative
operators" at the cost of picking up a minus sign.
So back to our earlier expression we find
Z +∞   
2 2 ? ∂Ψ ? ∂ ∂Ψ
= dx x |Ψ| + λ h̄Ψ x − λ h̄Ψ xΨ + λ h̄
−∞ ∂x ∂x ∂x
Z +∞  2 
2 2 ? ∂Ψ ? ∂ 2 2 ?∂ Ψ
= dx x |Ψ| + λ h̄Ψ x − λ h̄Ψ (xΨ) − λ h̄ Ψ
−∞ ∂x ∂x ∂ x2
2.4 Proof of the uncertainty principle for position and momentum 35

Now we use that

∂ i ∂2 1
= p̂x and 2
= − 2 p̂2x
∂ x h̄ ∂x h̄
to write this as
Z +∞
dx x2 |Ψ|2 + λ iΨ? x p̂x Ψ − λ iΨ? p̂x xΨ + λ 2 Ψ? p̂2x Ψ
 
=
−∞

noticing that the middle term contains the famous commutator

 
Z +∞
= dx x2 |Ψ|2 + λ iΨ? [x, p̂x ] Ψ + λ 2 Ψ? p̂2x Ψ
−∞ | {z }
=ih̄

we finally find (using usual definition of expectation values):

hx2 i − λ h̄ + λ 2 hp2x i = (∆x)2 − λ h̄ + λ 2 (∆px )2 ≡ F(λ ) ≥ 0 .

In the final step we have to find the minimum of F

dF h̄
= −h̄ + 2λ (∆px )2 = 0 → λmin =
dλ 2(∆px )2

and plug back in

h̄2
F(λmin ) = (∆x)2 − ≥0.
4(∆px )2

From this we find

2 h̄2 2 2 h̄2 h̄
(∆x) ≥ 2
→ (∆x) (∆px ) ≥ → ∆x∆px ≥
4(∆px ) 4 2
 3. The Expansion Theorem (Exp Thm I)

3.1 Energy Eigenstates are Stationary States

Energy eigenstates are particularly important in applications of quantum mechanics. Here
we discuss some of their properties, which we will see later in the course in a more general
form. This section and the following ones also provide practice at some of the mathematical
manipulations common to many calculations in quantum mechanics.
First we recall that energy eigenstates are solutions of both the TDSE and TISE and have
the form

ΨE (x,t) = ψE (x)e−iEt/h̄

Since this is common to many other examples, we adopt the more convenient notation:

Ψn (x,t) = ψn (x)e−iEnt/h̄ (3.1)

Since the complex conjugate wave function has the opposite sign on the time-dependent
exponential,

Ψn∗ (x,t) = ψn∗ (x)e+iEnt/h̄ (3.2)

the time-dependence drops out of the probability density because these exponentials cancel,

P(x,t) := |Ψn (x,t)|2 (3.3)

= ψn (x) ∗ ψn (x)e+iEnt/h̄ e−iEnt/h̄ (3.4)
= |ψn (x)|2 (3.5)
= P(x), independent of t (3.6)
38 Chapter 3. The Expansion Theorem (Exp Thm I)

It also follows that both Ψn (x,t) and ψn (x) are normalized to 1:

Z +∞
1 = Ψn∗ (x,t)Ψn (x,t) dx (3.7)
−∞
Z +∞
= ψn∗ (x)ψn (x) e+iEnt/h̄ e−iEnt/h̄ dx (3.8)
−∞
Z +∞
= ψn∗ (x)ψn (x) dx (3.9)
−∞

States whose probability density |Ψn (x,t)|2 is time-independent are known as station-
ary states. Furthermore, the time-dependence drops out of all expectation values for time–
independent physical observables, O,
b again, because the time–dependent exponentials cancel
for energy eigenstates:
Z +∞
hOi
b t := Ψn∗ (x,t)OΨ
b n (x,t) dx (3.10)
−∞
Z +∞
= ψn∗ (x)Oψ
b n (x) e+iEnt/h̄ e−iEnt/h̄ dx (3.11)
−∞
Z +∞
= ψn∗ (x)Oψ
b n (x) dx (3.12)
−∞
= hOi,
b independent of t (3.13)

In the above I have continued to use the notation ψn∗ (x) for the space–dependent part of
the eigenstate even though it is real. The reason for calling these states stationary is that their
momentum expectation value is zero: they do not move, at least on the average. The proof
depends on the fact that the x-dependent part of the wave function is real, ψn∗ (x) = ψn (x).
Using the representation of the momentum operator,

∂
pb = −ih̄ ,
∂x
the complex conjugate of the momentum expectation value is

 Z +∞
∗
∗ ∂
(h pbi) = −ih̄ ψn (x) ψn (x) dx (3.14)
−∞ ∂x
Z +∞
∂
= +ih̄ ψn (x) ψn (x) dx, since ψn (x) & x are real (3.15)
−∞ ∂x
= −h pbi, (3.16)

i.e. the expectation value is imaginary. But, being an observable quantity, it must be real:
the only possible conclusion therefore is that

h pbi = 0, for energy eigenstates. (3.17)

3.1 Energy Eigenstates are Stationary States 39

What is the expectation value of the energy? For an energy eigenstate we expect it to
be the energy eigenvalue itself; let us see if this is so. The operator representing energy in
quantum mechanics is the hamiltonian, so the expectation value is 1
Z +∞
<H
b > := Ψn∗ (x,t)HΨ
b n (x,t) dx (3.18)
−∞
Z +∞
= ψn∗ (x) Hψ
b n (x) dx (3.19)
−∞ | {z }
=En ψn (x)
Z +∞
= En ψn∗ (x,t)ψn (x,t) dx, (3.20)
−∞
= En (3.21)
where we first used the fact that ψn (x) is an eigenstate,
b n (x) = En ψn (x),
Hψ (3.22)
and finally that it is normalized,
Z +∞
ψn∗ (x)ψn (x) dx = 1. (3.23)
−∞
Since the expectation value is the energy eigenvalue itself, the eigenstate has a definite
energy; therefore we expect it’s energy uncertainty to be zero. First we calculate
Z +∞
b 2 i :=
hH Ψn∗ (x,t)H
b 2 Ψn (x,t) dx (3.24)
−∞
Z +∞
= ψn∗ (x)H
b 2 ψn (x) dx (3.25)
−∞
Z +∞
= En2 Ψn∗ (x,t)Ψn (x,t) dx, (3.26)
−∞
= En2 . (3.27)
Here we simply acted with the hamiltonian on the eigenstate twice in succession:
h i
2
H ψn (x) = H Hψn (x)
b b b (3.28)
h i
= En Hψ b n (x) (3.29)
= En2 ψn (x). (3.30)
Hence, the energy uncertainty vanishes,
q
∆E := <H b2 > − < H
b >2 (3.31)
q
= En2 − (En )2 (3.32)
= 0 (3.33)
1 In quantum mechanics one often uses the shorthand notation < E > for the expectation value hHi b even for
states which are not energy eigenstates. I shall give a simple example of such a state at the end of this section.
 40 Chapter 3. The Expansion Theorem (Exp Thm I)

3.2 Orthornormality of Energy Eigenstates and Reality of Eigenvalues

Perhaps the most important mathematical property of energy eigenstates is the orthogonality
of eigenstates corresponding to different energies. In this section we suppress the argument (x)
of the eigenstates ψi (x). To prove both orthogonality and the reality of the energy eigenvalues
we consider the integral:
Z +∞
Hi j := ψi∗ Hψ
b j dx (3.34)
−∞ |{z}
=E j ψ j
Z +∞
= Ej ψi∗ ψ j dx, (3.35)
−∞
where we used the fact that the ψi are eigenstates of the Hamiltonian.
The trick now is to somehow move the hamiltonian so that it acts on the other eigenstate.
We will show below that this is possible, with the amazingly simple result:
Z +∞ Z +∞  ∗
∗b
Hi j := ψi Hψ j dx = Hψ
b i ψ j dx (3.36)
−∞ −∞
This shows that the Hamiltonian is a Hermitian operator; we claimed already earlier that
this is a property of all operators representing physical observables in quantum mechanics
and for the Hamiltonian we will show this now explicitly. Note, we only do this here in 1D
but the proof in 3D follows exactly the same line of arguments – so the results we find will be
true for systems in any number of dimensions!
Proof. To proceed we recall the explicit form of the Hamiltonian:
2 2
b = − h̄ ∂ +V (x) ,
H (3.37)
2m ∂ x2
and perform two successive integrations by parts.
First let us deal with the first term in the hamiltonian:
Z +∞ 2  +∞ Z +∞   
∗ ∂ ∗ ∂ ∂ ∗ ∂
ψi ψ j dx = ψi ψj − ψ ψ j dx. (3.38)
−∞ ∂ x2 ∂x −∞ −∞ ∂x i ∂x
The first term in this expression evaluates the wave functions and their first derivative
at ±∞ where all wave functions vanish: the probability of finding our particle at infinity is
zero – see our discussion from Chapter 1 on physical boundary conditions! From now on we
shall always omit such “boundary terms" because well behaved wave functions must vanish at
spatial infinity. We now carry out the second integration by parts, omitting any surface terms
generated:
Z +∞ 2 Z +∞   
∗ ∂ ∂ ∗ ∂
ψi ψ j dx = − ψ ψ j dx by parts, as above (3.39)
−∞ ∂ x2 −∞ ∂x i ∂x
Z +∞  2 
∂ ∗
= + ψ ψ j dx by parts again (3.40)
−∞ ∂ x2 i
Z +∞  2  ∗
∂
= + ψi ψ j dx. (3.41)
−∞ ∂ x2
3.2 Orthornormality of Energy Eigenstates and Reality of Eigenvalues 41

In the last line we used the fact that x is real to place the derivative inside the complex
conjugation; similarly the potential V (x), being a real function, can be placed anywhere. The
result is that the entire hamiltonian has been moved over to act on the i-th eigenfunction. Thus
we have proved equation (3.36) showing that the Hamiltonian is a Hermitian operator. 
This enables us to obtain a second, alternative expression for our integral:
Z +∞
Hi j = Ψi∗ Hψ
b j dx, our first expression
−∞ |{z}
E jψ j
Z +∞
= Ej ψi∗ ψ j dx, our first result (3.42)
−∞
Z +∞ n o∗
= Hψi ψ j dx, our second expression
b
−∞ | {z }
=(Ei ψi )∗
Z +∞
= Ei∗ ψi∗ ψ j dx, our second result (3.43)
−∞
where we have used the fact that ψi is an energy eigenfunction.
Notice that the same integral occurs in each result; since they are equal, their difference is
zero:
Z +∞
(E j − Ei∗ ) ψi∗ ψ j dx = 0 (3.44)
−∞

There are two cases to consider:

Case 1: For i = j the expression gives

Z +∞
(E j − E j∗ ) |ψ j |2 dx = 0 (3.45)
−∞

Since the integrand is positive the integral cannot vanish; indeed it must be equal to 1 since
the eigenstate is normalised. Hence the other factor must vanish, i.e. the energy eigenvalue
is a real number,
E j∗ = E j (3.46)
Case 2: For i 6= j, and using the fact that the energies are real, the expression gives

Z +∞
(E j − Ei ) ψi∗ ψ j dx = 0, for i 6= j (3.47)
−∞

The energy factor is not zero if we assume that there is no degeneracy, i.e. that the energies
corresponding to different eigenfunctions are different. 2
2 Thecase of degeneracy does occur in situations where there are symmetries. They can be easily handled by
making particular linear combinations of the eigenstates in question, leading to the same results.
42 Chapter 3. The Expansion Theorem (Exp Thm I)

Hence the integral, involving different eigenstates, must vanish:

Z +∞
ψi∗ ψ j dx = 0, for i 6= j (3.48)
−∞

This is the statement that eigenstates corresponding to different eigenvalues are orthogonal.
We can combine this statement with the normalization condition in one elegant statement of
orthonormality:
Z +∞
ψi∗ ψ j dx = δi j (3.49)
−∞

where the kronecker delta is defined as

δi j = 1, for i = j (3.50)
= 0, for i 6= j (3.51)

Interpretation of orthonormality:
• Different energy eigenstates have nothing in common:
If the system is in state ψi we find always Ei if energy is measured, similarly if the
system is in state ψ j we find always E j but never Ei if i 6= j. This non-mixing property
is the orthogonality of ψi and ψ j .
• One can push this geometric analogy further and we will soon see that we can think of
states/wavefunctions ψ as vectors and the eigenstates ψi as orthogonal unit vectors.
∗
R
This means we should think of integrals like ψi (x) ψ j (x)dx as inner products (dot
products) of state (vectors) ψi and ψ j . And this inner product measures how much
these two states have in common. In the case of energy eigenstates we get zero if i 6= j
which just expresses intuitively that these to states “have nothing in common"; one
describes a state with energy Ei and the other one a state with energy E j .

• We elaborate this geometric picture later in the course and this lead to Dirac’s bra-ket
notation and Matrix Quantum Mechanics.
 3.3 The Expansion Theorem 43

3.3 The Expansion Theorem

Theorem 3.3.1 We come now to one of the most fundamental aspects of quantum me-
chanics. We have already noted the superposition principle: mathematically this is a
consequence of the Schrödinger equation’s linearity in Ψ. For all potentials of physical
interest it can be proved that the most general solution of the TDSE can be written as a
linear combination of energy eigenstates. This is the Expansion Theorem:

Ψ(x,t) = ∑ cn Ψn (x,t) = ∑ cn ψn (x)e−iEnt/h̄ (3.52)

n n

where the sum extends over all eigenstates and the expansion coefficients cn are complex
numbers.

The expansion theorem encodes an important implication of the TDSE:

Even though quantum mechanics is indeterministic, the time–evolution of the wave func-
tion itself is deterministic. By indeterministic we mean that quantum mechanics can only
predict the probability of a given outcome for a measurement; but the time-evolution of the
wave function is deterministic because, given Ψ(x, 0), its value at t = 0, the TDSE precisely
determines its evolution to later times t > 0, Ψ(x,t).

The proof follows from the expansion theorem and orthonormality:

Set t = 0 in eq. (3.52) – this sets e−iEnt/h̄ → e−iEn 0/h̄ = 1, multiply the equation by ψm∗ (x),
integrate over all x, and use the fact that all integrals in the series vanish except the m-th one –
multiplication by ψm∗ (x) and integration projects out all except the m-th term.

Z +∞ Z +∞
ψm∗ (x)Ψ(x, 0) dx = ∑ cn ψm∗ (x)ψn (x) dx
−∞ n −∞
= ∑ cnδmn
n
= cm (3.53)
Hence we have an expression for the expansion coefficients, cm , in terms of known
quantities: the initial wave function, Ψ(x, 0), and the eigenstates ψm (x),
Z +∞
cm = ψm∗ (x)Ψ(x, 0) dx (3.54)
−∞

Having found these expansion coefficients from this integral, we can then insert them into
the expansion theorem, eq. (3.52), to find the wave function Ψ(x,t) at all later times t > 0,
given that we already know the energy eigenvalues En belonging to each known eigenstate
ψn (x). There is a proviso, however: the wave function evolves according to this prescription
provided the system is left undisturbed after the initial time t = 0.
 44 Chapter 3. The Expansion Theorem (Exp Thm I)

3.3.1 Interpretation of expansion coefficients

The complex expansion coefficients cn clearly play an important role in determining the wave
function and its evolution. But what is their physical significance? Here we have to introduce
an additional postulate into our development of quantum mechanics; just as we did with the
momentum operator, we shall justify this new postulate by showing that it is a reasonable one.
We begin by calculating the energy expectation value,

Z +∞
hHi
b ≡ hEi = Ψ ∗ (x,t)HΨ(x,t)
b dx (3.55)
−∞

for the wave function given by the expansion theorem. Inserting the series eq. (3.52) for
each of the wave functions, we obtain a double sum,

Z +∞
hEi = Ψ ∗ (x,t)HΨ(x,t)
b dx
−∞
Z +∞
= ∑∑ cm∗ cn ei(Em −En )t/h̄ ψm∗ (x)Hψ
b n (x) dx
m n −∞
Z +∞
= ∑ ∑ cm∗ cnEnei(Em−En)t/h̄ −∞
ψm∗ (x)ψn (x) dx using Hψ
b n = En ψn ,
m n
= ∑∑ cm∗ cn En ei(Em −En )t/h̄ δmn using orthonormality,
m n
= ∑ |cn|2En . (3.56)
n

Using orthonormality to evaluate the integral, only the term with m = n survives in the sum
over m which also makes the time–dependence drop out.
In summary, the average energy is a time-independent, weighted sum over all the energies
with the weighting |cn |2 for energy En :

hEi = ∑ |cn |2 En (3.57)

n

This suggests that |cn |2 is the probability (which we will denote by Pn ) that the energy En
is the result of a measurement of energy for a particle in the state specified by the wave
function Ψ(x,t). To check that the Pn = |cn |2 are indeed probabilities, we use the fact that the
wave function is normalised; this calculation is identical to the preceding one, but without the
Hamiltonian and therefore without the factor En :
3.3 The Expansion Theorem 45

Z +∞
1 = Ψ ∗ (x,t)Ψ(x,t) dx
−∞
Z +∞
= ∑∑ cm∗ cn ei(Em −En )t/h̄ ψm∗ (x)ψn (x) dx
m n −∞

= ∑ ∑ cm∗ cnei(Em−En)t/h̄δmn
m n
= ∑ |cn|2, (3.58)
n

confirming that the |cn |2 indeed have the normalisation necessary to be probabilities:

∑ |cn|2 = ∑ Pn = 1 where Pn = |cn |2 . (3.59)

n
 46 Chapter 3. The Expansion Theorem (Exp Thm I)

3.4 Measurement Theorem

Postulate 6 Finally we are led to the key Measurement Postulate of quantum mechanics:
For a particle in the quantum state,

Ψ(x,t) = ∑ cn Ψn (x,t) (3.60)

n

the only possible result of a single energy measurement is one of the eigenvalues En with
probability Pn = |cn |2 .

There is more to the Measurement Postulate – the ‘Collapse of the Wave Function’ or
the ‘Reduction of the Wave Packet’:
If the energy measurement yields a result En at time t, then immediately after the measure-
ment the wave function ‘collapses’ to the corresponding eigenstate,

Ψafter (x,t) = Ψn (x,t) (3.61)

Note the absence of the cn : the state after the measurement must of course be a normalised
state.

Furthermore, if the energy is subsequently measured, with the particle now in the state
Ψafter (x,t) = Ψn (x,t), there is only one possible outcome: the energy will be En – with
probability 1. Of course the particle should not be disturbed between the first measurement
and this subsequent one.

It is the process of state reduction on measurement which is still controversial, although

it is consistent with all experiments so far performed. Some claim that the process can in
principle be described by the TDSE, the so–called unitary evolution; but others say this is not
possible and that quantum mechanics is incomplete.

We shall see all the components of the expansion and measurement postulates illustrated
and confirmed in experiments on spin one half particles. Other experimental evidence for
these postulates comes from the oscillations of neutral K-mesons and also of neutrinos. Also
notice that we can use the measurement process and collapse of the wavefuntion as a practical
tool to prepare a system in a definite state.
 3.5 Examples of Expansion Theorem 47

3.5 Examples of Expansion Theorem

 Example 3.1 Time evolution and expectation value of energy of a non-stationary state.
(Comment: We will look at the simplest example where the state is a linear combination of two
energy eigenstates but generalisations to more complicated states are straightforward. This
also illustrates that using the facts we have learned about the properties of energy eigenstates
and the expansion theorem can be used efficiently to avoid integrations to obtain expectation
values of the energy.)
Assume the state of a 1D system at t = 0 is

Ψ(x, 0) = c1 ψ1 (x) + c2 ψ2 (x) .

where c1,2 are complex numbers and ψ1,2 normalised energy eigenstates.
Questions:
1. What is the state Ψ(x,t) at a later time t > 0? √
2. By normalising Ψ find a relation between c1 and c2 . Assuming that c1 = 1/ 2 find c2
assuming it is real and positive.
3. What is the probability to find in a measurement of energy at t = 0 the value E1 , and
what is the wavefunction of the system after the measurement?
4. Compute the expectation value of the energy and uncertainty ∆E for measuring the
energy for Ψ with coefficients found in 2.
Answers:
1. Simple application of the expansion theorem gives
∞
Ψ(x,t) = ∑ cne−iEnt/h̄ψn(x) = c1ψ1(x)e−iE1t/h̄ + c2ψ2(x)e−iE2t/h̄
n=1

where we have first written the general formula (3.52) and used the fact that cn = 0 for
n ≥ 3 since we have a linear combination of Ronly the first two energy eigenstates.
2. We could do this by considering the integral Ψ∗ Ψdx = ..., or recalling that Pn = |cn |2
are the probabilities to find En in an energy measurement and these probabilities need to
some to one. Since most are zero we find
∞
∑ |cn|2 = |c1|2 + |c2|2 = 1
i=1
√
If c1 = 1/ 2 we find
√
1/2 + |c2 |2 = 1 ⇒ |c2 |2 = 1/2 ⇒ c2 = 1/ 2 using assumptions!

3. Using the measurement theorem we know the probability to find E1 is P1 = |c1 |2 = 1/2.
√
The wave function after the measurement is (note the coefficient is 1, not c1 = 1/ 2
since the new wave function must be normalised correctly!)

Ψa f ter = 1ψ1 (x)e−iE1t/h̄

48 Chapter 3. The Expansion Theorem (Exp Thm I)

4. Using (3.57) we immediately find (if you want you can also perform the integral)

b = ∑ |cn |2 En = 1 E1 + 1 E2
hEi = hHi
2 2

as a sum over the energy eigenvalues weighted by the probabilities Pn = |cn |2 . Using the
b 2 ψn (x) = En2 ψn (x) one also has
same steps as in the proof for (3.57) and using that H

b 2 i = ∑ |cn |2 (En )2 = 1 E12 + 1 E22

hE 2 i = hH
2 2

Now

1
(∆E)2 = hE 2 i − hEi2 = (E2 − E1 )2
4

hence, assuming E2 > E1 we find ∆E = (E2 − E1 )/2.



 Example 3.2 Expansion theorem for general wave function in the infinite square well:
At t = 0 a particle in an infinite square well is prepared in a state corresponding to the
normalised top-hat wave function, Ψ(x, 0), illustrated:
V =∞ V =∞ For this infinite square well potential
6 Ψ(x, 0) 6 the energy eigenstates are,
for n = 1, 2, . . .,
q
2 nπ
ψn (x) = L sin( L x) for 0 ≤ x ≤ L
?
N
= 0 elsewhere
- x
0 L
Questions:
1. Calculate the normalisation constant N assuming it is real and positive.
2. Obtain the probabilities Pn that the result of an energy measurement is En . Evaluate
numerically the probability to find the system in the groundstate.
3. Show that for even n the probabilities are zero. Give a general reason for this result.
Answers:
1. Note that Ψ(x, 0) = N if 0 ≤ x ≤ L otherwise Ψ = 0. Normalise:
Z L
1
Z ∞
2
1= |Ψ(x, 0)| dx = |N|2 dx = |N|2 L ⇒ N = √ using assumptions
−∞ x=0 L

2. The probabilities are Pn = |cn |2 where the expansion coefficients are given by the general
3.5 Examples of Expansion Theorem 49

formula Z ∞
cn = ψn∗ (x)Ψ(x, 0)dx (3.62)
−∞
Z L r
2  nπx  1
= sin √ dx (3.63)
x=0 L L L
√  x=L
2 L  nπx 
= − cos (3.64)
L nπ L x=0
√
2
= (− cos(nπ) + cos(0)) (3.65)
nπ
√    
2 −1 for n = 1, 3, 5, . . .
= − +1 (3.66)
nπ 1 for n = 2, 4, 6, . . .
 √
 2nπ2 for n = 1, 3, 5, . . .
= (3.67)

0 for n = 2, 4, 6, . . .
Hence, the probabilities are
 8
 n2 π 2 for n = 1, 3, 5, . . . parity EVEN eigenfucntion
Pn =
0 for n = 2, 4, 6, . . . parity ODD eigenfucntion


with P1 = 8/π 2 ∼ 0.81 for the groundstate.

3. As indicated above only the cn and ψ with n an odd number contribute to the expansion
theorem. The wavefunction Ψ is a parity EVEN function (under reflection around the
midpoint x = L/2), hence we expect only EVEN eigenstates to contribute; the contribute
exactly to the eigenstates ψn with n an odd integer.


 Example 3.3 Animations of time evolution of general wave functions in the infinite
square well and the simple harmonic oscillator using Mathematica. 
 50 Chapter 3. The Expansion Theorem (Exp Thm I)

3.6 Optional: simple example of a non-stationary state

Suppose we illuminate an ensemble of systems (such as a gas of H–atoms at low temperature)
in their ground state with a short laser pulse. Let the laser have a low enough frequency that it
can only excite the system into its first excited state. Also suppose that there is a 50:50 chance
of this excitation happening. The resulting wave function describing the ensemble prepared in
this way is the particular linear combination of eigenstates,

1
Ψ(x,t) = √ [Ψ1 (x,t) + Ψ2 (x,t)]
2
1 h −iE1 t/h̄ −iE2 t/h̄
i
= √ ψ1 (x)e + ψ2 (x)e (3.68)
2

With a linear combination of two eigenstates the exponentials do not cancel in expectation
values but generate time dependence; for example, the probability density,
1
Ψ ∗ (x,t)Ψ(x,t) = |Ψ1 (x,t)|2 + |Ψ2 (x,t)|2 + Ψ1∗ (x,t)Ψ2 (x,t) + Ψ2∗ (x,t)Ψ1 (x,t)

2
1
|Ψ1 (x,t)|2 + |Ψ2 (x,t)|2 + 2ℜe (Ψ1∗ (x,t)Ψ2 (x,t))

=
2  
1 2 2 (E2 − E1 )t
= |ψ1 (x)| + |ψ2 (x)| + 2ψ1 (x)ψ2 (x) cos , (3.69)
2 h̄

is oscillatory with angular frequency proportional to the energy difference,

(E2 − E1 )
ω12 = . (3.70)
h̄
However, not all expectation values are time–dependent. For example, the normalization
of this wave function remains fixed for all time.

Proof:

Using the expression (3.69) for the probability density:

Z +∞ Z +∞ Z +∞ Z +∞ 
2 1 2 2
|Ψ(x,t)| dx = |ψ1 (x)| dx + |ψ2 (x)| dx + 2 cos ω12t ψ1 (x)ψ2 (x) dx ,
−∞ 2 −∞ −∞ −∞
1
= [1 + 1 + 0]
2
= 1 (3.71)

where we have used unit normalization of the two eigenstates in the first two integrals and
their orthogonality in the third.
3.6 Optional: simple example of a non-stationary state 51

A more illuminating proof shows the generality of our result by using the full wave
functions Ψ1 (x,t), Ψ2 (x,t) and the expression (3.69) for the probability density:
Z +∞ Z +∞ Z +∞
∗ 1 2
Ψ (x,t)Ψ(x,t) dx = |Ψ1 (x,t)| dx + |Ψ2 (x,t)|2 dx
−∞ 2 −∞ −∞
Z +∞ Z +∞ 
∗ ∗
+ Ψ1 (x,t)Ψ2 (x,t) dx + Ψ2 (x,t)Ψ1 (x,t) dx
−∞ −∞
1
= [1 + 1 + 0 + 0]
2
= 1 (3.72)

where we have again used unit normalization of the two wave functions in the first two
integrals and their orthogonality in the third and fourth.

The expectation values for the energy and its powers are also time–independent.

Proof:

Z +∞
hHi
b := Ψ ∗ (x,t)HΨ(x,t)
b dx
−∞
+∞ Z +∞
Z
1
= Ψ1∗ HΨ
b 1 dx + Ψ2∗ HΨ
b 2 dx
2 −∞ −∞
Z +∞ Z +∞ 
+ Ψ1∗ (x,t)HΨ
b 2 (x,t) dx + Ψ2∗ (x,t)HΨ
b 1 (x,t) dx
−∞ −∞
 Z +∞ Z +∞
1
= E1 Ψ1∗ Ψ1 dx + E2 Ψ2∗ Ψ2 dx
2 −∞ −∞
Z +∞ Z +∞ 
+E2 Ψ1∗ (x,t)Ψ2 (x,t) dx + E1 Ψ2∗ (x,t)Ψ1 (x,t) dx
−∞ −∞
1
i.e. hHi
b = [E1 + E2 ] (3.73)
2

where we have used the fact that the two states are energy eigenstates,

b n = En ψn ,
Hψ

and the normalization and orthogonality of the two states, as before.

Note that the expectation value is just the average energy, hEi = hHi,
b with equal weighting
for the two possibilities, E1 and E2 . We shall see the explanation for this in more general form
shortly under the heading ‘The Measurement Postulate’. You should check for yourself that if
instead we used a wave function,

Ψ(x,t) = c1 Ψ1 (x,t) + c2 Ψ2 (x,t), (3.74)

52 Chapter 3. The Expansion Theorem (Exp Thm I)

then its normalization requires that

|c1 |2 + |c2 |2 = 1, (3.75)

and furthermore, that for this wave function,

hEi = |c1 |2 E1 + |c2 |2 E2 . (3.76)

The weighting now is in the ratio |c1 /c2 |2 , as one might have guessed.

What about the uncertainty in energy, ∆E? Here we encounter the essentially quantum
mechanical feature: a system with this wave function does not have a definite energy – indeed,
we might say that it does not ‘have’ an energy at all until we measure it. First we must calculate
hHb 2 i. This is easy, especially with a word–processor such as the one I am using for writing
these notes: all I do is replace Hb everywhere in the calculation of hHi
b with H b 2 by inserting an
extra H b next to each H:
b

b n = En Ψn → H
HΨ b n = En2 Ψn
b HΨ (3.77)

Therefore every En is replaced by an En2 , but otherwise the calculation proceeds as before,
with the result:

b2i = 1 2
E1 + E22

hH (3.78)
2
Hence the energy uncertainty for this state is:

q
∆E := <H b2 > − < H b >2
r
1 2 1
= [E1 + E22 ] − [E1 + E2 ]2
2 4
r
1 2 1
= [E1 + E22 ] − E1 E2
4 2
r
1
= [E2 − E1 ]2
4
1
i.e. ∆E = [E2 − E1 ] (3.79)
2
This is not perhaps surprising: since the system has equal probability of being in either
state, the average spread in energy is determined by the energy difference (E2 − E1 ).
 4. Matrix Q.M. and Bra-Ket Notation

4.1 Motivation
In this chapter we want to put general properties of wave functions that we discovered in
earlier chapters into a more general (and partially geometric) context. This will also allow
us to (re)discover alternative formulations of quantum mechanics. For example Q.M. was
originally formulated by Heisenberg using matrices (since they do not commute, a property
of observables he was trying to model), before Schrödinger introduced his now more widely
used formalism with the wave functions and observables defined as differential operator like
pbx . It was later found that these two are equivalent, but for certain quantities, in particular spin,
the matrix formulation is unavoidable.
As you have seen in Q.M. we often encounter mathematical expressions like
Z Z
∗
χ (x)ψ(x)dx for example cn = ψn∗ (x)Ψ(x, 0)dx (4.1)
Z
hAi
b = Ψ∗ (x,t)A
bΨ(x,t)dx
b (4.2)
Z
Hi j ≡ ψi∗ (x)Hψ
b j (x)dx (4.3)

Goals:
1. Expose the similarity (equivalence) of states and operators with complex vectors and
matrices, the similarity of the expressions above with vector dot products and matrix
multiplication.
2. Introduce a unifying notation ⇒
Dirac’s BRA-KET notation, whereby (normalised) states are identified with state
vectors (of unit length) living in a “Hilbert space" = space of allowed (well-behaved,
normalisable) wave functions.
 54 Chapter 4. Matrix Q.M. and Bra-Ket Notation

4.2 Basics of Matrix Quantum Mechanics

4.2.1 States in Matrix Quantum Mechanics
It all starts and ends with the all important Expansion Theorem; for now we hold time fixed
t = 0 and consider a system with N energy eigenstates ψn (x) with n = 1, 2, . . . N where N can
be finite or infinite depending on the system under consideration. So a general state can be
written as a linear combination
N
ψ(x) = ∑ cnψn(x) (4.4)
n=1

The crucial observation here is that ψ(x) and the set of coefficients cn contain exactly the same
information about the system. If you know one you know the other!
N
cn −→ ψ(x) = ∑ cnψn(x) (4.5)
n=1
Z
ψ(x) −→ cn = ψn∗ (x)ψ(x)dx , (4.6)

also since the cn comprise a list of N complex numbers we can repackage them into a complex
vector C. To make this totally obvious let us write this out as an N-component column vector
       
c1 1 0 0
 c2  0 1 0
ψ(x) ↔ C =  ..  = c1  ..  +c2  ..  + · · · cN  ..  , (4.7)
       
. . . .
cN 0 0 1
| {z } | {z } | {z }
ê1 ê2 êN

where we have made manifest another property of vectors, namely that they can be “expanded"
in terms of N orthogonal unit-vectors ên – by analogy we can also call these orthonormal,
since:
Z
ψm∗ (x)ψn (x)dx = δmn ←→ ê†m ên = δmn (4.8)

This already hints at the following identifications:

eigenstate ψn (x) ←→ ên column unit-vector (4.9)

complex conjugate eigenstate ψn∗ (x) ←→ ê†n row unit-vector (4.10)

but let us try to push this much further. Introduce another state and its complex conjugate
N
χ(x) = ∑ bmψm(x) (4.11)
m=1
N
χ ∗ (x) = ∑ b∗mψm(x) (4.12)
m=1
4.2 Basics of Matrix Quantum Mechanics 55

with bm = ψm∗ (x)χ(x)dx. As column/row vectors these become

R
 
b1
 b2 
χ(x) ↔ B =  ..  , χ ∗ (x) ↔ B† = b∗1 b∗2 · · · b∗N .


(4.13)
 
 . 
bN
Integrals are dot-products!
Now let us consider the first expression we mentioned:
Z Z
∗
χ (x)ψ(x)dx = (∑ b∗m ψm∗ (x))(∑ cn ψn (x)) (4.14)
m n
N Z
= ∑ b∗m cn ψm∗ (x)ψn (x)dx (4.15)
m,n=1 | {z }
=δmn
N
= ∑ b∗mcm (4.16)
m=1
†
= B C complex vector multiplication (4.17)
 
c1

 c2 

= b∗1 b∗2 · · · b∗N  ..  , (4.18)

.
cN
arriving finally at complex vector multiplication ⇒ dot (inner) vector product. One crucial
property of the dot product is that if it vanishes then the two vectors are orthogonal. So in a
very concrete sense we are starting to geometrise properties of wave functions/states.
With the help of the dot product we can also define the length squared of a vector, obviously
this should be related to the norm squared of a state ψ(x). Normalisation:
Z Z Z
1= |ψ(x)|2 dx = ψ ∗ (x)ψ(x)dx = (∑ c∗m ψm∗ (x))(∑ cn ψn (x)) (4.19)
m n
N Z
= ∑ c∗m cn ψm∗ (x)ψn (x)dx (4.20)
m,n=1 | {z }
=δmn
N
= ∑ |cm|2 (4.21)
m=1
= C C = |C|2 complex vector multiplication
†
(4.22)
 
c1

 c2 

= c∗1 c∗2 · · · c∗N  ..  . (4.23)

.
cN
It seems we are reinventing Linear Algebra!
 56 Chapter 4. Matrix Q.M. and Bra-Ket Notation

4.2.2 Operators in Matrix Quantum Mechanics

So states are vectors, but what are operators?

Consider an operator A b acting on a general state ψ(x) producing a new state χ(x).

χ(x) = Aψ(x)
b ⇒ ∑ bk ψk (x) = A(
b ∑ cn ψn (x)) (4.24)
k n

Now we need to know bk =? To get the expansion coefficients, multiply the equation above
with ψm∗ (x), integrate and use orthonormality...

Z Z
∑ bk ψm∗ (x)ψk (x)dx = ∑ cn ψm∗ (x)Aψ
b n (x)dx , (4.25)
k | {z } n | {z }
=δmk =Amn

where we introduced the matrix-form of the operator

Z
Amn = ψm∗ (x)Aψ
b n (x)dx (4.26)

and we find

N
bm = ∑ Amncn (4.27)
n=1

which can equivalently be written in matrix notation

    
b1 A11 A12 · · · A1N c1
 b2   A21 A22 · · · A2N   c2 
B = AC ⇔  ..  =  .. (4.28)
    
.. .. ..   .. 
 .   . . . .  . 
bN AN1 AN2 · · · ANN cN
| {z }
=Amn

General expectation values: Now let us compute the expectation value for a completely
4.2 Basics of Matrix Quantum Mechanics 57

general state ψ(x)

Z
hAi
b = ψ ∗ (x)Aψ(x)dx
b (4.29)
Z
= (∑ c∗m ψm (x))A(
b ∑ cn ψn (x))dx (4.30)
m n
N Z
= ∑ c∗m cn ψm∗ (x)Aψ
b n (x)dx (4.31)
m,n=1 | {z }
=Amn
N
= ∑ c∗m Amn cn = C† AC (4.32)
m,n=1
  
A11 A12 · · · A1N c1
 A21 A22 · · · A2N   c2 
c∗1 c∗2 · · · c∗N  ..


= (4.33)
  
.. ... ..   .. 
 . . .  . 
AN1 AN2 · · · ANN cN
| {z }
=Amn

and again we can express it completely in terms of matrix products!

Lessons:
(1) The column vector C contains all information about the state ψ(x).
(2) If we know the matrix form of the operator Amn we can find all possible expectation values
for all possible states without any integration!

 Example 4.1 Hamiltonian:

Z Z
Hmn = ψm∗ (x) Hψ
b n (x) dx = En ψm∗ (x)ψn (x)dx = Em δmn (4.34)
| {z }
=En ψn (x)

 
E1 0 0 ···
 0 E2 0 · · ·
Hmn =  0 0 E (4.35)
 
 3 · · ·

.. .. .. ...
. . .


In this case the expectation value of energy immediately evaluates to

N
hEi = C† HC = . . . = ∑ |cn|2En . (4.36)
n=1
 58 Chapter 4. Matrix Q.M. and Bra-Ket Notation

Comments:
(1) This is a special case because for the Hamiltonian the matrix form of the operator is a diag-
onal and Hermitian matrix Hmn = En δmn (just put all energy eigenvalues in the diagonal).
(2) For general operators, for example xb and pbx , we will see that the corresponding matrix is
Hermitian but not diagonal.

4.2.3 Multiplication of operators in matrix form

This is very simple; operator multiplication becomes matrix multiplication of the matrices
representing the operators (in the same order as the operators appear):

N
bBb ←→ (AB)mn =
A ∑ Amk Bkn this sum is just matrix multiplication (4.37)
k=1

For example for the matrix form of the Hamiltonian squared we find
 
E12 0 0 ···
 0 E2 0 · · ·
2 2
H = HH =  0 0 E 2 , (4.38)
 
 3 · · ·
.. .. .. ..
. . . .

this was particularly easy since H happens to be diagonal, in general this is more work.
Once we know how to multiply the matrix form of operators we can also compute commu-
tators

[A, bBb − BbA

b =A
b B] b ←→ [A, B] = AB − BA using matrix multiplication (4.39)

4.2.4 Matrix form of TISE and TDSE

First we consider the TISE

b n (x) = En ψn (x)
Hψ (4.40)

and using the maps

Z
ψn (x) ←→ ên and H
b ←→ Hmn = ψm∗ Hψ
b n dx (4.41)

we bring it to the matrix form of the TISE

H ên = En ên (4.42)

4.2 Basics of Matrix Quantum Mechanics 59

Since H is a diagonal matrix the eigenvectors indeed are the unit-vector ên introduced earlier.
1At the danger of being pedantic we make this fully explicit.
         
E1 0 0 · · · 1 1 E1 0 0 · · · 0 0
 0 E2 0 · · · 0 0  0 E2 0 · · · 1 1
 0 0 E3 · · · 0 = E1 0 ,  0 0 E3 · · · 0 = E2 0 , . . . (4.43)
         
         
.. .. .. . . .. .. .. .. .. . . .. ..
. . . . . . . . . . . .

On to the TDSE: we have to bring back the time dependence that we suppressed for now.
First recall the expansion theorem for a general solution of the TDSE
N
Ψ(x,t) = ∑ c|ne−iEn t/h̄
{z } n
ψ (x) (4.44)
n=1
≡cn (t)
N
Ψ(x, 0) = ∑ cn ψn (x) (4.45)
n=1
|{z}
≡cn (0)

where we have introduced the time-dependent expansion coefficents to absorb the time-
dependent exponential

cn (t) ≡ cn e−iEnt/h̄ (4.46)

Hence, the time dependent solution of the TDSE Ψ(x,t) is mapped to a time-dependent vector
C(t)
 
c1 (t)
 c2 (t) 
Ψ(x,t) ←→ C(t) =  ..  (4.47)
 
 . 
cN (t)

The TDSE in matrix form therefore becomes

∂ Ψ(x,t) d
HΨ(x,t)
b = ih̄ ←→ HC(t) = ih̄ C(t) (4.48)
∂t dt

Proof. Writing out HC(t) = ih̄ dtd C(t) and performing the matrix multiplication on the lefthand
side of the equation one obtains
     
E1 c1 (t) c1 (t) E1 c1 (t)
d 
HC(t) = E2 c2 (t) = ih̄ c2 (t) = E2 c2 (t) , (4.49)
    
.. dt .. ..
. . .
1 You need to remember that for a diagonal matrix the eigenvalues are simply the entries in the diagonal and
the eigenvectors are the ên , there is no need to go through the diagonalisation procedure required for general
matrices.
60 Chapter 4. Matrix Q.M. and Bra-Ket Notation

where in the last step we have used (4.46) to show that

dcn (t)
ih̄ = ih̄cn (−iEn /h̄)e−iEnt/h̄ = En cn (t) .
dt

 4.3 2-State System in Matrix Q.M. 61

4.3 2-State System in Matrix Q.M.

We will now discuss in detail a simple system in which the particle can only be in two
energy eigenstates N = 2 in the matrix formulation of Q.M. On the one hand this will allow
us to see the machinery introduced in the previous section in action, but this case also has
superimportant applications:
1. QUBITs (quantum computers) are quantum generalisation of the smallest information
units in computers – the bit (0 or 1). Crucial difference between bits and qubits is that
for bits there are only 2 binary choices, while the qubit can be in a superpositions of the
0 and 1 state!
2. Spin 12 Particles: such particles can in two different spin states:
   
1 0
ψ1 , ψ2 ←→ , (4.50)
0 1
|{z} |{z}
spin up spin down

3. In some sense these are the most quantum mechanical systems since they have no
classical limit. There is no quantum number that you can make large
To set the stage we introduce two operators (Hermitian matrices ⇒ real eigenvalues,
orthonormal eigenvectors):
   
E1 0 0 1
H= , A= (4.51)
0 E2 1 0
with two corresponding sets of eigenvalues and eigenvectors:

Eigenvalues of H Eigenvectors
  Eigenvalues of A Eigenvectors
 
1 1 1
E1 ê1 = λ1 = 1 V1 = √2
0 1 (4.52)
0 1
E2 ê2 = λ2 = −1 V2 = √12
1 −1
obeying eigenvalue equations

H ê1 = E1 ê1 , H ê2 = E2 ê2 , AV(i) = λiV(i)

Please briefly check these yourselves.

 Example 4.2 A general state at t = 0 is a linear combination of the energy eigenvectors
     
c1 1 0
C= = c1 ê1 + c2 ê2 = c1 + c2
c2 0 1

Find the normalisation condition for the state and compute the expectations values of H, H 2 , A, A2 .
Normalisation
 
† ∗ ∗

c1
C C = c1 c2 = |c1 |2 + |c2 |2 = 1
c2
62 Chapter 4. Matrix Q.M. and Bra-Ket Notation

Expectation values
    

E1 0 c1
c1 E1
†
hEi = C HC = c∗1 c∗2 ∗ ∗
= c1 c2 = |c1 |2 E1 + |c2 |2 E2
0 E2 c2 c2 E2

hE 2 i = C† HHC = . . . = |c1 |2 E12 + |c2 |2 E22

    

0 1 c1
c2
† ∗
hAi = C AC = c1 c2 ∗ ∗
= c1 c2∗ = c∗1 c2 + c∗2 c1 = 2ℜc∗1 c2
1 0 c2 c1
  
† † 0 1 0 1
hAi = C AAC = C C = C†C = 1
1 0 1 0
| {z ! }
1 0
=
0 1


Matrix A has the special property to exchange the two energy eigenvectors (A : ê1 ↔ ê2 ),
since
    
0 1 1 0
Aê1 = = = ê2 and Aê2 = ê1 .
1 0 0 1

Note also the remarkable fact that the two matrix operators have different eigenvectors.

What is the reason for this? The reason is that if two operators (matrices) do not commute
then they do not have simultaneous eigenstates(eigenvectors), this means both operators have
different eigenvectors. We will soon show this in the course.

Check:
     
E1 0 0 1 0 1 E1 0
[H, A] = − (4.53)
0 E2 1 0 1 0 0 E2
     
0 E1 0 E2 0 E1 − E2
= − = 6= 0 (4.54)
E2 0 E1 0 E2 − E2 0

if E2 > E1 .

 Example 4.3 Time evolution: Assume the system at t = 0 is in the general state
     
c1 1 0
C(0) = = c1 ê1 + c2 ê2 = c1 + c2 .
c2 0 1

Questions:
1. Find the general state C(t) at later times t > 0.
2. At t = 0 A is measured and eigenvalue λ1 = +1 is found. Find the state for t > 0.
4.3 2-State System in Matrix Q.M. 63

3. Now the system evolves without disturbances until at a later time t > 0 we measure A
again. Find the probability to find again the value λ1 = +1.
Answers:
1. Simple application of the expansion theorem gives
   
−iE1 t/h̄ 1 −iE2 t/h̄ 0
C(t) = c1 e + c2 e .
0 1

Note since we expand in energy eigenvectors êi we only need to multiply the expansion
coefficient with the usual time-dependent exponentials.

2. Slight generalisation of Measurement Theorem. At t = 0 A is measured and λ1 = +1, the

wave function collapses to the corresponding eigenvector V(1) . Hence immediately after the
measurement the state must be
     
1 1 1 1 1 0
C(0) = V(1) =√ =√ +√
|{z} 2 1 2 0 2 1
eigenvector of A

Hence we find c1 = c2 = √1 and at t > 0 the state is

2

1 e−iE1t/h̄
     
1 −iE1t/h̄ 1 1 −iE2t/h̄ 0
C(t) = √ e +√ e =√
2 0 2 1 2 e−iE2t/h̄

3. Now we measure again A at time t > 0. To find the probability to find again the eigenvalue
λ1 = +1 we need to remember that its corresponding eigenvector is V(1)
In order to find the probability we need to find out: How much of V(1) is inside C(t)?
Mathematically this is given by the “dot product"

1 e−iE1t/h̄
 
† 1 1  −iE1t/h̄ −iE2 t/h̄

V(1)C(t) = √ 1 1 √ = e +e
2 2 e−iE2t/h̄ 2

This is the analogue of expansion coefficents cn = ψn∗ (x)ψ(x)dx, so in order to get probability
R

we need to take the absolute value squared

†
P+1 (t) = |V(1) C(t)|2
1  iE1t/h̄  
= e + eiE2t/h̄ e−iE1t/h̄ + e−iE2t/h̄
4 
1
= 1 + 1 + ei(E2 −E1 )t/h̄ + ei(E1 −E2 )t/h̄ 
4 | {z }
=2 cos(∆Et/h̄) with ∆E=E2 −E1
 
1 2 ∆Et
= (1 + cos (∆Et/h̄)) = cos (4.55)
2 2h̄
 64 Chapter 4. Matrix Q.M. and Bra-Ket Notation

Figure 4.1: Here we demonstrate how the probability to find again eigenvalue +1 evolves with
time. Initially this probablity is close to 1, but after some time it even vanishes before it goes
again up to 1. The probability P−1 to find the eigenvalue −1 shows the opposite behaviour.

Interestingly this probability is time-dependent: if the 2nd measurement occurs immedi-

ately after the 1st the probability is 1 but if we let time pass this probability decreases until it
vanishes, and then increases again for even larger values of t. This is highly counterintuitive
(and quantum mechanical): after all we measure the property of the system at t = 0 and believe
with certainty to know what it was, but quantum evolution makes this property “fade in and
out" periodically! 

4.3.1 Neutrino Oscillations

This example is a toy model for some very real physical phenomena, in particular the neutrino
oscillations! You might know or will hear in a particle physics course that neutrinos come
in 3 flavours: electron-, muon- and tau-neutrinos. In order to parallel the discussion with the
previous example we focus now only on the first two flavours:
νe electron neutrinos, λ = +1, ∼ V(1)
νµ muon neutrinos, λ = −1, ∼ V(2)
where we indicated that the two different flavours are distinguished by their λ eigenvalue.
Now we need to introduce a few more facts: we assume that the Hamiltonian including
all forces has eigenvalues E1 and E2 with eigenvectors ê1,2 . However, the particles that we
call neutrinos are produced in the so-called β -decay which is governed by a piece of the full
Hamiltonian Hweak that controls the weak interactions; but as it turns out this Hamiltonian has
different eigenstates of the form
   
cos θ − sin θ
Vνe = and Vνµ =
sin θ cos θ
where θ is called the “mixing-angle" and is not a geometric angle. Our simple model above
would have θ = π/2 but nature chose a different value (for details ask my particle physics
colleagues).
4.3 2-State System in Matrix Q.M. 65

The up-shot of all this is that if an electron-neutrino νe is produced at t = 0 for example by

nuclear reactions in the sun, then the probability to find again an νe at a later time on earth is
 
2 2 ∆Et
1 − sin (2θ ) sin
2h̄

where ∆E ∼ c3 (m22 − m21 )/(2p). In particular this depends on the masses of the particles
and a non-trivial effect is only expected if at least one of the two types of particles has a
non-vanishing mass. On the other hand for a long time people believed that neutrinos are
massless! Which brings us to another story...

(Resolution) of the solar neutrino problem: The sun is a huge nuclear reactor and since we
know its size we can very well estimate the flux of neutrinos that should hit the earth. The
travel time is 499 seconds, but the nuclear reactions take place deep in the sun’s core and
it takes thousands of years for photons (sunlight) to escape the core while neutrinos escape
immediately. This would mean that if the rate of nuclear reactions changes we would detect
this immediately from a change in the rate of neutrinos but it would take thousands of years to
observe a change in the intensity of sunlight.
In any case researchers observed that only ca. 1/3 of the expected νe ’s were arriving on
earth – leaving a 2/3 deficit. This initially caused panic since if the neutrinos are massless as
assumed at the time this would indicate that the nuclear reactor of the sun is shutting down and
in some hundred years the earth might freeze over. Bruno Pontecorvo (1957) had the brilliant
idea that if the neutrinos have tiny mass then the discrepancy can be explained by the fact that
the νe ’s produced at the sun convert (oscillate) into other types of neutrinos during their 499
second trip to earth. This idea turned out to be correct, it is now accepted that neutrinos carry
tiny masses and to this date neutrino physics is an extremely active area of research.
 66 Chapter 4. Matrix Q.M. and Bra-Ket Notation

4.4 Bra-Ket Notation

This notation was introduced by Dirac to simply and unify the different representations.
Using the expansion theorem we can write two wave functions ψ(x) = ∑n cn ψn (x) , χ(x) =
∑n bn ψn (x). We often consider integrals of the form which in matrix quantum mechanics
correspond to an inner (dot) product:

Z
χ ∗ (x)ψ(x)dx ≡ hχ | ψi ≡ ∑ b∗n cn = B†C ψ(x) ↔ |ψi ↔ C and χ ∗ (x) ↔ hχ| ↔ B†
|{z} |{z} n
hBRA| |KET i

(4.56)
where on the left we have the usual representation using wave functions and on the far right
the equivalent expression in matrix quantum mechanics. Both are represented in the middle as
the inner- or dot-product of the abstract hBRA| and |KET i states, the point being that it is their
product, the only physically relevant quantity, that should not depend on the representation or
framework we use to express states.
Similar expressions exist where operators are placed in the middle, in Dirac’s notation
these become
Z
χ ∗ (x) Aψ(x)
b dx ≡ hχ|ρi ≡ hχ|A|ψi
b = ∑ b∗m Amn cn ≡ B† AC (4.57)
| {z } m,n
=ρ(x)

This is only a very convenient notation valid in any dimension and allows to perform many
computations more efficiently and compactly, for example many computations that seemed to
involve long integrations collapse to simple algebra. We will use it mainly as a useful book
keeping device and postpone a full mathematical development to later courses.

Examples:

• b i i = Ei |ψi i
TISE: H|ψ
• Orthonormality: hψn |ψm i = δnm
• Operator in matrix form: Hi j = hψi |H|ψ
b ji
• Expectation values: hAi
b = hΨ|A|Ψi
b
• Expansion coefficients: cn = hψn |Ψ(x, 0)i

Properties:

• hχ|cψi = chχ|ψi
• hcχ|ψi = c∗ hχ|ψi
• hχ|c1 ψ1 + c2 ψ2 i = c1 hχ|ψ1 i + c2 hχ|ψ2 i
• hχ|ψi∗ = hψ|χi
Here we stop because that’s all you need to know about Dirac formalism in this course.
 5. Hermitian Operators & Exp Thm II

5.1 Hermitian Matrices

From standard mathematics courses you should know the definition and properties of Hermitian
operators:
Definition 5.1.1 A matrix A is called hermitian if

A† ≡ (A∗ )T = A or equivalently in components A∗mn = Anm (5.1)

Note: If a matrix is Hermitian its eigenvalues λi are all real numbers and its eigenvectors U(i)
are guaranteed to be orthonormal:
†
A† = A ⇒ AU(i)) = λiU(i)) with U(i)U( j) = δi j and λi real numbers (5.2)

Properties of Hermitian Conjugation † :

 †
A† = A (5.3)
(cA)† = c∗ A† (5.4)
(A + B)† = A† + B† (5.5)
(AB)† = B† A† (5.6)

Need to be able to check this property.

 Example 5.1
   
i 1−i † −i 1 − 2i
A= : A = 6= A NOT Herm.
1 + 2i 3 1+i 3
 68 Chapter 5. Hermitian Operators & Exp Thm II

Spin operator

h̄ 0 i T h̄ 0 −i
     
h̄ 0 −i †
Sy = : Sy = = = Sy YES, Herm.
2 i 0 2 −i 0 2 i 0


5.2 Hermitian Operators

b† .
First we define the Hermitian Adjoint of an operator A
b† is the Hermitian Adjoint of an the operator A
Definition 5.2.1 A b if:
Z Z  ∗
χ ∗ (x)Aψ(x)dx
b = b† χ(x) ψ(x)dx
A (5.7)

for all functions χ(x) and ψ(x).

or in BRA-KET notation

hχ|A|ψi
b = hχ|Aψi b† χ|ψi
b = hA (5.8)

It will be very important to check if an operator is Hermitian or not. Since we often deal
with composite operators it is useful to quote the properties of the Hermitian Adjoint which
are completely identical to the rules for matrices:
1.
 †
b† = A
A b

2.
 †
b = c∗ A
cA b†

3.
 †
A b† + Bb†
b + Bb = A

4.
 †
bBb = Bb† A
A b†

which can be proven by applying Definition 6.2.1 twice

Z Z Z
∗ bb b† χ)∗ Bψdx b† χ)∗ ψdx
χ ABψdx = (A b = (Bb† A
 5.2 Hermitian Operators 69

Let us repeat the important Postulate 4 from...:

“In quantum mechanics every physical observable corresponds to a linear and Hermitian
operator. Such operators can be a function F(~x), a differential operator or a matrix, but in
any case they must have these two properties in order to be a potential observable, something
we can measure in an experiment."
Linearity is important for the superposition principle, while Hermiticity guarantees that all
eigenvalues are real (as they should since they are possible outcomes of measurements) and
all eigenstates (eigenvectors) are orthonormal. We know this already for matrices and H, b and
for general operators we will prove this momentarily.
But first we want to discuss a general method to check Hermiticity of (composite) operators
using the properties reviewed above. For this we first need to check Hermiticity of the simplest
building blocks of all operators – the position and momentum operators – and then we can
proceed to general operators.

1. Position Operators are Hermitian. For example xb† = xb = x because

Z Z
∗
χ xψdx = (xχ)∗ ψdx

This also true for yb and b

z.
2. Momentum Operators: we only deal with pbx but the result will also apply to pby,z .
Z Z
∂ψ
χ pbx ψdx = −ih̄ χ ∗
∗
dx now integrate by parts
∂x
∂ χ∗
Z
= +ih̄ ψdx dropping boundary terms
∂x 
∂χ ∗
Z 
= −ih̄ ψdx using x∗ = x
∂x
Z
= ( pbx χ)∗ ψdx (5.9)
Therefore, pb†x = pbx . All momentum operators are Hermitian.
3. Hamiltonian:

Hb = 1 pb2x +V (x)
2m
Since m and V (x) are real we find

Hb † = 1 pb†x pb†x +V (x)∗ = 1 pb2x +V (x) = H b

2m∗ 2m
which confirms our earlier proof using integrations by parts.
4. Consider the operator A b = xbpbx . Using the properties above we find
b† = (b
A x pbx )† = pb†x xb† = pbx xb 6= A
b

The problem here is that at the last step we cannot swap the positions of x and pbx do not
commute! Had we considered a symmetrised version A b = 1 (bbx + pbx xb) we would have
2 xp
had more luck. Please check for yourself!
70 Chapter 5. Hermitian Operators & Exp Thm II
bz = xbpby − ybpbx :
5. Angular Momentum Operators L

bz† = pb†y xb† − pb†x yb† = pby xb− pbx yb = xbpby − ybpbx = L
L bz

Here in the last step we were allowed to swap the positions of operators because the
operators are commuting. Hence, we conclude that all Angular Momentum operators
are Hermitian.
 5.3 General Properties of General Linear, Hermitian Operators 71

5.3 General Properties of General Linear, Hermitian Operators

5.3.1 Eigenvalues and Eigenstates
b† have real eigenvalues λn and
b=A
Theorem 5.3.1 ALL Linear and Hermitian Operators A
orthonormal eigenstates (eigenfunctions) ϕn (x):

b n (x) = λn ϕn (x)
Aϕ

For this reason these are our preferred operators as they correspond (potentially) to physical
observables.
We now perform the proofs which closely follow the proofs presented earlier for H,
b but we
will use the more compact/efficient BRA-KET notation. Note that these results are independent
of the number of dimensions we are working in and are completely general.

b n (x) = λn ϕn (x) ⇒ A|ϕ

Aϕ b n i = λn |ϕn i , (5.10)

where we will assume that all eigenstates are normalised. Recall that the Hermiticity condition
implies in BRA-KET notation

hχ|A|ϕi
b = hχ|Aϕi
b = hAχ|ϕi
b

Now we compute in two different ways:

(1)

hϕi | A|ϕ
b j i = hϕi |λ j ϕ j i = λ j hϕi |ϕ j i (5.11)
| {z }
=λ j ϕ j

(2) this is equal to

b |ϕ i = hλi ϕi |ϕ j i = λi∗ hϕi |ϕ j i
h Aϕ (5.12)
|{z}i j
λ i ϕi

Now we subtract the RHSs of these two equal equations to get the master equation

(λ j − λi∗ )hϕi |ϕ j i = 0 (5.13)

Finally consider the two cases:

• For i = j we get (λi −λi∗ ) hϕi |ϕi i = λi −λi∗ = 0 from which we learn that all eignevalues
| {z }
=1
λi are real numbers λi = λi∗ .
• For i 6= j we have (λ j − λi∗ )hϕi |ϕ j i from which we infer hϕi |ϕ j i = δi j , the eigenstates
| {z }
6=0
are orthonormal.
 72 Chapter 5. Hermitian Operators & Exp Thm II

5.3.2 Generalised Expansion Theorem and Generalised Measurement Theorem

This theorem apply to all Hermitian operators not only the Hamiltonian H. b Still, the Hamil-
tonian always plays a special role as it is controls the time-evolution of the wave function –
other operators do not provide that.
For now we fix time at t = 0 and any state can be expanded in eigenstates of any hermitiann
operator A
b

N
Ψ(x, 0) = ∑ di ϕi (x) (5.14)
i=1

where the expansion coefficients are

Z
di = ϕi∗ (x)Ψ(x, 0)dx = hϕi |Ψ(x, 0)i . (5.15)

These have a well-known interpretation:

|di |2 . . . probability to find in a measurement of A
b the eigenvalues λi

Also expectation values can be evaluated with ease

hAi
b = hΨ(x, 0)|A|Ψ(x,
b 0)i = ∑ |di |2 λi (5.16)
i

All proof are the same as for A

b and will not repeated here.

Generalised Measurement Theorem

If the system is in the state Ψ(x, 0) at t = 0 and a measurement is performed at t = 0 of the
observable Ab then the outcome will be one its eigenvalues λi with probability |di |2 .

Collapse of wave function: In the measurement process the general state Ψ(x, 0) = ∑i di ϕi (x)
and the state immediately after the measurement (still at t = 0) will be Ψa f ter (x, 0) = ϕi (x) if
λi was found.

IMPORTANT: there is crucial difference compared to the measurement of energy. In general

b are not energy eigenstates ϕi (x) 6= ψn (x). So we have to think harder to
the eigenstates of A
obtain the state Ψa f ter at later times t > 0.

Ψa f ter (x,t > 0) 6= ϕi (x)e−iEit/h̄ (5.17)

The reason is that the time-evolution is governed by uniquely by H

b and not A.
b So we have
to expand

Ψa f ter (x, 0) = ϕi (x) = ∑ cn ψn (x)

n
5.3 General Properties of General Linear, Hermitian Operators 73

in terms of energy eigenstates where the coefficients are (as usual)

Z
cn = ψn∗ (x)ϕi (x) = hψn |ϕi i .

Only now can we infer the wave function at later times

Ψa f ter (x,t) = ∑ cn ψn (x)e−iEnt/h̄ ,

n

b for t > 0.
and because of the different exponential factors this will not be an eigenstate of A
Comments:
• If we measure energy and find En then all subsequent energy measurements give the
same outcome.
Not so for other observables since the state Ψ(x, 0) = ϕi (x) immediately after the
measurement is an eigenstate of A b but not of H,
b hence, with time it evolves into a
much more complicated state that is a general linear combination of the ϕ j , and not an
eigenstate of A.
b This is exactly the behaviour we observed earlier for the 2-state system
and is the mechanism that underlies neutrino oscillation.
• If we measure A b again at t > 0 what is the probability Pi (t) to find λi ?

Answer: Z
Pi (t) = |di (t)| = | ϕi∗ (x)Ψa f ter (x,t)dx |2
2
(5.18)
| {z }
How much of ϕi is in Ψa f ter ?
2
= hϕi |Ψa f ter (x,t) (5.19)
• This is actually an answer to a more general question. Assume the system is in a general
state Ψ(x,t) (it really does not matter what happened at earlier stages) and at a random
time t you measure a observable A b of your choice, what is the probability to find any of
its eigenvalues λi .
Z
Pi (t) = |di (t)|2 = | ϕi∗ (x)Ψ(x,t)dx |2 = |hϕi |Ψ(x,t)i|2 (5.20)
| {z }
How much of ϕi is in Ψ?
 6. The Simple Harmonic Oscillator

6.1 Intro
The career of a young theoretical physicist consists of treating the harmonic oscillator in
ever-increasing levels of abstraction, Sidney Coleman.

Without exaggeration the simple harmonic oscillator (also SHO in short) is the most
important system in physics and describes a wide range of phenomena from children enjoying
a swing to the small oscillations of the electro-magnetic field that give rise to photons which
carry off energy in multiples of h̄ω.
Physicists love the SHO because it is one of the rare systems that we can fully understand,
because it is solvable. This means we can know everything about it, we can solve it completely
without approximations. And unlike earlier courses I will introduce you to a nifty, "grown-up"
approach that will completely avoid having to solve any tedious second order differential
equations. This is a first example of how to fully exploit the power of operators and do
everything algebraically.
Let us first recap some of the facts you know already. First its total energy is
1 2 1
H = K +T = p + mω 2 x2 , (6.1)
2m x 2
where we used the relation between the spring constant K and the angular frequency
K
ω2 = . (6.2)
m
Using the operator postulate (reference) you can find directly the quantum mechanical Hamil-
tonian
2 2
b = 1 pb2x + 1 mω 2 x2 = − h̄ ∂ + 1 mω 2 x2 ,
H (6.3)
2m 2 2m ∂ x2 2
76 Chapter 6. The Simple Harmonic Oscillator

and in earlier courses you have seen how to solve this with "brute-force" methods by solving
the associated TISE which is a second order differential equation
b n (x) = En ψn (x) ,
Hψ (6.4)
with the result En = h̄ω(n + 12 ) for n = 0, 1, 2, . . . and
2 x2 /2 mω
ψn (x) = e−β Hn (x) , β 2 = . (6.5)
h̄

Figure 6.1: Illustrating the energy levels and the first two energy eigenstates of the SHO
You might remember that in order to arrive at the energy eigenvalues one had to impose
normalisability, well every respectable wave function has that property, i.e. only for fine-
tuned values of the En the infinite series for Hn would collapse to polynomials (Hermite’s
polynomials) with a finite number of terms. This technical procedure obfuscates the nature of
the eigenfunctions and begs the question: "Can we do better than this?"
The good news is: "Yes, we can!"
As a first take away lesson you should remember that for n even/odd the ψn (x) are even/odd
functions under parity, which transforms x → Px = −x. The simplest examples are
2 x2 /2 2 x2 /2 2 x2 /2
ψ0 = N0 e−β , ψ1 = N1 xe−β , ψ2 = N2 (2β 2 x2 − 1)e−β ,... (6.6)
Another important feature is that even if the system is in the groundstate the energy E0 = h̄ω/2
is above the classical groundstate energy E = 0, this is a common feature for particles trapped
in potentials and has been causing theoretical physicists a conceptual headache for almost a
century.
What’s the claim to fame of the SHO?
1. Light comes in finite chunks of energy. If you unify special relativity and quantum
mechanics then all fields, for example the electromagnetic field, are promoted to a
collection of an infinite number of harmonic oscillators with arbitrary frequencies ω –
after all light can come with arbitrary wavelength. Now in this interpretation the state
ψn corresponds to a particular excitation of the electromagnetic field with n photons
of frequency ω. The total energy is En − E0 = nh̄ω so the energy of one photon is h̄ω.
That is exactly what Planck postulated in order to derive blackbody radiation, but you
now know why it is quantised. He didn’t have a clue at his time.
6.1 Intro 77

2. Approximation of particles moving around the minimum of potential wells. (figure) In

stable atoms electrons are trapped in an effective and usually very complicated potential
for which the TISE is impossible to solve. Usually the electron like to hang around the
groundstate unless they are excited, so we can approximate the full potential by a Taylor
series around the minimum of the potential. Assume that we have shifted our coordinate
system such that minimum sits at x = 0 and at the moment we focus on a 1-dimensional
SHO, so that
1
V (x) = V (0) +V 0 (0) x + V 00 (0)x2 + . . . , (6.7)
| {z } 2
=0

where we have suppressed "higher order" terms like 16 V 000 (0)x3 + 16 V 0000 (0)x4 + . . . which
is valid if x is small. Notice that the linear in x term vanishes since we expand around
the minimum and the potential just looks that of a SHO with 12 V 00 (0) = 12 mω 2 and the
groundstate shifted by V (0). Hence the energy eigenvalues near the groundstate can be
approximated by
r
1 V 00 (0)
En = V (0) + h̄ω(n + ) , with ω = . (6.8)
2 m

3. The anharmonic oscillator: If you are more ambitious you might ask: “Can we do better
than this and include the "higher-order" terms?" The answer is: "Yes, we can!" But this
only can be done in an approximate fashion and uses time-independent perturbation
theory. It simply says that if I perturb the Hamiltonian by adding a small piece to the
potential V (x) → V (x) + δV (x) then the energy eigenvalues are shifted by the amount
Z
δ En = ψn∗ δV (x)ψn dx . (6.9)

In other words you need to evaluate the expectation value of the perturbing potential
δV (x) in the unperturbed energy eigenstates, which we know explicitly. We will have a
look at this later and is a cute precursor of the sort of computations you do in quantum
field theory that lead to the famous Feynman diagrams.
 78 Chapter 6. The Simple Harmonic Oscillator

6.2 SHO in Operator Formalism

We now set out to solve the SHO in a completely algebraic fashion
1. Solve SHO without solving second order differential equations and imposing boundary
conditions.
2. Operator Formalism: Key concept, introduce raising/lowering operators also called
ladder operators. Very useful! See also later for Angular Momentum Operators.
3. Similar operators appear in Quantum Field Theory and particle physics where they
describe the creation/annihilation of particles → creation/annihilation operator.
4. Illustration of bra-ket formalism and Matrix Quantum Mechanics.
The main idea is very simple; the Hamiltonian is essentially a sum of two squares and a
sum of squares can be written as the modulus square of a complex number
a2 + b2 = (a − ib)(a + ib) = c∗ c = |c|2 .
Hence,
1 2 m 2 2
H = p + ω x
2m x 2
1 2
= (p + (ωmx)2 )
2m x
1
= (ωmx − ipx )(ωmx + ipx ) .
2m
In classical mechanics this does not gain us anything, so on to QM. But we will pull out the
factor h̄ω since we know it has dimension of energy:
b = 1 pb2x + m ω 2 x2
H
2m  2 
1 2 mω 2
= h̄ω pb + x
2h̄ωm x 2h̄
 2 
β 2 1 mω
= h̄ω x + 2 pbx with β 2 =
2
. (6.10)
2 2h̄ β 2 h̄
This leads us to introduce two new, non-Hermitian operators which are related by Hermitian
conjugation (since operators are involved now complex conjugation does not suffice):
   
1 1 1 1 ∂
ab = √ β x + i pbx = √ β x + (6.11)
2 h̄β 2 β ∂x
   
† 1 1 1 1 ∂
ab = √ β x − i pbx = √ β x − (6.12)
2 h̄β 2 β ∂x
Later it will also be useful to invert these relations and write position and momentum operators
in terms of these new operators:
1
xb = √ (b a + ab† ) (6.13)
2β
ih̄β †
pbx = √ (b a − ab) (6.14)
2
6.2 SHO in Operator Formalism 79

Now let us work out some key properties of the raising operator ab† and the lowering
operator ab. The mysterious names will become clear in a moment!
Commutator: Using the standard commutator [bx, pbx ] = ih̄ you can easily convince yourself
that

a, ab† ] = 1 .
[b (6.15)

This looks rather harmless but turns out to have rather profound implications.
Hamiltonian: We can now write the Hamiltonian of the SHO in terms of the raising and
lowering operators. But we have to be a bit careful since we are dealing with operators. Let us
look at the quantity
 
† 1 2 2 1 2 i i
ab ab = β x + 2 pbx + x pbx − pbx x
2 h̄ β 2 h̄ h̄
 
1 2 2 1 i
= β x + 2 pb2x + [x, pbx ]
2 h̄ β 2 h̄ | {z }
  ih̄
1 1
= β 2 x2 + 2 pb2x − 1 . (6.16)
2 h̄ β 2
Multiplying both sides with h̄ω we find
 
† h̄ω 2 2 1 2 h̄ω
h̄ω ab ab = β x + 2 pbx −
2 h̄ β 2 2

= H b − h̄ω (6.17)
2
so this is almost the Hamiltonian we want up to a finite shift, hence,
 
† 1
H = h̄ω ab ab +
b (6.18)
2

Note often the number operator N b = ab† ab is introduced, since in an appropriate sense it counts
the excitation or quantum number n.
So far this was simply a rewriting of the Hamiltonian but what have we gained? To
address this and finally explain the terms raising and lowering operators consider the following
commutation relations. The only thing we need to remember is (6.15) and (6.18):

b ab† ] = h̄ω[b 1
[H, a† ab + , ab† ]
2
† †
= h̄ω[b
a a, a ] + 0
b b
 

= h̄ω ab† [b
a, ab† ] + [b
a† , ab† ] ab = h̄ω ab† , (6.19)
| {z } | {z }
=1 =0
80 Chapter 6. The Simple Harmonic Oscillator

so

b ab† ] = h̄ω ab†

[H, (6.20)

Let us take a moment and appreciate what this relation means; we take an operator whose
eigenvalues and eigenstates we want to know, H b in this case, commute it with another operator,
†
ab , and get that very operator times a number, h̄ω in this case. We will see soon that this tells
us almost everything about the allowed eigenvalues of H! b
To see this let’s rewrite the commutation relation
b ab† ] = H
[H, b ab† − ab† H
b = h̄ω ab† → H
b ab† = ab† H
b + h̄ω ab† , (6.21)
b E = Eψ without assuming what the eigenvalues
and act on an eigenstate ψE that obeys Hψ
are. We want to derive them from first principles. So we find
b ab† ψE = H(b
H b a† ψE ) = ab† Hψ a† ψE ) = (E + h̄ω)(b
b +h̄ω(b a† ψE ) . (6.22)
| {zE}
=EψE

Amazing! This is an eigenvalue equation and means that by acting with ab† on the energy
eigenstate ψE we have "created" a new energy eigenstate ab† with eigenvalue E + h̄ω. In
essence ab† has raised the eigenvalue by the amount h̄ω – hence we call ab† raising operator.
We can play similar games with ab and find

b ab] = −h̄ω ab ,
[H, (6.23)

and following similar steps as above we find

b aψE ) = (E − h̄ω)(b
H(b aψE ) , (6.24)

and, hence, we call ab lowering operator, because it lowers the energy eigenvalue if it acts on
an energy eigenstate ψE .
6.2 SHO in Operator Formalism 81

Illustration of ladder operators

We learn from that that the difference between two eigenvalues is an integer multiple of
h̄ω. Therefore, the eigenvalues must have the form

En = E0 + nh̄ω , (6.25)

where n is an integer, but we still have to find out what E0 is and what the allowed values of n
are.
First, note that if we kept acting repeatedly with the lowering operator ab on an eigenstate
we would be able to generate eigenstates with arbitrarily low energy unless this process
terminates at some point. In other words there must be a lowest energy eigenvalue otherwise
the system does not have a groundstate which would be a sick situation – we would be able to
extract an arbitrary amount of energy from such a system. So we have to show there exists a
lowest energy eigenvalue E0 such that

abψE0 = 0 . (6.26)

This is easy to show by considering the expectation value of our Hamiltonian for an arbitrary,
normalised state ψ
h̄ω h̄ω h̄ω
hEi = hψ|H|ψi
b a† ab|ψi +
= h̄ωhψ|b hψ|ψi = h̄ω h(b aψi +
aψ)|b ≥ . (6.27)
2 | {z } | R {z } 2 2
=1 aψ|2 dx≥0
= |b

This shows that the energy is bounded from below and that there is a groundstate with energy
E0 = h̄ω
2 . Furthermore, we now know the values of all possible energy eigenvalues
 
1
En = h̄ω n + , n = 0, 1, 2, 3, . . . (6.28)
2
 82 Chapter 6. The Simple Harmonic Oscillator
h̄ω
Furthermore, the groundstate is the state with energy E0 = 2 which saturates the lower
bound and must obey the condition
abψE0 = 0 . (6.29)
This also guarantees that the process of acting repeatedly with the lowering operator must end
at some point, while we can keep acting with the raising operator to create states with higher
and higher energy eigenvalues.
So if we know the groundstate wavefunctions ψE0 we can generate all other eigenstates by
acting repeatedly with the raising operator:
ψE0 , ψE1 ∝ ab† ψE0 , ψE2 ∝ (b
a† )2 ψE0 , · · · , ψEn ∝ (b
a† )n ψE0 , · · · (6.30)
For the SHO we often use the more convenient convention to label eigenstates by the
quantum number n rather than the energy eigenvalue En . We also do this for states in the
bra-ket notation:
ψEn → ψn and, equivalently |ψEn i → |En i → |ni . (6.31)
Therefore we can write
ψ0 , ψ1 ∝ ab† ψ0 , ψ2 ∝ (b
a† )2 ψ0 , · · · , ψn ∝ (b
a† )n ψ0 , · · · , (6.32)
and
|0i , |1i ∝ ab† |0i , |2i ∝ (b
a† )2 |0i , · · · , |ni ∝ (b
a† )n |0i , · · · . (6.33)
Note that I have deliberately written ∝ instead of = since these relations are correct up to
normalisation.

6.3 Applications of SHO in operator formalism

In the following I want to demonstrate the power of this new formalism by going through a
number of examples.

6.3.1 Eigenfunctions reloaded

Using (6.11), (6.12) and (6.29) we want to determine now the ψn (x) without solving the TISE,
which is a second order differential equation. We start with (6.29) to find ψ0 .
   
1 1 ∂ 1 1 d
abψ0 = 0 = √ β x + ψ0 = √ β x + ψ0 (6.34)
2 β ∂x 2 β dx
dψ0
⇒ = −β 2 xψ0
dx
dψ0
⇒ = −β 2 xdx
ψ0
⇒ log ψ0 = −β 2 x2 /2 + log N0 , where log N0 is an integration const.
β 2 x2
⇒ ψ0 = N0 e− 2 . (6.35)
 6.4 Examples 83
√
β
By direct integration you can find the normalisation constant N0 = π 1/4 .
Now lets get ψ1 from ψ0 by acting with the raising operator (6.11). Here we will ignore
the correct normalisation but come back to that shortly.
 
1 1 d β 2 x2
†
ψ1 ∝ ab ψ0 = N0 √ β x − e− 2
2 β dx
√
 
1 1 β 2 x2 β 2 x2
= N0 √ β x − (−β x)) e− 2 = 2N0 β xe− 2
2
(6.36)
2 β
I hope you get the pattern. No more differential equations to be solved, only act with the
first order differential operator ab† as many times as you like. So ψ2 ∝ ab† ψ1 ∝ (b
a† )2 ψ0 ; this is
completely mechanical now and you will quickly get bored because it is so easy.

6.3.2 Raising/lowering operators and normalisation

At the end we will completely avoid using the explicit form of the ψn , but for later we need
the precise relation between different eigenstates including normalisations. Let me first quote
the main formulas and then prove one of them. In practise you do not have to memorise them,
but you will need to be able to use them:
√
ab† ψn = n + 1ψn+1 (6.37)
√
abψn = nψn−1 (6.38)
1  † n
ψn = √ ab ψ0 (6.39)
n!

Note that all ψn are assumed to have the correct normalisation. And the above means that if
ψn is correctly normalised then ψm±1 are if they are defined as above.
Let us now prove the second relation, which can be written as ψn−1 = √1n abψn . Now

1 1
hψn−1 |ψn−1 i = h(b
aψn )|baψn i = hψn |b a† abψn i using def. of Herm. conjugation
n !n
1 Hb 1 1
= hψn | − ψn i a† ab + )
b = h̄ω(b
using H
n h̄ω 2 2
 
1 h̄ω(n + 1/2) 1
= hψn | − ψn i using Hψ
b n = h̄ω(n + 1/2)ψn
n h̄ω 2
n
= hψn |ψn i = 1 Q.E.D. (6.40)
n

6.4 Examples
Now we have all the ingredients to compute all possible expectation values of all possible
operators without performing a single integral. How does this work?
First observe that any operator Ob in the system is a function of position and momentum
operators, and, hence, because of (6.13) and (6.14) is a function of ab and ab† . The only other
84 Chapter 6. The Simple Harmonic Oscillator

a, ab† ] = 1. From now on we also use bra-ket notation

identities we need are (6.37), (6.38), [b
with |ψn i = |ni and orthonormality hn|mi = δnm .

 Example 6.1 Expectation values of energy eigenstates:

Compute the expectation value of x, x2 and ∆x assuming that the SHO is in the state ψn . Using
(6.13) we find
1
a + ab† |ni
hxi = √ hn|b
2β
1  
= √ a|ni + hn |b
hn |b a† |ni
2β √
|{z} √
| {z }
n|n−1i n+1|n+1i
1 √ √ 
= √ n hn|n − 1i + n + 1 hn|n + 1i = 0 due to orthogonality (6.41)
2β | {z } | {z }
=0 =0

Next
1 1
hx2 i = 2
a + ab† )2 |ni = 2 hn|ψi ,
hn |(b (6.42)
2β | {z } 2β
≡|ψi

where

a + ab† )(b
|ψi = (b a + ab† )|ni
√ √ 
a + ab† ) n|n − 1i + n + 1|n + 1i
= (b
√ √ √  √ √ √ 
= n n − 1|n − 2i + n|ni + n + 1 n + 1|ni + n + 2|n + 1i
p p
= n(n − 1)|n − 2i + (2n + 1)|ni + (n + 1)(n + 2)|n + 2i . (6.43)

Due to orthonormality we find without effort

p p
hn|ψi = n(n − 1)hn|n − 2i + (2n + 1)hn|ni + (n + 1)(n + 2)hn|n + 2i
= 0 + (2n + 1) × 1 + 0 , (6.44)

and, hence,
1
hx2 i = hn|x2 |ni = (2n + 1) , (6.45)
2β 2

and
r
1 1
q q
∆x = hx2 i − hxi2 = hx2 i − 0 = n+ . (6.46)
β 2

6.4 Examples 85

 Example 6.2 The anharmonic oscillator:

This example will teach you the first baby steps that need to be taken go from time-indepencent
perturbation theory in Quantum Mechanics to Feynman rules. The manipulations involving
raising/lowering operators you will see below are very similar in both situations.
To set the stage, we want to consider a modification of the SHO by adding a quartic term
to the potential of the form
δV (x) = λ x4 (6.47)
so that the total Hamiltonian of the system is
b=H
H bSHO + δV (x) . (6.48)
We assume here that λ is small so we can think of this term as a small modification/perturbation
of the SHO.
Since no-one knows how to compute these corrections exactly, hence, in the following we
want to use an appropriate approximation scheme — in this case 1st order, time-independent
perturbation theory to find the modified eigenvalues of this system En = h̄ω(n + 12 ) + δ En .
As mentioned earlier the corrections to the eigenvalues are given by
Z
δ En = ψn∗ δV (x)ψn dx = hn|δV (x)|ni . (6.49)

I will show this only for the groundstate, but you can easily generalise this to n > 0. So we
need to work out
λ
δ E0 = h0|δV (x)|0i = λ h0|x4 |0i = 4 h0|(b a + ab† )4 |0i (6.50)
4β
If we blindly expand the 4-th power of the bracket we will find 16 terms including ab4 , (b a† )4 ,
abab† abab† etc. (please complete the list!). However, before evaluating these 16 terms further it
pays to pause and think what raising and lowering operators do! Because of orthonormality
you only get a non-zero contribution if after the four operators in each term have acted on |0i
you land again on the same state. Obviously this means that only terms with an equal number
of raising and lowering operators can possibly contribute! Just think of hopping up and done a
ladder according to the raising/lowering operators and each rung corresponds to an energy
eigenstate.1
This leaves only six options: ab2 (b a† )2 , abab† abab† , ab† ab2 ab† , ab† abab† ab, ab(b
a† )2 ab, (b
a† )2 ab2 . In our
case we sandwich them between h0| and |0i, hence, the last three operators in the list do not
contribute since ab|0i = 0. On the other hand
√ √ √
ab2 (ba† )2 |0i = ab2 ab† |1i = ab2 ( 2|2i) = ab( 2 2|1i) = 2|0i (6.51)
† † † †
abab abab |0i = abab ab|1i = abab |0i = ab|1i = |0i (6.52)
† 2 † † 2 †
ab ab ab |0i = ab ab |1i = ab ab|0i = 0 (6.53)
|{z}
=0
1 This
is also behind the vanishing of hxi or any expectation number of an odd power of x. If you have an odd
power of operators you can never have the same number of raising and lowering operators, hence, you always get
zero.
86 Chapter 6. The Simple Harmonic Oscillator

Hence, we find

a + ab† )4 |0 = 1h0|0i + 2h0|0i + 0 = 3 ,

h0|(b (6.54)

and therefore we find the correction to the groundstate energy of the SHO to be

λ 3λ 3λ h̄2
δ E0 = h0|δV (x)|0i = λ h0|x4 |0i = h0|(b
a + a † 4
) |0i = = . (6.55)
4β 4 4β 4 4m2 ω 2
b

With some more effort you find for general eigenstates

λ † 4 (6n2 + 6n + 3)λ (6n2 + 6n + 3)λ h̄2

δ En = hn|(b
a + a ) |ni = = . (6.56)
4β 4 4β 4 4m2 ω 2
b

 Example 6.3 The SHO as an example of Matrix Quantum Mechanics:

Finally we wish to discuss the SHO as a beautiful example of Matrix Quantum Mechanics.
Let us first recap that the matrix corresponding to a general operator O
b has components given
by

Omn = hm|O|ni
b , (6.57)

where the indices m and n run over all possible eigenstates which means that for the SHO
m, n = 0, 1, 2, 3, · · · . One of the simplest examples is the Hamiltonian

Hmn = hm |H|ni
b = h̄ω(n + 1/2)hm|ni = h̄ω(n + 1/2)δmn (6.58)
| {z }
=h̄ω(n+1/2)|ni

 
h̄ω/2 0 0 0 ···

 0 3h̄ω/2 0 0 ··· 

H =
 0 0 5h̄ω/2 0 ··· 
 (6.59)

 0 0 0 7h̄ω/2 · · · 

.. .. .. .. ..
. . . . .

State (vectors) representing the energy eigenstates are

   
1 0
 0   1 
   
ψ0 ↔ u(0) =  0  , ψ1 ↔ u(1) =  0  , . . . (6.60)
   
 0   0 
   
.. ..
. .

in terms of which the TISE becomes a matrix eigenvalue equation

b n = En ψn ↔ Hu(n) = En u(n)
Hψ (6.61)
6.4 Examples 87

Matrix corresponding to lowering operator

√ √
ab → Amn = hm |b a|ni = nhm|n − 1i = nδm,n−1 (6.62)
√
|{z}
n|n−1i

this can be written as

 √ 
0 1 0 0 ···
 0 0 √2 0 ··· 
 √ 
A= 0 0
 0 3 ··· .

(6.63)
 0 0
 0 0 ··· 

.. .. .. .. ..
. . . . .
Similarly for the raising operator
√ √
ab† → A†mn = hm |ba† |ni = n + 1hm|n + 1i = n + 1δm,n+1 (6.64)
√
| {z }
n+1|n+1i

this can be written as

 
√0 0 0 0 ···
 1 0 0 0 ··· 
 0 √2 0
 
†
A = √ 0 ··· .

(6.65)
 0
 0 3 0 ··· 

.. .. .. .. ...
. . . .

General operators O b are functions of ab and ab† and will contain arbitrary products of these.
It is completely trivial to convert them to matrix operators O since we only have to replace the
raising/lowering operators by the matrics A/A† .
abab† abab† ↔ AA† AA† . (6.66)
General states and general expectation values: General states are described by infinite
dimensional, complex column vectors
 
c0
 c1 
∞  
ψ(x) = ∑ cn ψn (x) ↔ C =  c2  (6.67)
 
n=0
 c3 
 
..
.

For such a general states the expectation value of general operators O

b can be computed
immediately from their matrix form Omn :
∞
hOi
b = hψ|O|ψi
b = ∑ c∗m Omn cn = C† OC , (6.68)
m,n=0

where in the last expression matrix multiplication is assumed. 

 7. Operators and Symmetries

7.1 Compatible Operators and Simultaneous Eigenstates – A fresh

look at commuting operators
The (by now familiar) generalised Heisenberg uncertainty principle
1 b b
∆A∆B ≥ |h[A, B]i| (7.1)
2
implies that two observables A and B can only be measured simultaneously to 100% if

[A, b =0,
b B] (7.2)

and, hence,

∆A∆B ≥ 0 . (7.3)

Importantly, note that ∆A = 0 implies that ψ is an eigenstate of A,

b and at the same time
∆B = 0 implies that ψ is an eigenstate of B.
b
Question: under which condition do we have simultaneous eigenstate?
b = aψ and Bψ
Aψ b = bψ (7.4)

Definition 7.1.1 A
b and Bb are compatible if there exists a complete set of eigenstates ψn
such that each ψn is eigenstate of both A
b and the Bb simultaneous eigenstates

b n = an ψn & Bψ
Aψ b n = bn ψn (7.5)

Claim: Two operators are compatible if and only if they commute [A, b = 0.
b B]
90 Chapter 7. Operators and Symmetries

Proof: First, show that if two operators are compatible then they commute. After that we will
show the converse.
If A
b and Bb are compatible this is equivalent to the definition above. Hence,

A b n = A(b
bBψ b n ψn ) = an bn ψn (7.6)
b n = B(a
BbAψ b n ψn ) = bn an ψn = an bn ψn (7.7)
and taking the difference of these equations we find:
bBb − BbA)ψ
(A b n = [A, b n=0.
b B]ψ (7.8)

Since this is true for a complete set of wave functions ψn and any wave function can be written
as a linear combination ψ = ∑n cn ψn , we can conclude that this actually is true for an arbitrary
wavefunction

[A, b = ∑ cn [A,
b B]ψ b n=0.
b B]ψ (7.9)
n

Since this is true for any wave function, we obtain an operator equation

[A, b =0.
b B] (7.10)

Now we need to prove the converse. Let us assume that [A, b = 0 and Aψ
b B] b n = an ψn . This
implies

A b n = A(
bBψ b n ) = Bb Aψ
b Bψ b = an (Bψ
b n) (7.11)
|{z}n
=an ψn

and looking at the second and last term we realise that Bψ

b n is also an eigenstate of A
b with the
same eigenvalue an . If we assume that the eigenvalues are non-degenerate have to conclude
that ψn and Bψ
b n must be equal up an overall proportionality constant. Mathematically this
means
b n = bn ψn ,
Bψ (7.12)

leads to the desired conclusion, namely, that if [A, b = 0 then there exists a complete set of
b B]
simultaneous eigenstates:

[A, b =0
b B] ⇔ b n = an ψn & Bψ
Aψ b n = bn ψn (7.13)

R
1. In this proof we assumed that the eigenvalues are non-degenerate i.e. that for
each eigenvalue an there is a unique eigenstate ψn up to an overall constant factor.
An eigenvalue would be called degenerate if there are more than one, linearly
independent eigenstates ψn;i with i = 1, 2, . . . dn such that Aψ
b n;i = an ψn;i for the
same eigenvalue.
7.1 Compatible Operators and Simultaneous Eigenstates – A fresh look at
commuting operators 91
2. Sometimes there exist more than two compatible observables.
3. An important concept is that of a Complete Set of Commuting/Compatible Ob-
servables. For a given system this is defined as the largest set of commuting
operators (and is in general not unique). Importantly, the eigenvalues of these
operators uniquely specify the eigenstates. For example if there are degenerate
energy eigenstates then we cannot distinguish them by there energy eigenvalues but
we are guaranteed that there is one or more additional operators whose eigenvalues
uniquely label the energy eigenstates.
An important example are the energy eigenstates of the hydrogen atom. For each
eigenvalue En there are n2 different (degenerate) eigenstates where n = 1, 2, 3, · · ·
is the principal quantum numbers. In this case the Complete Set of Operators
is {H, b2 , L
b ~L b } with eigenvalues {E , h̄2 l(l + 1), h̄m}. This is why we label the
z n
hydrogen energy eigenstates by three labels

ψnlm (r, θ , ϕ) = Rnl (r)Ylm (θ , ϕ) . (7.14)

You might have heard about this in previous courses and we will discuss this more
later!
4. Final comment – this will be discussed at length in the next section: Operators that
commute with the Hamiltonian H b are related to symmetries! And symmetries, as
you already know from classical mechanics are related to conserved quantities.
 92 Chapter 7. Operators and Symmetries

7.2 Commutators and Symmetries

We now want to explore in detail why certain operators commute with each other and what
this means. In the important case when one of the operators is the Hamiltonian we will find a
deep connection to symmetries and conserved quantities i.e. a quantum version of Nöther’s
Theorem.

7.2.1 Parity
One of the simplest example of a symmetry that exists in any dimension is parity with
corresponding operator P:
1D P : x → Px = −x and 3D P :~x → P~x = −~x (7.15)

Definition 7.2.1 Symmetry: is a transformation of the coordinates ~x (that could be a

translation, rotation or parity) that leaves the Hamiltonian H
b unchanged.

For now we stay in 1D so

2 2
Hb = − h̄ ∂ +V (x) (7.16)
2m ∂ x2
Under parity
∂ ∂ ∂
x → −x : → =− (7.17)
∂x ∂ (−x) ∂x
∂ 2 ∂ 2
→ (7.18)
∂ x2 ∂ x2
so the kinetic term is always invariant. However,
V (x) → V (−x) (7.19)
which is only invariant if V (−x) = V (x) or in other words if the potential is an even function.
You know already two important examples: the Simple Harmonic Oscillator (SHO) and the
(in)finite square well. Importantly, the eigenstates of both systems are either even or odd! This
is of course no coincidence...

Figure 7.1: Examples of parity symmetric and non-symmetric potentials.

7.2 Commutators and Symmetries 93

Let us now look at the commutator of the Hamiltonian and the parity operator. For this we
need to note that
Pψ(x)
c = ψ(−x) . (7.20)
Now
[P,
c H(x)]ψ(x)
b = P(
c H(x)ψ(x))
b −Hb Pψ(x)
c (7.21)
= H(−x)ψ(−x)
b − H(x)ψ(−x)
b (7.22)
= 0 IF H(−x)
b b ⇒ V (−x) = V (x)
= H(x) (7.23)
So we learn that
b has Parity Symmetry ⇔ [P,
H b =0
c H] (7.24)

which, as it will turn out, is only one simple example of the very general relation between
symmetries and the existence of operators that commute with the Hamiltonian.
Now we can put to practise what we learned in the previous section: if two operators
commute then they have simultaneous eigenstates, which should imply that energy eigenstates
b must also be eigenstates of the parity operator P.
ψn (eigenstates of H) c
Let’s see how this goes: take the TISE
H(x)ψ
b E (x) = EψE (x) , (7.25)
and apply the parity operator
H(−x)ψ
b E (−x) = EψE (−x) (7.26)
If H
b is parity symmetric we find

H(x)ψ
b E (−x) = EψE (−x) (7.27)
which implies that ψE (x) and ψE (−x) have the same eigenvalue E (degeneracy). In one
dimension energy eigenvalues are non-degenerate which implies the two functions must be
equal up to an overall factor
x→−x
ψE (−x) = λ ψE (x) −−−→ ψE (x) = λ ψE (−x) (7.28)

⇒ ψE (x) = λ 2 ψE (x) ⇒ λ 2 = 1 ⇒ λ = ±1 (7.29)

So the allowed eigenvalues of P

c are ±1:

Pψ
c E (x) = ψE (−x) = ±ψE (x) (7.30)

Profound consequence: the groundstate (lowest energy state) is always Parity even
(Pψg.s. = +ψg.s. , this means it has NO NODES!!
c
94 Chapter 7. Operators and Symmetries

 Example 7.1 Double-well potential

Figure 7.2: Here we see a parity symmetric double well potential: classically there are two
minima and the groundstate is degenerate. In quantum mechanics however there exists a unique
parity even groundstate wavefunction (no degeneracy). The particle oscillates between the
two minima. The classical situation of the particle localised in one of the minima corresponds
to an excited state!

More consequences:
d c
hPi = 0 (7.31)
dt
The expectation value of the parity operator is time independent = Conserved Quantity. Later
b P]
we show that this is because [H, c = 0. We can also show this directly, for example for the
SHO:
∞
Ψ(x,t) = ∑ cnψne−iEnt/h̄ (7.32)
n=0
∞
PΨ(x,t)
c = ∑ cn(−1)nψne−iEnt/h̄ (7.33)
n=0

Due to the orthonormality of the eigenfunctions we find

Z ∞
∗
hPi
c = Ψ (x,t)PΨ(x,t)dx
c = ... = ∑ (−1)n|cn|2 , (7.34)
n=0

where as in the case of the computation of hEi in an earlier chapter the time-dependent
exponentials cancelled out completely leaving a time-independent result, although the wave
function is time-dependent!
Summary: If we have a parity symmetric system then V (−x) = V (x) ⇒ H(−x) b = H(x)
b
which implies
1. We have simultaneous eigenstates of Hb and Pc and, hence, the ψn have definite parity
(eigenvalue) equal to ±1.
 7.2 Commutators and Symmetries 95

2. [P, b =0
c H]

R Central potentials in 3D system, like the hydrogen atom, are parity symmetric since

V (~x) = V (r) , (7.35)

p ~x→−~x
with r = |~x| = x2 + y2 + z2 −−−→ r. For this reason energy eigenstates of systems with
central potentials have definite parity = ±1.

7.2.2 Translations and Rotations

A Translation is a transformation that simply shifts the coordinates by a constant amount

1D: x → x + a and in 3D: ~x →~x +~a , (7.36)

and acts on a wavefunction in 1D as

∂ ψ(x)
ψ(x) → ψ(x + a) = ψ(x) + a + O(a2 ) (7.37)
∂x
The infinitesimal change or variation of the wavefunction under an infinitesimal translation in
the x-direction
∂ ψ(x) i
δ ψ(x) ∼ a = a pbx ψ(x) (7.38)
∂x h̄
is generated by the momentum operator pbx up to some numerical factor. You see here
the connection between geometric Translations (moving objects in space) and momentum
operators – note that this story generalises to translations in the y and z directions in a natural
fashion.
Now when is the Hamiltonian invariant under translations? First note that
∂ x→x+a ∂ ∂x ∂ ∂
−−−−→ = = (7.39)
∂x ∂ (x + a) ∂ (x + a) ∂ x ∂ x

so the kinetic term is not affected at all. What about the potential? For an infinitesimal
translation we find using similar expressions above for ψ

∂V
δV (x) = a , (7.40)
∂x
∂V
so if the Hamiltonian is invariant under translations in the x-direction we require ∂x = 0!!
On the other hand in earlier chapters we showed that

∂V
[H,
b pbx ] = ih̄ . (7.41)
∂x
96 Chapter 7. Operators and Symmetries

So not unsurprisingly we conclude:

b has Translation Symmetry in x-direction ⇔ ∂ H = 0 ⇔ [ pbx , H]

b
H b =0 (7.42)
∂x

and we will prove very soon that this implies conservation of momentum in the x-direction

dh pbx i
=0. (7.43)
dt
It should not be very surprising that Rotational symmetry is linked to the orbital angular
momentum operators. E.g.

bz = −ih̄ ∂
L (7.44)
∂ϕ
makes it clear by analogy that the angular momentum operator around the z-axis is related
to rotations around the z-axis, which are nothing but shifts of the angle ϕ in spherical polar
coordinates.
So even less surprisingly we conclude:

b has complete Rotational Symmetry ⇔ [~L,

H b =0
b H] (7.45)

if there is rotational symmetry around all three axis.

If the symmetry is only around the z-axis then:

b has Rotational Symmetry around z-axis ⇔ [L

H bz , H]
b =0 (7.46)

Can you state what are the corresponding conserved quantities in these cases?
 7.3 Comments on degeneracy 97

7.3 Comments on degeneracy

If a particle in 1D is moving in a non-trivial potential V (x) then the energy eigenstates are
non-degenerate. This means for each different En there is a unique (up to an overall numerical
factor) energy eigenstate ψn (x).
In higher dimension (2D or 3D) the situation changes and energy eigenvalues En can be
degenerate. This would mean that there exist several linearly independent eigenstates ψn,i
with i = 1, . . . , dn where the number of independent eigenstates dn is called the degeneracy of
the eigenvalue dn , which all obey
b n,i = En ψn,i .
Hψ (7.47)

Of course such things never happen without a reason...

Theorem 7.3.1 The energy eigenvalues of H

b will in general be degenerate (although there
may be isolated exceptions) if there exist at least two operators A
b and Bb that commute with
b (symmetries) but which do not commute with each other [A,
H b 6= 0.
b B]

Proof. Since [H, b = 0: there exist simultaneous eigenstates ψn,i of H

b A] b and Ab with energy
eigenvalue En .
Also, since [H, b = 0: there exist simultaneous eigenstates ψ 0 of H
b B] b and A b with energy
n,i
eigenvalue En .
But on the other hand [A,
b B]b 6= 0 which implies that they cannot have simultaneous eigenstates,
0
and therefore ψn,i 6= λ ψn,i . Note the last relation implies that there are at least two linearly
independent eigenstates for the energy eigenvalue En .
We cannot determine like this the number dn of linearly independent eigenstates (the
degeneracy), in most cases this can only be found by explicit computation. But intuitively it
should be clear that the more symmetry there is the larger the degeneracy dn . 
 Example 7.2 In the case of the Hydrogen atom (central potential) we have [H,
b Lbx,y,z ] = 0
(rotational symmetry about all three coordinate axis) but on the other hand angular momentum
operators do not commute [L bx , L
by ] = ih̄L
bz etc. Also recall from earlier the complete set of
2
commuting operators {H,b ~L
b ,L b }. Therefore, the energy eigenstates E of the hydrogen atom
z n
must be degenerate, and indeed in this case dn = n2 with n = 1, 2, 3, . . .. 
 98 Chapter 7. Operators and Symmetries

7.4 Heisenberg’s equations of motion

You can think of them as quantum mechanical equations of motion directly expressed in terms
of physical quantities – in this case the expectation values. Note that the Schrödinger is the
fundamental dynamical equation in quantum mechanics but is expressed in terms of the wave
function which has no direct physics meaning.

Question: How do quantum mechanical expectation values change with time?

This result will also give us a quantum mechanical version of conservation laws which
are related to symmetries. So we will get quantum versions of (Emmy) Nöther’s Theorem in
classical physics.
Without loss of generality we can perform the derivation in 1D. Consider the general state
Ψ(x,t), a solution of the TDSE:

∂Ψ i b
∂ Ψ  ∂t = − h̄ HΨ

b = ih̄
HΨ ⇒ (7.48)
∂t  ∂ Ψ∗ = i (HΨ)
 b ∗
∂t h̄

where b ∗ = H.
we Rused H b In the following we also set for convenience Ψ(x,t) ≡ Ψ and
R +∞
−∞ dx ≡ dx. Now:

dhAi d
Z
Ψ∗ HΨdx
b
= b (7.49)
dt dt
∂ Ψ∗ b
Z  
∂A
Z
∗ b∂ Ψ
dx + Ψ∗ Ψdx
b
= AΨ + Ψ A (7.50)
∂t ∂t ∂t
 
* +
i  b ∗b ∂A
Z
∗bb
b
= (HΨ) AΨ −Ψ A HΨ + (7.51)

h̄ | {z } ∂t
Ψ∗ H
b AΨ
b
* +
i ∂A
Z h i
Ψ∗ H
b
= b− A
bA bHb Ψdx + (7.52)
h̄ ∂t
* +
i ∂A
Z h i
Ψ∗ H,
b
= b Ψdx +
b A (7.53)
h̄ ∂t
* +
i Dh b biE ∂A
b
= H, A + (7.54)
h̄ ∂t

This is the so-called Heisenberg equation of motion for the operator A

b
* +
dhAi
b i Dh b biE ∂A
b
= H, A + (7.55)
dt h̄ ∂t
7.4 Heisenberg’s equations of motion 99
∂A
In most cases the operator is time-independent = 0, in which case this simplifies to
b
∂t

dhAi
b i Dh b biE ∂A
b
= H, A IF =0 (7.56)
dt h̄ ∂t
The important conclusion is that if [H, b = 0, which is due to some symmetry as discussed
b A]
earlier in this chapter, then
dhAi
b
=0. (7.57)
dt
This means hAi
b does not change with time, or in other words A
b is a conserved quantity!
 Example 7.3 Translation invariance in all three directions:
b ~bp] = 0 = ih̄~∇V (~x)
[H, (7.58)
⇒ V (~x) = const., which implies that the system and the Hamiltonian H
b is translation invariant.
⇒ dtd h~bpi = 0 or in words momentum in all three dimensions is conserved. This corresponds
to a free particle moving in 3D. 

 Example 7.4 Rotation invariance around all three axis:

b ~L]
[H, b =0 (7.59)
p
⇒ V (~x) = V (r) with r = x2 + y2 + z2 (Central Potential). This implies that the system and
the Hamiltonian Hb is symmetric under rotations around all three coordinate axis. ⇒ d h~Li
b =0
dt
or in words angular momentum around all three axis is conserved. 

R These two examples were special cases. In general we can have situations where
translations only in some of the three directions are symmetries and/or rotation only
around one or two directions are symmetries.
The general method to find out is to check all the commutators [H,
b pbx,y,z ] and [H,
b Lbx,y,z ].
But often you can use intuition and direct inspection. For example let us consider a
particle moving in 3D with potential
V (~x) = ax2 , (7.60)
where a is some real, positive number. Since the potential depends on x but not the other
coordinates, translations in the directions y and z are still symmetries but not in the x
direction.
What about rotations? Rotation around the x-axis mix y and z which does not appear in
V , so that rotation will be a symmetry. However, the other two rotations will mix x with
y or z so they cannot be symmetries. So we get three conserved quantities in this case:
h pby i, h pbz i, hL
bx i. Computing the corresponding commutators you would find the same
result.
Question: what would you find for
V (~x) = a(x2 + y2 ) ? (7.61)
 100 Chapter 7. Operators and Symmetries

7.5 Ehrenfest’s Theorem

This was used in the early days of quantum mechanics to justify the form of, for example, the
momentum operator. Importantly it gives a connection between classical equations of motion
(Newton) and quantum mechanical (Heisenberg) equations of motion. It guarantees that in the
classical limit (large quantum numbers or h̄ → 0) quantum mechanics reproduces the classical
equations – otherwise we would not be convinced about our new theory, would we?
We are considering only the 1D version, with H b = 1 pb2x +V (x) but extensions to 2D and
2m
3D are immediate.
b = x. The corresponding Heisenberg equation of motion is
1. Ehrenfest relation: choose A

d i b i 1 1
hxi = h[H, x]i = h[ pb2x , x]i = h pbx i (7.62)
dt h̄ h̄ |2m{z } m
=− ih̄
mp bx

b = pbx . The corresponding Heisenberg equation of motion is

2. Ehrenfest relation: choose A

 
d i b i ∂V (x)
h pbx i = h[H, pbx ]i = h[V (x), pbx ]i = − (7.63)
dt h̄ h̄ | {z } ∂x
=ih̄ ∂V
∂x

Where we have used throughout commutators that can be computed with methods developed
in earlier chapters.
These two relations (at least at first sight) have striking similarities with the classical
relation if we make the following replacements/identifications
xcl ↔ hxi
(7.64)
px,cl ↔ h pbx i
So the 1. Ehrenfest relation becomes
dxcl 1
= px,cl , (7.65)
dt m
which is as expected. But what about the connection of the 2. Ehrenfest relation with Newton’s
3rd law? Let us start with the latter and rewrite it in terms of expectation values:
d ∂V (xcl ) d ∂V (hxi)
px,cl = − −→ h pbx i = − = Fx (hxi,t) (7.66)
dt ∂ xcl dt ∂ hxi
So naively we expect the right hand side of Newton’s law would be the force (= minus the
gradient of the potential) evaluated at the expectation value (average) of the positions. But
instead, the 2. Ehrenfest relation gives
 
d ∂V (x)
h pbx i = − , (7.67)
dt ∂x
7.5 Ehrenfest’s Theorem 101

which is the expectation value of minus the gradient of the potential, or the quantum mechanical
average of the classical force. Hence, these two expressions are not identical, but maybe this
was asking too much. After all classical physics is not equal to quantum mechanics, we should
only recover it in appropriate situations.
Question: under which conditions do we recover the classical/naive expectation?
Let us look more closely at
 
∂V (x)
− = hFx (x))i , (7.68)
∂x
and Taylor expand the force arouns x = hxi. This yields
∂ Fx (hxi,t) (x − hxi)2 ∂ 2 Fx (hxi,t)
Fx (x,t) ∼ Fx (hxi,t) + (x − hxi) + . (7.69)
∂x 2 ∂ x2
Now using that
h(x − hxi)i = hxi − hxi = 0 (7.70)
h(x − hxi)2 i = hx2 i − 2hxihxi + hxi2 = hx2 i − hxi2 = ∆x2 (7.71)
we finally find that
∆x2 ∂ 2 Fx (hxi)
hFx (x)i ∼ Fx (hxi) + . (7.72)
2 ∂ x2
Now we see that the first term is exactly the naive expectation (force at the average position),
while the second term represents true quantum corrections which probe the breadth or "spread",
∆x, of the particle’s wave function. If the particles are well localised, and hence ∆x is
small compared to distances over which the potential changes significantly (slowly varying
potential/force), then we recover the classical equations of motion.

Figure 7.3: Illustration of situations of when or when not the Ehrenfest theorem reduces to
classical equations of motion.
 8. Angular Momentum in Q.M.

8.1 Intro
• Angular momentum (AM) is only relevant in 3D (and 2D), but in 1D there are no
rotations and no AM:

b = 1 pb2x + pb2x + pb2x +V (x, y, z)

H
2m
h̄2 ∂ 2 ∂2 ∂2
 
=− + + +V (x, y, z) (8.1)
2m ∂ x2 ∂ y2 ∂ z2
| {z }
=4=~∇2

• In many quantum mechanical systems, angular momentum is conserved or almost

conserved. In particular
p this is true if the system is described by a central potential
V (x, y, z) = V (r = x + y2 + z2 )
2
• Surprisingly for physicists at the time Q.M. was developed, angular momentum is
quantised, it comes in multiples of h̄. This contradicts classical expectations but was
implicitly assumed already in Bohr’s model. Crucially, later direct experimental ev-
idence for quantisation was found. Famous examples are the Zeeman effect, which
I will now briefly describe, and the Stern-Gerlach experiment that we will discuss later:

Consider the hydrogen atom (you need to recall a few facts about this system from
earlier courses). The energy eigenvalues are
13.6eV
En ∼ − (8.2)
n2
where the principal quantum number is n = 1, 2, 3, 4, . . .. For n = 1 there is a unique state
with zero angular momentum. But for n = 2 there are 4 states, one with zero angular
104 Chapter 8. Angular Momentum in Q.M.

momentum ` = 0 and 3 with non-zero AM ` = 1. Recall, ` tells us about the length of

the AM momentum vector ~L2 ∼ h̄2 `(` + 1) and there is another quantum number, m,
that gives the size of the z-component Lz which is h̄m. For ` = 0 we have m = 0, while
for ` = 1 we have three choices m = −1, 0, +1.
Normally, for n = 2, ` = 1 and m = −1, 0, +1 there is one energy eigenvalue E2 . So
there is degeneracy because there are three different eigenfunctions with the same energy.

Zeeman effect: Things become interesting if we put the atoms in a homogeneous

magnetic field ~B = (0, 0, Bz ), because then the magnetic moment ~µ = − 2me e~L produced
by the rotating electron interacts with the magnetic field and shifts the energy by
e
∆V = −~µ · ~B = Bz Lz (8.3)
2me
Now following our earlier discussion the z-component Lz = h̄m where m can take
non-zero values if ` 6= 0 and lead to shifts of the energy eigenvalues.
e
∆V = Bz h̄m (8.4)
2me
How can this give evidence for quantisation? Classically we would assume Lz can take
any value and m would not be quantised, which would lead to a continuum range of
energy shifts, however what is observed are shifts that are integer multiples of 2me e h̄Bz .
What happens is further illustrated in the figure below

Figure 8.1: A non-zero magnetic field "lifts" the degeneracy of the E2 energy eigenvalue, and
instead of one sharp spectral line three sharp lines are seen. In classical physics we would have
expected a broadening or smearing out of the sharp line because m could be any real number.
 8.2 Orbital angular momentum ~L
b and its commutators 105

8.2 Orbital angular momentum ~L

b and its commutators
For a classical particle with momentum ~p and position vector ~x, the Orbital Angular Mo-
mentum is:
~Lcl = ~x ×~p. (8.5)
To find the quantum mechanical orbital angular momentum we follow our usual recipe to
turn classical observables in Q.M. operators – the Operator Postulate: replace all classical
dynamical variables by the appropriate operator,
 
∂ ∂ ∂
~p → ~bp = −ih̄ ~∇ = −ih̄ , , (8.6)
∂x ∂y ∂z
and ~x → ~b x = ~x = (x, y, z), (8.7)
giving the QM Orbital Angular Momentum Operator:
b =~x × ~bp = −ih̄~x × ~∇,
~L (8.8)

or, in cartesian components,1

 
∂
bx = y pbz − z pby = −ih̄ y − z ∂
L ,
∂z ∂y
 
∂ ∂
Ly = z pbx − x pbz = −ih̄ z − x
b , (8.9)
∂x ∂z
 
∂ ∂
Lz = x pby − y pbx = −ih̄ x − y
b .
∂y ∂x

First we note that, ~L

b are good quantum mechanical observables, i.e. they are Hermitian
(self-adjoint) operators (real eigenvalues, orthonormal eigenstates)
b† = L
L bi , i = 1, 2, 3 (8.10)
i

Note that we have introduced the convenient notation here whereby

b1 = L
L bx , L
b2 = L
by , L
b3 = L
bz , x1 = x , x2 = y , x3 = z and so on. (8.11)

Proof:
 †
L x pby − ybpbx )† = pb†y xb† − pb†x yb† = pby xb− pbx yb = xbpby − ybpbx = L
bz = (b bz

1I remember the cross-product by drawing a circle in my mind with the components x, y, z arranged clockwise.
The components of the cross-product, reading from left to right in the equation, always occur in cyclic order (ie.
clockwise order) for the first (positive sign) term and anti-cyclic for the second (negative sign) term. Thus, for the
z-component Lz , you start the cyclic sequence with z, giving z → x → y, giving Lz = xpy . . .; the anti-cyclic one
is z → y → x, giving Lz = xpy − ypx . This is just another way of expressing the rule based on the determinant,
but it’s easier to keep in your head without the need to write down any intermediate steps.
106 Chapter 8. Angular Momentum in Q.M.

where we have used that position and momentum operators are Hermitian and that

[xi , pbj ] = ih̄δi j , (8.12)

which, since in this case i 6= j, implies that the order of position and momentum operator can
be swapped.
Next we study the commutation relations between the three components of the angular
momentum operator using the canonical commutation relations. These state that coordinates
commute with alien components of the momentum operator, but do not commute with their
own, [xi , pbj ] = ih̄ δi j , and that coordinates commute with coordinates and momenta with
momenta, [xi , x j ] = 0, [ pbi , pbj ] = 0:

[L
bx , L
by ] = [(y pbz − z pby ), (z pbx − x pbz )] and using the distributive law for commutators,
= [y pbz , z pbx ] − [y pbz , x pbz ] − [z pby , z pbx ] + [z pby , x pbz ]
= y pbx [ pbz , z] − 0 − 0 + x pby [z, pbz ] taking out factors which commute;
= (x pby − y pbx )[z, pbz ],
= ih̄ (x pby − y pbx ), using [z, pbz ] = ih̄
= ih̄ L
bz . (8.13)

We can proceed similarly for the other possible commutators or more efficiently use cyclic
permutations of the x, y, z in the derivation above, and we find

[L
bx , L
by ] = ih̄ L
bz , (8.14)
[L
bz , L
bx ] = ih̄ L
by , (8.15)
[L
by , L
bz ] = ih̄ L
bx . (8.16)

Notice that the commutators give +ih̄ when, reading from left to right, the components are
in cyclic order just as in a cross product. In MT4/QMS you will hear that this commutator
algebra is an example of an SU(2) or SO(3) Lie algebra, but this is just jargon – this is just
saying these operators are related with rotations in 3D. However we will see later that this
structure will give us all the possible eigenvalues without solving differential equations...
Interestingly but not very usefully, these commutators can be written in an even more
succinct form as a cross product:

~L
b ×~L
b = ih̄~L.
b (8.17)

This form of the commutators shows how unintuitive non-commuting operators can be: the
cross product of two parallel ordinary vectors vanishes, but not when they are operators and
their components don’t commute!
8.2 Orbital angular momentum ~L
b and its commutators 107

More importantly let us look at the physical consequences of this failure to commute. They
are remarkable and follow from the generalised Heisenberg uncertainty relation
1 b b
∆A ∆B ≥ |h[A, B]i|,
2
6 = 0 if [A, b B]b 6= 0. (8.18)
Hence only one component of this vector, conventionally taken to be L bz , can be measured
with perfect precision (∆Lz = 0). Once this measurement has been performed, we can learn
nothing whatsoever about the other two components (∆Lx 6= 0 and ∆Ly 6= 0) because the
operator Lbz commutes with neither L bx nor Lby . Formally we can say that an eigenstate of L
bz
cannot also be an eigenstate of either L
bx or L
by .
This is very different from for example ~bp; you can measure (and know) all three compo-
nents of the momentum simultaneously to arbitrary precision!
Also there cannot be simultaneous eigenstates of L bz AND Lbx (or L
by ), or in other words
wave functions can only be eigenstates of one of the components!
But is there another operator that commutes with ~L? b
2
Yes, the square of the angular momentum, ~L b commutes with L b . Hence the length of
z
the angular momentum vector can also be measured with perfect precision along with its
b2 , which is defined as
z-component. To prove this we consider separately each term in ~L
b2 = L
~L bx2 + L
by2 + L
bz2 . (8.19)

First,2
bx2 , L
[L bz ] = L
bx [L
bx , L
bz ] + [L
bx , L
bz ]L
bx
= −ih̄(Lbx Lby + L bx ) using [L
by L bz ] = −ih̄L
bx , L by , (8.20)
Similarly,
by2 , L
[L bz ] = L
by [L
by , L
bz ] + [L
by , L
bz ]L
by
= +ih̄(Lby Lbx + L by ), using [L
bx L by , L
bz ] = ih̄L
bx (8.21)
while, finally and obviously, L
bz commutes with its own square,
bz2 , L
[L bz ] = 0 (8.22)

b2 indeed commutes with L

Adding eqs. (8.20), (8.21), and (8.22), shows that ~L bz :

b2 , L
[~L bx2 , L
by ] = [L by2 , L
by ] + [L bz2 , L
by ] + [L by ] = 0 (8.23)
2 Using the Leibniz rule for operators from Chapter 2
b2 , B]
[A b = A[
b A, b + [A,
b B] b B]
bAb
108 Chapter 8. Angular Momentum in Q.M.

Similar calculations for the other components give the same result, which we can summarise
as:

b2 , L
[~L bi ] = 0, for all i = 1, 2, 3 (8.24)

Since we have chosen to measure L b2 , L

bz exactly, the relevant commutator is [~L bz ] = 0 (i = 3),
2
which shows that we can also measure ~L
b exactly: it is possible to measure exactly both
the length and one component of the angular momentum vector, but at the price of knowing
nothing whatsoever about the other two components. This can be illustrated with Figure 1,
which shows a vector of known length and projection onto the z-axis, but which can have any
position on the circle obtained by rotating the vector around the z-axis. In formal terms our
result tells us that eigenfunctions of L b2 , but not of L
b are also eigenfunctions of ~L b nor of L
b.
z x y
Of course the choice of L
bz rather than L
bx or L
by is an arbitrary convention; the point is that only
one may be chosen from the three components because the corresponding operators do not
commute.

z
6

~L
6
K
A


A


A


A q

2
hL
bz i A
~b
A
hL i
A


A

? A

-y

x
Figure 8.2: Picturing what can be measured exactly. This quasi-classical picture is not to
b2 i are the eigenvalues of the correspond-
bz i and h~L
be taken too literally - for an eigenstate hL
ing operators. ~L is a quasi-classical vector representing the quantum mechanical angular
momentum.
 8.3 Angular momentum operators in spherical polar coordinates 109

8.3 Angular momentum operators in spherical polar coordinates

In many physical applications, such as the quantum mechanics of the hydrogen atom, we
either have exact or approximate spherical symmetry. This, together with angular momentum’s
association with rotation, demands that we consider writing down the angular momentum op-
erators in spherical polar coordinates; this also leads to a great simplification and a recognition
that the orbital angular momentum operators only involve the angular variables θ and ϕ.
We have exact rotational symmetry around all axis if our system has a central potential:
p
V (x, y, z) = V (r) with r = x2 + y2 + z2 , (8.25)
i.e. it does not depend on the angles.
Let us briefly recall the relation between cartesian and spherical polar coordinates:
x = r sin θ cos ϕ
y = r sin θ sin ϕ (8.26)
z = r cos θ
and its inverse
p
r = x2 + y2 + z2
p
x2 + y2
θ = arctan (8.27)
z
y
ϕ = arctan
x
with the usual coordinate ranges
r = 0...∞
θ = 0...π (8.28)
ϕ = 0 . . . 2π
which ensure we cover 3-space exactly once, and need to be remembered if we integrate over
all of space. For example when we want to normalise a 3D wave functions
Z ∞Z ∞Z ∞ Z ∞ Z π Z 2π
2
1= |Ψ(x, y, z,t)| dxdydz = |Ψ(r, θ , ϕ,t)|2 r2 sin θ drdθ dϕ
−∞ −∞ −∞ r=0 θ =0 ϕ=0 | {z }
dxdydz

(8.29)
Now in Q.M. we have a bunch of differential operators like the momentum ~bp and the AM
operators ~L
b that we want to express in the new coordinates. In practise this is tedious, not very
illuminating but also conceptually trivial (most textbooks will have more details on this). It is
repeated applications of the chain rule for differentiation, for example
∂ ∂x ∂ ∂y ∂ ∂z ∂ ∂
= + + , = ...
∂ϕ ∂ϕ ∂x ∂ϕ ∂y ∂ϕ ∂z ∂θ
(8.30)
∂ ∂r ∂ ∂θ ∂ ∂ϕ ∂ ∂
= + + , = ...
∂x ∂x ∂r ∂x ∂θ ∂x ∂ϕ ∂y
110 Chapter 8. Angular Momentum in Q.M.

Of course we know how to take derivatives and we are only interested in the end result, but to
illustrate this I pick the simplest example and use the first line in (8.30) together with (8.26):

∂ ∂x ∂ ∂y ∂ ∂z ∂
= + +
∂ϕ ∂ϕ ∂x ∂ϕ ∂y ∂ϕ ∂z
∂ ∂ ∂
= r sin θ (− sin ϕ) + r sin θ (cos ϕ) + 0 (8.31)
∂x ∂y ∂z
∂ ∂ ∂ ∂
= −y + x = x − y .
∂x ∂y ∂y ∂x

The keen eye now recognises that

∂ ∂ ∂
−ih̄ = (−ih̄)(x − y ) = x pby − y pbx = L
bz (8.32)
∂ϕ ∂y ∂x

So we have found that the z-component of the AM operator is very simple

bz = −ih̄ ∂
L (8.33)
∂ϕ

which intuitively can be understood by realising that rotations around the z-axis correspond
to shifts of ϕ while θ remains untouched.
We could do this also for the other components, with some more effort, but will below
only quote results (they do not have to be remembered – formulas will always be given in
exams – but they have to be used).
But first let us make a small detour: since this operator is so simple let us just find the
eigenstates of Lbz

∂ Φ(ϕ)
−ih̄ = λ Φ(ϕ) (8.34)
∂ϕ

which gives

dΦ(ϕ) λ
= i Φ(ϕ) (8.35)
dϕ h̄

The solution is very similar to a de Broglie wave (eigenfunction of momentum operator) and
is given by (please check by plugging into the differential equation)

Φ(ϕ) = ceiλ ϕ/h̄ . (8.36)

For momentum operators we would have no further restrictions on the eigenvalues, but here
we have to think again. ϕ is an angular coordinate and as such periodic: ϕ and ϕ + 2π describe
the same point in space and from Chapter 1 we remember that wave functions must be single
8.3 Angular momentum operators in spherical polar coordinates 111

valued or in other words must be periodic:

Φ(ϕ) = Φ(ϕ + 2π)

⇒ eiλ ϕ/h̄ = eiλ (ϕ+2π)/h̄ = eiλ ϕ/h̄ e|iλ{z
2π/h̄
(8.37)
}
≡1
λ
⇒ ≡ m = 0, ±1, ±2, ±3, . . .
h̄
So the eigenvalues λ = h̄m are quantised, where m is an integer and the eigenstates are

Φm = eimϕ with m integer (8.38)

This shows that the orbital AM around the z-axis is quantised – since there is nothing
special about this axis this is also true for any axis. But in this coordinate system it is easiest
to see for L
bz .
We are now ready to quote the three components of the AM operators and ~L b2 in spherical
polars (without proof):
 
bx = ih̄ sin ϕ ∂ ∂
L + cot θ cos ϕ
∂θ ∂ϕ
 
by = −ih̄ cos ϕ ∂ ∂
L − cot θ sin ϕ
∂θ ∂ϕ
(8.39)
bz = −ih̄ ∂
L
∂ϕ
∂2
2
   
b = −h̄2
~L 1 ∂ ∂ 1
sin θ + 2
sin θ ∂ θ ∂θ sin θ ∂ ϕ 2

As mentioned we will rederive the simultaneous eigenfunctions of L b2 using a different

bz and ~L
method, but we will do this a bit later once we have developed the methods.
Let us stress at this point that all the AM operators are independent of r and contain
only θ , ϕ derivatives, so

~L
b f (r) = 0 (8.40)

showing that any function that depends only on r has zero AM around all three axis.
Remember: the orbital angular momentum operator only acts on the angular variables!
 112 Chapter 8. Angular Momentum in Q.M.

8.4 The 3D Schrödinger equation in spherical polar coordinates

b2 appears also in the kinetic term of the 3D Hamiltonian: recall in Cartesian
More remarkably ~L
coordinates

b = 1 pb2x + pb2y + pb2z +V (x, y, z)

H
2m
(8.41)
h̄2 ∂ 2 ∂2 ∂2 h̄2 ~ 2
 
=− + + +V (x, y, z) = − ∇ +V (x, y, z)
2m ∂ x2 ∂ y2 ∂ z2 2m
where (without proof)

1 ∂

∂

1

1 ∂

∂

1 ∂ 2 
~∇ =
2
r 2
+ 2 sin θ + 2 .
r2 ∂ r ∂r r sin θ ∂ θ ∂θ sin θ ∂ ϕ 2
 2
∂2
    
∂ 2 ∂ 1 1 ∂ ∂ 1
= + + 2 sin θ + 2 (8.42)
∂ r2 r ∂ r r sin θ ∂ θ ∂θ sin θ ∂ ϕ 2
 2
∂ 2 ∂
 b2
~L
= + − 2
∂ r2 r ∂ r h̄ r2
Therefore we find
2
b = − h̄ ~∇2 +V (x, y, z)
H
2m
h̄2 ∂ 2

2 ∂
 b2
~L (8.43)
=− + + +V (r, θ , ϕ)
2m ∂ r2 r ∂ r |2mr
2
{z }
"Centrifugal Term"

where the middle term has a nice classical analogue, namely the classical rotational energy
~L2 ~L2
cl
= cl 2
2I 2mr
where I is the moment of inertia of a point particle of mass m rotating around the origin with
distance r and angular momentum ~Lcl .
Let us now specialise to potentials that do not depend on the direction, i.e. a central
Ze2
potentials with V (~x) ≡ V (r, θ , ϕ) = V (r). Examples of this are V (r) = − 4πε0r
for an electron
m 2 2 2 2 m 2 2
around a nucleus, and V = 2 ω (x + y + z ) = 2 ω r for an isotropic 3D simple harmonic
oscillator.
Now the Hamiltonian simplifies slightly

2 
∂2 2 ∂
 b2
~L
b = − h̄
H + + +V (r) (8.44)
2m ∂ r2 r ∂ r |2mr
2
{z }
"Centrifugal Term"
8.4 The 3D Schrödinger equation in spherical polar coordinates 113

Now remembering that the angular momentum operators only act on angle coor-
dinates but not on r we see that H
b of a central potential commutes with all angular
b2 ,~L]
momentum operators, because V (r) commutes with all AM operators and [~L b = 0!
So in summary:

For a central potential: [~L, b =0

b H] (8.45)
which means that each component and any power of the angular momentum commutes with
the potential:
For a central potential: [L
bi , H]
b = 0 i = 1, 2, 3 (8.46)
2
and [~L
b , H]
b = 0 (8.47)
Very Important Comments:

• In a central potential AM is conserved!

d ~b i b ~b
hLi = h[H, L]i = 0
dt h̄
And the reason is complete rotational symmetry around any axis.
• For systems with central potentials we have the following maximum set of commuting
operators: {H, b2 , L
b ~L b }. This guarantees that energy eigenstates of such systems are si-
z
multaneous eigenstates of these three operators and are labelled by three quantum
numbers

b Elm (r, θ , ϕ) = EψElm (r, θ , ϕ)

Hψ
b2 ψ (r, θ , ϕ) = h̄2 `(` + 1)ψ (r, θ , ϕ) ` = 0, 1, 2, . . . ,
~L (8.48)
Elm Elm
bz ψElm (r, θ , ϕ) = h̄ m ψElm (r, θ , ϕ) m = −`, −` + 1, . . . , 0, . . . , ` − 1, `.
L

Note that they cannot also be eigenstates of Lbx or L

by since they do not commute with L bz .
• The existence of this set of commuting operators and ultimately rotational symmetry is
responsible for the fact that we can use separation of variables for the energy eigenstates:

|m|
ψE`m = RE` (r)Y`m (θ , ϕ) = RE` (r)P` (θ )Φm (ϕ) (8.49)
This is not merely a mathematical trick, we can do this because of symmetry!
• Shortly, we will find a new method to find the spherical harmonics Y`m (θ , ϕ) without
solving the complicated set of equations

b2Y (θ , ϕ) = h̄2 `(` + 1)Y (θ , ϕ) ` = 0, 1, 2, . . . ,

~L `m `m (8.50)
bzY`m (θ , ϕ) = h̄ m Y`m (θ , ϕ) m = −`, −` + 1, . . . , 0, . . . , ` − 1, `.
L
114 Chapter 8. Angular Momentum in Q.M.

Notice that we could pull RE` (r) out of the 2nd and 3rd equation of (8.48) because AM
operators do not depend on r!
Instead we will use raising/lowering operators as for the SHO, and as a byproduct find
more possibilities for the allowed eigenvalues than anticipated (and missed by the brute
force method).
 8.5 The radial equation for central potentials 115

8.5 The radial equation for central potentials

Before looking at abstract AM operators and the novel method to find their eigenstates. We
will take a closer look at the TISE in the case of a central potential and have a first look at the
hydrogen atom.
First, let us look at the structure of the energy eigenstates (8.49) again
ψE`m = RE` (r)Y`m (θ , ϕ) (8.51)
which contains a universal factor, Y`m (θ , ϕ), which carries all the angular dependence and
does not depend on the shape of the potential! So once we know these, the only interesting
piece is the radial wave function RE` (r).
So let us plug ψE`m = RE` (r)Y`m (θ , ϕ) into the TISE (first line of (8.48)):

h̄2

∂ 2 ψE`m 2 ∂ ψE`m b2 ψ
~L

E`m
HψE`m = −
b + + +V (r)ψE`m
2m ∂ r2 r ∂r 2mr2

h̄2 d 2 RE` (r) 2 dRE` (r)

  b2Y (θ , ϕ)
~L `m
=− 2
+ Y`m (θ , ϕ) + RE` (r) +V (r)RE` (r)Y`m (θ , ϕ)
2m dr r dr 2mr2
h̄2 d 2 RE` (r) 2 dRE` (r) h̄2 `(` + 1)
   
= − + + RE` (r) +V (r)RE` (r) Y`m (θ , ϕ)
2m dr2 r dr 2mr2
= ERE` (r)Y`m (θ , ϕ)
(8.52)
dropping now Y`m in the last two lines we obtain the radial Schrödinger equation

h̄2 d 2 RE` (r) 2 dRE` (r) h̄2 `(` + 1)

 
− + + RE` (r) +V (r)RE` (r) = ERE` (r) (8.53)
2m dr2 r dr 2mr2
So thanks to spherical symmetry we have effectively arrived at a one-dimensional TISE!
It almost has the form of a standard 1D TISE, except for two differences:
1. First there is a first order derivative in r (second term) due to the fact that we are coming
from 3D.
2. There is the additional "centrifugal term" due to the rotational energy (third term) – it is
convenient to combine this with the actual potential into an effective potential
h̄2 `(` + 1)
Ve f f (r) = V (r) + (8.54)
2mr2
where for electrons in atoms
Ze2
V (r) = − . (8.55)
4πε0 r
Notice that if the particle has zero angular momentum (` = 0) then the potential is unmodified,
while for ` 6= 0 the centrifugal potential is strongly repulsive near r = 0 and dominates the
negative Coulomb potential.
116 Chapter 8. Angular Momentum in Q.M.

Figure 8.3: Illustration of the effective potential in the radial equation produced by a nucleus.
Note the difference between ` = 0 and ` 6= 0 (zero or non-zero orbital angular momentum).

Example 8.1 Groundstate of hydrogen atom

Let us now solve the radial equation in the simplest case Z = 1 (hydrogen) and zero angular
momentum ` = 0. The wavefunction the depends only on r:
ψE1 00 (r, θ , ϕ) ≡ ψ(r) (8.56)
Plugging in the radial equation and setting ` = 0 we get
h̄2 d 2 ψ(r) 2 dψ(r) e2
 
− + +0− ψ(r) = E1 ψ(r) (8.57)
2m dr2 r dr 4πε0 r
We solve this equation by using a trial function, this is also called making an ansatz. Obviously
we want the function to be normalisable and fall off to infinity. A natural (but not unique)
choice (usually you have to try a few times) is:
ψ = Ne−r/a (8.58)
with N a normalisation constant and a some unknown positive constant with dimension of
length. From the radial equation we now want to determine a and E1 , after that we can find N
and compute some expectation values.
For the radial equation we need (dropping for now N)
dψ 1 1 d2ψ 1
ψ = e−r/a , = − e−r/a = − ψ , 2
= 2ψ .
dr a a dr a
Plugging this in the radial equation and dividing by ψ we obtain
h̄2 1 e2
 
2
− − − = E1 (8.59)
2m a2 ar 4πε0 r
8.5 The radial equation for central potentials 117

This equation contains terms of order r0 and r1 and since we want this to hold for any r we
need to compare terms of the different orders:

h̄2 1 e2 4πε0 h̄2

from r1 we find: − =0⇒a= ∼ 5.3 × 10−11 m
2m a 4πε0 me2
(8.60)
h̄2 1 h̄2 1 me4
from r0 we find: − = E1 ⇒ E1 = − =− ∼ −13.6eV
2m a2 2m a2 2(4πε0 h̄)2

Of course a is the famous Bohr radius (the size of the electron cloud around the proton) and
we confirm that there is a finite groundstate energy despite the problematic 1/r potentially
that classically has a groundstate at −∞. So quantum mechanics has resolved successfully this
classical disaster and establishes that there are stable boundstates of protons and electrons.
However also notice that the groundstate has zero angular momentum so it is not the centrifugal
force balancing the Coulumb attraction, this is the uncertainty principle at work!
From earlier courses you might remember the relation between E1 and the energy eigen-
values for excited states (which have an n2 fold degeneracy – more on that later)

E1
En = with n = 1, 2, 3, . . . (8.61)
n2

Now we want to correctlyR normalise the wave function and compute expectation values hri
and hr2 i. Useful formula r=0
∞
e−r/b rk dr = k!bk+1 . Normalisation:
Z ∞ Z π Z 2π
1= |ψ|2 r2 sin θ drdθ dϕ
r=0 θ =0 ϕ=0
Z ∞ Z π Z 2π
2 2 −2r/a
= |N| r e dr sin θ dθ dϕ = |N|2 (2!(a/2)3 )(2)(2π) = πa3 |N|2 (8.62)
r=0 θ =0 ϕ=0
1
⇒N= √
πa3
Expectation values:
Z Z Z
n
hr i = ψ ∗ rn ψr2 sin θ drdθ dϕ
Z Z Z
= |ψ|2 rn+2 sin θ drdθ dϕ
Z ∞ (8.63)
2 −2r/a n+2
= |N| e r dr × 4π
r=0
4π
= (n + 2)!(a/2)n+3 = (n + 2)!an 2−n−1
πa3
Hence
√
3a 3
q
hri = , hr2 i = 3a2 and hence ∆r = hr2 i − hri2 = a (8.64)
2 2
118 Chapter 8. Angular Momentum in Q.M.

Figure 8.4: Sketches of |ψ|2 and the radial probability P(r) = r2 |ψ|2 for the groundstate
wavefunction of the hydrogen atom.


 8.6 Abstract angular momentum operators 119

8.6 Abstract angular momentum operators

One of the great surprises in the history of 20th century physics was the discovery of spin, a
form of angular momentum intrinsic to an isolated particle such as an electron. The surprise is
that spin cannot be understood quantitatively as the analogue of a spinning top or the earth
spinning on its axis - these are really examples of orbital angular momentum in disguise: the
material making up a top or the earth is simply orbiting about the spin axis. In fact the electron
is known experimentally to be so small that it’s impossible to imagine how it could produce
its angular momentum h̄/2 by spinning no faster than the speed of light.
This is our motivation for studying angular momentum in a purely abstract way, assuming
that the general angular momentum operator b J~ represents all forms of angular momentum:
orbital, when b J~ = ~L; J~ = ~S; and situations which occur when spinning particles are
b spin, when b b
also orbiting - such as an electron in an atom - for which the total angular momentum is the
vector sum b b + ~b
J~ = ~L S.

Goals:
1) Derive the allowed eigenvalues of the abstract AM operators b J~ algebraically, not assuming
that the operators are differential operators or matrices or something else.
2) Find a better method to find the Y`m (θ , ϕ) using general properties of the operators similar
to the SHO.
We only assume that angular momentum in QM is represented by the Hermitian opera-
tors Jbx , Jby , Jbz which obey the same commutation relations as orbital AM:

[Jbx , Jby ] = ih̄Jbz , [Jby , Jbz ] = ih̄Jbx , [Jbz , Jbx ] = ih̄Jby (8.65)
2
From these equations we can deduce that b J~ commutes with all the components of b ~ The
J.
proof follows exactly the same steps for the analogous proof for orbital angular momentum
operators. Hence, we find
2
J~ , Jbi ] = 0,
[b for all i = 1, 2, 3 (8.66)

Simultaneous eigenstates (normalised): | jm j i ≡ ψ jm j with h jm j | jm j i = 1:

2
J~ , Jbz ] = 0
[b
2
(8.67)
J~ | jm j i = h̄2 j( j + 1)| jm j i
b

Jbz | jm j i = h̄m j | jm j i

with as yet unknown eigenvalues h̄2 j( j + 1) and h̄m j respectively

√ (we only know they must
be real numbers but they could be non-integers like π or 2).
120 Chapter 8. Angular Momentum in Q.M.

Using the commutation relations alone we want to derive the allowed values of j and
m j algebraically; this will lead to a surprise, a new possibility that can only be found by
working with abstract angular momentum operators.
In order to simplify things we rename for now

mj ≡ m .

and switch back to the original convention after the derivation.

Step 1: m2 ≤ j( j + 1) or m is bounded

Assume the system is in the AM eigenstate | jmi, then

2 2
J~ i = h jm|b
hb J~ | jmi = h jm|h̄2 j( j + 1)| jmi = h̄2 j( j + 1)
= hJbx2 i + hJby2 i + hJbz2 i ≥ m2 h̄2 , (8.68)
|{z} |{z} |{z}
≥0 ≥0 =m2 h̄2

where we used that expectation values of squares of Hermitian operators are non-negative.
This gives us an inequality between the eigenvalues:3

p
m2 ≤ j( j + 1) or, taking the square root of these positive numbers, |m| ≤ j( j + 1).
(8.69)

This implies in particular that m has a maximum value mmax and by symmetry a mini-
mum mmin = −mmax .
This result is not surprising when we refer to the quasi-classical Figure 10, since the
projection of the angular momentum vector onto the z−axis can never be larger than the length
of the vector.

3 Remember that the eigenvalue µ of an operator O

b is just the expectation value of that operator when the
system is in the corresponding eigenstate:

Z Z
hOi
b = ψ∗ O
b ψ dx = µ ψ ∗ ψ dx for an eigenstate, Ôψ = µ ψ,
= µ for a normalised eigenstate.
8.6 Abstract angular momentum operators 121

z
6
2
J~ | jmi = h̄2 j( j + 1)| jmi
b

J~
6

 Jbz | jmi = h̄m | jmi









h̄ m
p

h̄ j( j + 1)





?


-y

x
2
J~ with
p
Figure 8.5: Picturing the inequality, |m| ≤ j( j + 1) for an eigenstate of Jbz and b
eigenvalues h̄m and h̄2 j( j + 1).

Step 2: Raising & lowering operators, Jb±

As we discovered in solving the SHO, the introduction of raising and lowering operators is
the key to the analysis. They are simple linear combinations of Jbx and Jby :

†
Jb± ≡ Jbx ± iJby with Jb+ = Jb− . (8.70)

Properties all follow from those of the J−operators:

b
2
1. Firstly, Jb commute with b
± J~ because Jb and Jb do: x y

2
J~ , Jb± ] = 0.
[b (8.71)

2
J~ with the same eigenvalue as | jmi.
This implies that Jb± | jmi are eigenstates of b

Proof:

2 2
J~ (Jb± | jmi) = Jb± (b
b J~ | jmi) = h̄2 j( j + 1)(Jb± | jmi) (8.72)
| {z }
=h̄2 j( j+1)| jm j i

2. More importantly
[Jbz , Jb+ ] = [Jbz , Jbx ] + i[Jbz , Jby ] (8.73)
= ih̄Jby + i(−ih̄)Jbx = h̄Jbx + ih̄Jby = +h̄Jb+ , (8.74)
122 Chapter 8. Angular Momentum in Q.M.

and similarly [Jbz , Jb− ] = −h̄Jb− .

In summary

[Jbz , Jb+ ] = +h̄Jb+ , [Jbz , Jb− ] = −h̄Jb− (8.75)

which are superimportant relations since they tell us that Jb± are raising/lowering
operators for angular momentum. We will come to this in a moment, but they are
very similar to ab† and ab – the raising/lowering operators for the SHO which obey
b ab† ] = +h̄ω ab† , [H,
[H, b ab] = −h̄ω ab . (8.76)

Recall that these relations imply that ab† and ab shift the energy eigenvalues by ±h̄ω.
3. Finally, with a little effort one can proof

[Jb+ , Jb− ] = +2h̄Jbz , (8.77)

but we will not use this relation.

Raising/lowering operators Jb± : rewriting the commutators in eqns. (8.75)

Jbz Jb± = Jb± Jbz ± h̄Jb± , (8.78)

and acting on AM eigenstates we find

Jbz (Jb± | jmi) = Jb± (Jbz | jmi) ± h̄Jb± | jmi from the above commutator,
= h̄m(Jb± | jmi) ± h̄Jb± | jmi since Jbz | jmi = h̄m| jmi. (8.79)

Collecting the (Jb± | jmi) terms on the right together gives the result:

Jbz (Jb± | jmi) = h̄(m ± 1) (Jb± | jmi) (8.80)

By now you know what is going on, Jb± are indeed raising/lowering operators, since they
increase/decrease the eigenvalue with respect to Jbz by h̄. Schematically:

Jb+ | jmi ∝ | jm + 1i and Jb− | jmi ∝ | jm − 1i (8.81)

or with correct normalisation factors:

p
Jb+ | jmi = h̄ j( j + 1) − m(m + 1)| jm + 1i
p (8.82)
Jb− | jmi = h̄ j( j + 1) − m(m − 1)| jm − 1i

Or even better pictorially...

8.6 Abstract angular momentum operators 123
2
Jbz Eigenvalue Eigenfunction J~ Eigenvalue
b
Jb+ | jmmax i = 0 h̄ mmax | jmmax i h̄2 j( j + 1)

h̄(m + 2) (Jb+ )2 | jmi h̄2 j( j + 1)

h̄(m + 1) Jb+ | jmi h̄2 j( j + 1)

6
Jb+
h̄ m | jmi h̄2 j( j + 1)
Jb−
? h̄(m − 1) Jb− | jmi h̄2 j( j + 1)

h̄(m − 2) (Jb− )2 | jmi h̄2 j( j + 1)

h̄ mmin | jmmin i h̄2 j( j + 1)

Jb− | jmmin i = 0
Figure 8.6: Illustrating the action of the raising and lowering operators on the eigenstates
2 2
of Jb and b
z J~ with eigenvalue h̄2 j( j + 1).
J~ . All are eigenstates of b

Step 3: Putting it all together: Since m is bounded j( j + 1) ≥ m2 (Step 1) the process of

lowering/increasing m must terminate in both directions!

Jb+ | jmmax i = 0 we cannot increase the eigenvalue even further (8.83)

Jb− | jmmin i = 0 we cannot dencrease the eigenvalue even further (8.84)

A first important conclusions from this is, since the states | jmmax i and | jmmin i are related
by an integer number of steps — everytime we act with Jb± to go up or down the runges we
increase or decrease m by 1 — we have

mmax − mmin = integer = 0, 1, 2, 3, . . . . (8.85)

Before using this result we need to find a relation between j and mmax by exploiting

Jb+ | jmmax i = 0 . (8.86)

124 Chapter 8. Angular Momentum in Q.M.

Now if this holds also the following must be true

0 = Jb− Jb+ | jmmax i

= (Jbx − iJby )(Jbx + iJby )| jmmax i
= (Jbx2 + Jby2 + i(Jbx Jby − Jby Jbx ))| jmmax i
2
J~ − Jbz2 + i [Jbx , Jby ])| jmmax i
= (b
| {z } (8.87)
ih̄Jbz
2
J~ − Jbz2 − h̄Jbz | jmmax i
= (b
= (h̄2 j( j + 1) − h̄2 m2max − h̄2 mmax )| jmmax i using known eigenvalues
⇒ j( j + 1) − mmax (mmax + 1) = 0

Hence we find, assuming j and mmax are positive

mmax = j and by symmetry mmin = − j (8.88)

and combining this with (8.85) we arrive at the key result!

mmax − mmin = 2 j = integer = 0, 1, 2, 3 . . . (8.89)

Importantly, this has two types of possible solutions!!! ⇒

Theorem 8.6.1 Quantisation Rules for General Angular Momentum:

A: j = 0, 1, 2, 3, . . . is an integer. This is the result we expect from the orbital angular

momentum where j = ` is indeed a non-negative integer.

B: j = 21 , 32 , 25 , . . . is a half-integer. This was an unexpected surprise...

For fixed j the allowed values of m for both cases are

m = mmin , mmin + 1, . . . , mmax − 1, mmax

= − j, − j + 1, . . . , j − 1, j (8.90)
| {z }
⇒ (2 j+1) possible values

 Example 8.2 We list here the allowed values of m (z-component of abstract AM) for the
8.6 Abstract angular momentum operators 125

first few values of j.

j=0:m=0
1 1 1
j = : m = − ,+
2 2 2 (8.91)
j = 1 : m = −1, 0, +1
3 3 1 1 3
j = : m = − ,− ,+ ,+
2 2 2 2 2


 Example 8.3 Remarkably, Nature makes use of both possibilities A. and B.!
A. j = 0, 1, 2, . . . INTEGER
1. Orbital Angular Momentum: b J~ = ~L
b
j = ` = 0, 1, 2, . . . and m = −`, −` + 1, . . . + `
| {z }
(2`+1) allowed values
bzY`m (θ , ϕ) = h̄mY`m (θ , ϕ)
L (8.92)
b2Y (θ , ϕ) = h̄2 `(` + 1)Y (θ , ϕ)
~L (8.93)
`m `m
2. Spin Angular Momentum: J~ = ~S (can only be realised using matrices)
b
j = 0: Scalar Particle ⇒ Higgs
j = 1: Photons, Gluons, W± and Z0 bosons are the force carriers for electromag-
netic, strong and weak interactions.
j = 2: The fundamental particle with the highest spin is the Graviton gµν .
B. j = 12 , 32 , . . . ⇒ Spin Angular Momentum b J~ = ~S, whose corresponding operators can
only be realised as matrices, NOT differential operators, for example Pauli matrices for
j = 12
j = 12 : electrons, quarks, neutrinos (all fundamental "matter" particles in nature), and
protons, neutrons,...
j = 32 : certain baryons (boundstates of 3 quarks), for example the ∆++ which is made
of three up-quarks.

126 Chapter 8. Angular Momentum in Q.M.

Summary: we have found, by purely algebraic means, the allowed eigenvalues (quantum
numbers) of the simultaneous eigenstates of the abstract angular momentum operators
2
J~ | j, m j i = h̄2 j( j + 1)| j, m j i
b (8.94)
Jbz | j, m j i = h̄m j | j, m j i (8.95)

The j and m j labelling an eigenstate are usually called the angular momentum quantum num-
bers. A special case of integral angular momentum is the orbital angular momentum b J~ = ~L;
b

we shall see later that spin angular momentum, b J~ = ~S, which cannot be pictured classically as
b
a microscopic spinning top, can have both integral and half-integral values.
In the figure below I have summarised the main result:

Jbz Eigenvalue Eigenfunctions Jbz Eigenvalue

m j h̄ | j, m j i | j, m j i m j h̄

+ jh̄ | j, ji | j, ji + jh̄

+2h̄ | j, 2i
| j, 3/2i + 32 h̄
| j, 1i
6
+h̄ Jb+
| j, 1/2i + 12 h̄
6
Jb+
0 | j, 0i
Jb− | j, −1/2i − 12 h̄
? −1h̄ | j, −1i Jb−
| j, −3/2i − 32 h̄ ?
−2h̄ | j, −2i

− jh̄ | j, − ji | j, − ji − jh̄

Angular Momentum j = 0, 1, 2, . . . Angular Momentum j = 12 , 32 , 52 , . . .

2
The angular momentum spectrum: All are eigenstates of b J~ with eigenvalue h̄2 j( j + 1).
On the left is shown the case of integral j, which allows m j = 0. Orbital angular momentum
always gives such a spectrum, with b J~ = ~L,
b j = ` and m = m = m; so does integral spin (eg.
j `
the photon has spin s=1) with J~ = ~S, j = s and m j = ms . On the right is shown half-integral
b b
angular momentum (eg. the electron and proton have spin s = 1/2), which forbids m j = 0.
 8.7 Using raising/lowering operators to find spherical harmonics 127

8.7 Using raising/lowering operators to find spherical harmonics

We have found in earlier that in central potentials V (r) because of complete spherical symmetry
the Hamiltonian commutes with all orbital angular momentum operator. Hence, energy
eigenstates are simultaneous eigenstates of three operators {H, b2 , L
b ~L b }: z

b n,`,m (r) = En ψn,`,m (r) ,

Hψ (8.96)
2
~L 2
n,`,m (r) = h̄ `(` + 1)ψn,`,m (r) ` = 0, 1, 2, . . . , (8.97)
b ψ
bz ψn,`,m (r) = h̄ mψn,`,m (r) m = −`, −` + 1, . . . , 0, . . . , ` − 1, ` ,
L (8.98)

and the energy eigenstates have this factorised form

ψn,`,m = Rn,` (r)Y`,m (θ , ϕ) . (8.99)

Importantly, the Y`,m (θ , ϕ) are universal and can be computed by a variety of methods.
Either using the old school approach by solving the 1st and 2nd order differential equations

b2Y (θ , ϕ) = h̄2 `(` + 1)Y (θ , ϕ) ` = 0, 1, 2, . . . ,

~L (8.100)
`,m `,m
bzY`,m (θ , ϕ) = h̄ mY`,m (θ , ϕ) m = −`, −` + 1, . . . , 0, . . . , ` − 1, ` ,
L (8.101)

directly.
We actually want to avoid this and use what we learned in the previous section about
raising/lowering operators Jb± to find an almost algebraic approach to find the Y`,m ’s. So
to begin let us recall the basic results from the previous section and write the operators in
spherical polar coordinates:
 
bx = ih̄ sin ϕ ∂ ∂
L + cot θ cos ϕ
∂θ ∂ϕ
 
by = −ih̄ cos ϕ ∂ ∂
L − cot θ sin ϕ (8.102)
∂θ ∂ϕ
bz = −ih̄ ∂
L
∂ϕ

so we can introduce raising/lowering orbital AM operators:

 
b+ = Lbx + iL
by = h̄ (cos ϕ + i sin ϕ) ∂ ∂
L + i cot θ (cos ϕ + i sin ϕ)
∂θ ∂ϕ
 
∂ ∂
= h̄eiϕ + i cot θ (8.103)
∂θ ∂ϕ
 
−iϕ ∂ ∂
L− = Lx − iLy = h̄e
b b b − + i cot θ
∂θ ∂ϕ
128 Chapter 8. Angular Momentum in Q.M.

For orbital angular momentum the (eigenstates) are Y`m (θ , ϕ) with the properties

b2 i = h̄2 `(` + 1)
h~L
hLbz i = h̄m
hL
b± i = hY`m |L
b± |Y i = 0 because of orthonormality
| {z`m} (8.104)
∝|Y`,m±1 i

= hL
bx i + ihL
by i
⇒ hL bx i = hL
by i = 0

Normalisation: with some effort one can work out the correct normalisation obtained by
acting with the raising/lowering operators L
b± . For example, consider

hL
b+Y`m |L
b+Y`m i = hY`m |L b+Y`m i
b− L
b2 − L
= hY`m |(~L bz2 − h̄L
bz )|Y`m i using eqn. (8.87) (8.105)
= h̄2 (`(` + 1) − m(m + 1))

Hence we find
p
b+Y`m = h̄ `(` + 1) − m(m + 1)Y`,m+1
L
p (8.106)
b−Y`m = h̄ `(` + 1) − m(m − 1)Y`,m−1
L

b± : first we recall that for fixed `, the z-component is bounded. Therefore,

Finding Y`,m using L
mmax = ` and mmin = −`. Using the first we know that the (highest) state Y`` has the following
properties
b+Y`` = 0 we cannot increase the eigenvalue any further since m = `
L
bzY`` = h̄`Y``
L (8.107)
if we also recall the eigenfunctions of L
bz derived earlier (see (8.38)), we know that

Y`` (θ , ϕ) = f (θ )ei`ϕ .

Notice the factorised form! If f would not be solely a function of θ then it would not be an
eigenfunction of Lbz anymore, but after all we want to find simultaneous eigenstates. Finally,
we impose
 
i`ϕ iϕ d f i`ϕ i`ϕ
0 = L+ f (θ )e = h̄e
b e + i cot(θ ) f i`e , (8.108)
dθ
giving the first order differential equation
df
− ` cot(θ ) f = 0 , (8.109)
dθ
8.7 Using raising/lowering operators to find spherical harmonics 129

which can easily be solved to give

f = N(sin θ )` . (8.110)

Now having found

Y`` = N(sin θ )` ei`ϕ = N(sin(θ )eiϕ )` ∝ (Y11 )` (8.111)

it becomes a mechanical exercise to find all other AM eigenfunctions by using the lowering
operator L
b− .
 Example 8.4 Let us go through this in one simple example, namely ` = 1. First note
r
3
Y11 = − sin(θ )eiϕ .
8π
Rπ R 2π 2
The sign is pure convention and Y11 was normalised as θ =0 ϕ=0 |Y11 | sin(θ )dθ dϕ = 1.
Next, using (8.106)

1
Y1,0 = p L
b−Y11
h̄ `(` + 1) − m(m − 1) `=m=1
r
∂ (sin(θ )eiϕ ) ∂ (sin(θ )eiϕ )
 
1 3 −iϕ
=− √ h̄e − + i cot θ
h̄ 2 8π ∂θ ∂ϕ (8.112)
r
3 −iϕ
− cos(θ )eiϕ + i cos(θ )ieiϕ


=− e
16π
r
3
= cos(θ )
4π
doing this again without showing the steps we find
r
1 b 3
Y1,−1 = √ L−Y10 = sin(θ )e−iϕ .
h̄ 2 8π

Similar steps for ` = 2 yield

2
Y2,±2 ∝ sin(θ )e±iϕ
Y2,±1 ∝ sin(θ ) cos(θ )e±iϕ (8.113)
2
Y2,0 ∝ 3 cos (θ ) − 1

 130 Chapter 8. Angular Momentum in Q.M.
1
8.8 Why ` = 2 does not work
We now want to illustrate why we could not discover half-integer quantisation angular momen-
tum using orbital angular momentum operator. So let us try to construct spherical harmonics
for the case ` = 12 where m = ± 12 .
Let us start with the ` = m = 12 :
1
Y 1 1 = (sin θ ) 2 eiϕ/2 which obeys L
b+Y 1 1 = 0 . (8.114)
22 22

So far so good, but you see this function is not periodic anymore

ϕ → ϕ + 2π : Y 1 1 → −Y 1 1 , (8.115)
22 22

instead under a full rotation the wave function comes back with a minus sign! But let us not
worry about this for now — later we will encounter this issue again for spin 12 particles. It
actually cannot be avoided.
We proceed by acting with the lowering operator (ignoring normalisations now)

Y 1 ,− 1 ∼ L
b−Y 1 1
2 2 22
1 1
!
∂ (sin(θ ) 2 eiϕ/2 ) ∂ (sin(θ ) 2 eiϕ/2 )
∼ h̄e−iϕ − + i cot θ
∂θ ∂ϕ
(8.116)
cos θ iϕ/2 cos θ √
 
−iϕ iϕ/2
∼e − √ e − sin θ e
2 sin θ 2 sin θ
cos θ −iϕ/2
∼√ e
sin θ
This is bad since this function diverges for θ = 0, π and even worse the process does not stop!
b−Y 1 1 6= 0. So this really does not work!
L
2 ,− 2
The way around, as you might suspect, is matrix quantum mechanics. For example for
spin j = s = 12 angular momentum (electrons, quarks, ...) the relevant operators b J~ = ~S = h̄2 ~σ
are matrices proportional to the Pauli matrices.
 
h̄ 0 1
Sx =
2 1 0
 
h̄ 0 −i
Sy = (8.117)
2 i 0
 
h̄ 1 0 h̄ h̄
Sz = ⇒ eigenvalues: + , − spin up/down in z-direction
2 0 −1 2 2
 8.9 More on particles in a central potential 131

8.9 More on particles in a central potential

We have found in earlier sections that if we have a central potential V (x, y, z) = V (r) then
we have complete rotational symmetry around any axis and, therefore, the Hamiltonian H b
commutes with all orbital angular momentum operators ~L. Furthermore, there is a complete
b
set of commuting operators with corresponding eigenvalues:

Operator H
b b2
~L L
bz
2
Eigenvalues E h̄ `(` + 1) h̄m
with ` = 0, 1, 2, . . . and m = −`, −`+, . . . , ` − 1, `. Or in full detail

b E,`,m (r, θ , ϕ) = EψE,`,m (r, θ , ϕ)

Hψ
b2 ψ
~L 2 (8.118)
E,`,m (r, θ , ϕ) = h̄ `(` + 1)ψE,`,m (r, θ , ϕ)
bz ψE,`,m (r, θ , ϕ) = h̄ mψE,`,m (r, θ , ϕ)
L

Notice that the energy eigenstates are not only labelled by the energy eigenvalues E, but
also the AM "quantum numbers" ` and m. Furthermore, the energy eigenstates have this
factorised form

ψE,`,m (r, θ , ϕ) = RE` (r)Y`m (θ , ϕ) (8.119)

where the spherical harmonics are universal – they do not depend on the shape of the potential
– and RE` (r) is usually called the radial wave function which does depend on the potential.

Normalisation: of course we need our energy eigenstates to be normalised.

Z ∞ Z π Z 2π
|ψE,`,m (r, θ , ϕ)|2 r2 sin(θ )drdθ dϕ = 1 (8.120)
r=0 θ =0 ϕ=0

and because of the factorised form of the eigenstates it is natural to demand

Z ∞ Z π Z 2π
2 2
|RE` (r)| r dr = 1 AND |Y`m |2 sin(θ )dθ dϕ = 1 (8.121)
r=0 | {z } θ =0 ϕ=0
Radial Probability

Expansion Theorem: using a natural extension of the expansion theorem we can write the
most general solution of the TDSE as:
∞ +`
Ψ(r, θ , ϕ,t) = ∑ ∑ ∑ cE,`,m ψE`m e−iEt/h̄ (8.122)
E l=0 m=−`

where we simply sum over all possible eigenvalues of all three operators.
132 Chapter 8. Angular Momentum in Q.M.

What is the interpretation of the coefficients? |cE,`,m |2 is the probability to find in a simultane-
ous measurement of energy, ~L b2 and Lb the values (E, h̄2 `(` + 1), h̄m).
z

How can we find the coefficients? As befits eigenstates of any Hermitian operator they are
orthonormal
Z ∞ Z π Z 2π
∗
ψE,`,m ψE 0 ,`0 ,m0 r2 sin(θ )drdθ dϕ = δE,E 0 δ`,`0 δm,m0 , (8.123)
r=0 θ =0 ϕ=0

which is simply saying that this integral is zero (the eigenstates are orthogonal) unless all
three quantum numbers/eigenvalues of the two eigenstates are equal! From this we infer
by analogy that
Z ∞ Z π Z 2π
∗
cE,`,m = ψE,`,m Ψ(r, θ , ϕ, 0)drdθ dϕ . (8.124)
r=0 θ =0 ϕ=0

Radial Equation:

h̄2 d 2 RE` (r) 2 dRE` (r) h̄2 `(` + 1)

 
− + + RE` (r) +V (r)RE` (r) = ERE` (r) (8.125)
2m dr2 r dr 2mr2

We show here the radial Schrödinger equation that we derived earlier – one important fact to
notice is that it does not depend on m. This implies that there at least 2` + 1 eigenstates for
each energy eigenvalue E because there are as many possible values for m. So the degeneracy
is at least 2` + 1 for particles in central potentials, but it can be higher as is the case of the 1/r
potential of atoms as we discuss in a moment.
 8.10 Properties of the spherical harmonics 133

8.10 Properties of the spherical harmonics

As we have already discussed the Y`m are simultaneous eigenstate
b2Y (θ , ϕ) = h̄2 `(` + 1)Y (θ , ϕ)
~L (8.126)
`,m `,m
LzY`,m (θ , ϕ) = h̄ mY`,m (θ , ϕ) ,
b (8.127)
but with the help of raising/lowering operators we found a method to find them without having
to solve the first of these equation which is 2nd order! We actually never checked that Y`m ’s
satisfy also the first differential equation – this is not needed if you have faith in our method,
but you are welcome to check in a few cases.
In the following I want to discuss some general properties of these functions. We will also
be sloppy with normalisation – at the end of the section there is a summary table with the
precise normalisation constants.
One nice exercise is to re-express the spherical harmonics in Cartesian coordinates using:

x = r sin(θ ) cos(ϕ)
y = r sin(θ ) sin(ϕ) (8.128)
z = r cos(θ )

which gives
` = 0 Y00 ∝ 1 1
z
` = 1 Y1,0 ∝ cos θ r
x±iy
Y1,±1 ∝ sin θ e±iϕ r
3z2 −r2 2z2 −x2 −y2
` = 2 Y2,0 ∝ 3 cos2 θ − 1 r2
= r2
z(x±iy)
Y2,±1 ∝ sin θ cos θ e±iϕ r2
(x±iy)2
Y2,±2 ∝ sin2 θ e±2iϕ r2

from which we can see the Y`m are polynomials of degree ` in x, y, z divided by r` . In particular
the Y1,m ’s are linear combinations of the components of the unit vector ~x/|~x| =~x/r.

Parity: P :~x → −~x: in this representation it is very easy to see how these function transform
under parity. Since they are polynomials of degree ` they pick a sign factor (−1)` and therefore:

P : Y`m → (−1)`Y`m (8.129)

Orthonormality: As expected for eigenfunctions of Hermitian operators the spherical har-

monics are orthonormal:
Z
Y`∗0 m0 (θ , ϕ)Y`m (θ , ϕ) dΩ = δ`0 ` δm0 m (8.130)
134 Chapter 8. Angular Momentum in Q.M.

where we use the solid angle notation as a convenient shorthand for the angular integration
variable, dΩ ≡ sin θ dθ dϕ, with the accompanying shorthand for the integration itself:
Z Z 2π Z π
dΩ ≡ dϕ sin θ dθ (8.131)
0 0

Z
|Y`m (θ , ϕ)|2 dΩ = 1, (8.132)

Finally note that the expansion theorem allows us to write the most
p general wave function for
a particle whose magnitude of orbital angular momentum is h̄ `(` + 1), ie. orbital angular
momentum quantum number `, as a linear combination of eigenstates with definite projections
along the z-axis:
+` +`
Y(`) (θ , ϕ) = ∑ cm Y`m (θ , ϕ), where ∑ |cm |2 = 1, (8.133)
m=−` m=−`

For a system in this normalised state the outcome of a measurement of L b 2 will certainly
yield h̄2 `(` + 1); but a measurement of L bz will yield one of its eigenvalues, h̄m, m ≡ m` =
−`, −` + 1, . . . , ` − 1, `, with probability |cm |2 ; if the outcome is h̄m then immediately after
the measurement the wave function will be Y`m (θ , ϕ). p
The angular momentum vector, in units of h̄, is a vector of length `(` + 1); the only
permissible orientations of the vector being those for which its projection onto the z−axis
have the values m = −`, −` + 1 . . . , `.
Thus for ` = 1 there are 3 allowed m−values, m = −1, 0, 1; for ` = 2 there are 5 allowed
values, m = −2, −1, 0, 1, 2, and so on.
 8.10 Properties of the spherical harmonics 135

8.10.1 Explicit form of spherical harmonics up to ` = 3

Quantum Associated Legendre Azimuthal

Numbers Function Function Spherical Harmonics
Parity = (−1)`
|m| |m|
` m P` (cos θ ) eimϕ Y`m (θ , ϕ) = N`m P` eimϕ

1

12
0 0 1 1 Y0,0 = 4π

3

12
1 0 cos θ 1 Y1,0 = 4π cos θ

12
±1 sin θ e±iϕ Y1,±1 = ∓ 3
8π sin θ e±iϕ

1 2 5

12
2 0 2 (3 cos θ − 1) 1 Y2,0 = 16π (3 cos2 θ − 1)

12
±1 3 sin θ cos θ e±iϕ Y2,±1 = ∓ 15
8π sin θ cos θ e±iϕ

12
±2 3 sin2 θ e±2iϕ Y2,±2 = 15
32π sin2 θ e±2iϕ

1 3 7

12
3 0 2 (5 cos θ − 3 cos θ ) 1 Y3,0 = 16π (5 cos3 θ − 3 cos θ )

21
±1 3 2
2 sin θ (5 cos θ − 1) e±iϕ Y3,±1 = ∓ 21
64π sin θ (5 cos2 θ − 1) e±iϕ

12
±2 15 sin2 θ cos θ e±2iϕ Y3,±2 = 105
32π sin2 θ cos θ e±2iϕ

21
±3 15 sin3 θ e±3iϕ Y3,±3 = ∓ 35
64π sin3 θ e±3iϕ

Z Z Z 2π Z π
Y`∗0 m0 (θ , ϕ)Y`m (θ , ϕ) dΩ = δ`0 ` δm0 m with dΩ ≡ dϕ sin θ dθ
0 0
 136 Chapter 8. Angular Momentum in Q.M.

8.11 More on (Hydrogen) Atoms

In this case the central potential has the form

Ze2
V (r) = − , (8.134)
4πε0 r
where Z is the number of protons in the nucleus – for the hydrogen atom Z = 1.

Energy eigenstates: are labelled by three quantum numbers (you need to remember their
ranges!)

ψn`m = Rn` (r)Y`m (θ , ϕ)

n = 1, 2, 3, . . . Principal qu. nr.:
(8.135)
` = 0, 1, 2, . . . n − 1
m = −`, −` + 1, . . . , ` − 1, `

Energy eigenvalues and Bohr radius

E1 me e4 Z 2 h̄2
En = with E 1 = − = − ∼ −13.6eV
n2 2(4πε0 h̄)2 2me a2
(8.136)
4πε0 h̄2
a= ∼ 5.29 × 10−11 m Bohr radius, "size of atoms"
me e2 Z
where me denotes the mass of the electron, to distinguish it from the AM quantum number m.
Note that the Bohr radius, a, is about 104 times larger than the nucleus! This also implies that
most of matter is made of empty space between the nucleus and the electrons.

Degeneracy: this is a special case because of the 1/r potential and the En do not depend on
the AM quantum number `. As a result, for each n there are n2 distinct eigenstates with
the same energy En !
The deeper reason is that this system has not only complete rotational symmetry (orbital
Angular momentum is conserved), but there is another conserved vector quantity – the so-
called Runge-Lenz vector:

~ = ~p ×~L +V (r)me~x
K (8.137)
8.11 More on (Hydrogen) Atoms 137

Examples: plots of radial wave functions Rm` (r) are on the last page of this section

2 −r/a
• n = 1, ` = m = 0, 1 unique groundstate wavefunction R10 = a3/2 e
Degeneracy = 1, h~Li = h~pi = 0 (unlike the original Bohr model!)
• n=2
– ` = 0, m = 0, 1 State, R20 = √2a1 3/2 (1 − r/(2a))e−r/(2a)
– ` = 1, m = −1, 0, +1, 3 States, R21 = √ 1
24a5/2
re−r/(2a)
Degeneracy = 1 + 3 = 4
• n=3  
2 2r 2r2 −r/(3a)
– ` = 0, m = 0, 1 State, R30 = √27a 3/2 1 − 3a + 27a2 e

3 States, R31 = 27√86a3/2 1 − 6ar r −r/(3a)

– ` = 1, m = 0, ±1, a e
4 r 2 −r/(3a)


– ` = 2, m = 0, ±1, ±2, 5 States, R32 = 81√30a 3/2 a e
Degeneracy = 1 + 3 + 5 = 9

Weakly bound electrons (Rydberg atom)

These are encountered if an electron in an atom is highly excited and its principal quantum
number n is very large. Then the binding energy of the electron E1 /n2 obviously becomes very
small. These excited states also have an unusually large lifetime, and the electron follows an
almost classical circular trajectory with a radius much larger than the Bohr radius (see below).
In this case the maximum value for AM is ` = n − 1 in which case the radial wave function
is very simple
 n+1/2
2 1 r
Rn,n−1 = p rn−1 e− na (8.138)
na (2n)!
and it turns out that the radial probability r2 |Rn,n−1 |2 has a maximum at r = n2 a. For n = 100
this is 10000 times the Bohr radius! In this case the "size" of the atom is of the order of µm!

Expansion Theorem: a general solution of the TDSE has the form

∞ n−1 +`
Ψ(r, θ , ϕ,t) = ∑∑ ∑ cn`m ψn`m (r, θ , ϕ)e−iEnt/h̄ . (8.139)
n=1 `=0 m=−`
Note that the ranges of the sums over the quantum numbers follow exactly the ranges men-
tioned earlier (8.135).
Once a general state is expanded in terms of energy eigenstates it is very simple (without
b2 and L
integration) to obtain the expectation values of energy, ~L b . This follows from the in-
z
terpretation of the cn`m which is a generalisation of the interpretation of cn found in 1D systems:

b2 and L
|cn`m |2 is the probability to find in a simultaneous measurement of energy, ~L bz
2
(remember they commute with each other) the values En , h̄ `(` + 1), h̄m.
138 Chapter 8. Angular Momentum in Q.M.

 Example 8.5 Assume at t = 0 the state of an electron in the hydrogen atom is

1 1 1
Ψt=0 = √ ψ1,0,0 + ψ2,0,0 + ψ2,1,1 (8.140)
2 2 2

Tasks:
1) Check that the wave function is correctly normalised,
2) Find the wave function at later times,
b h~Lb2 i, hL
3) Find the expectation values hHi, bz i

Answers:

1) ∑n,`,m |cn`m |2 = | √12 |2 + | 12 |2 + | 21 |2 = 21 + 14 + 14 = 1. OK.

2) Ψt>0 = √1 ψ1,0,0 e−iE1 t/h̄ + 1 ψ2,0,0 e−iE2 t/h̄ + 1 ψ2,1,1 e−iE2 t/h̄
2 2 2

3) Expectation values:

b = ∑n,`,m |cn`m |2 En = | √1 |2 E1 + | 1 |2 E2 + | 1 |2 E2 = E1 +E2

hHi 2 2 2 2

b2 i = h̄2
h~L ∑n,`,m |cn`m |2 h̄2 `(` + 1) = | √12 |2 × 0 + | 12 |2 × 0 + | 12 |2 × h̄2 1(1 + 1) = 2

bz i = ∑n,`,m |cn`m |2 h̄m = | √1 |2 × 0 + | 1 |2 × 0 + | 1 |2 × 1h̄ =

hL h̄
2 2 2 4


8.11 More on (Hydrogen) Atoms 139

The radial wave functions Rn` for n = 1, 2, 3 and all possible values of `:
2
R10 = e−r/a
a3/2
1  r  −r/(2a)
R20 = √ 1− e
2a3/2 2a
1 r
R21 = √ e−r/(2a)
24a 3/2 a
(8.141)
2r2
 
2 2r
R30 = √ 1− + 2
e−r/(3a)
27a 3/2 3a 27a
8  r   r  −r/(3a)
R31 = √ 1− e
27 6a3/2 6a a
4  r 2
R32 = √ e−r/(3a)
81 30a 3/2 a

Figure 8.7: Plots of the |Rn` (r)|2 and the radial probabilities r2 |Rn` (r)|2 for n = 1, 2
 9. Spin Angular Momentum

One of the great surprises in the history of 20th century physics was the discovery of spin, a
form of angular momentum intrinsic to an isolated particle such as an electron. Even famous
scientists like Wolfgang Pauli – who eventually got a Nobel Prize for the description of
spin – were initially strongly opposed to this crazy idea. The surprise is that spin cannot be
understood quantitatively as the analogue of a spinning top or the earth spinning on its axis
- these are really examples of orbital angular momentum in disguise: the material making
up a top or the earth is simply orbiting about the spin axis. In fact the electron is known
experimentally to be so small that it’s impossible to imagine how it could produce its angular
momentum h̄/2 by spinning no faster than the speed of light.
Before developing quantum mechanics of spin, let us look at experimental evidence for
quantisation of orbital angular momentum and the existence of spin.

9.1 Periodic table and the missing quantum number

In previous chapters we discovered that electrons moving in the Coulomb potential produced
by an atomic nucleus are described by energy eigenstates labelled by three quantum numbers
n, `, m:

ψn`m = Rn` (r)Y`m (θ , φ )

n = 1, 2, 3, 4, . . .
(9.1)
` = 0, 1, 2, . . . , n − 1
m = −`, −` + 1, . . . , ` − 1, `
142 Chapter 9. Spin Angular Momentum

However, it was understood that this could not explain the electron configurations and
chemical properties of atoms. In particular, all Z electrons in an atom with Z protons in the
nucleus could be in the groundstate and such atoms would be chemically inert and no bonding
to produce molecules would occur.
Therefore, in 1924/1925 Pauli postulated that each electron state is labelled by an ad-
ditional quantum number, which can take two possible values, and formulated the Pauli
exclusion principle in order to explain correctly the observed electron configurations of atoms.
Later this additional quantum number was identified with the z-component of the spin angular
momentum of electrons – more precisely the spin operator (matrix) Sz
 
h̄ 1 0
Sz =
2 0 −1

has eigenvalues ms h̄ with ms = ± 12 . (Initially Pauli strongly opposed the idea of spin.) The
Pauli exclusion principle demands that electrons must differ in at least one of the four quantum
numbers n, `, m, ms . Shortly after in 1925 electron spin was discovered by Goudsmit and
Uhlenbeck, but it took Pauli until 1940 to prove his exclusion principle (now also called
spin-statistics theorem).
The exclusion principle applies actually to all half-integer spin particles (which are also
called fermions), while integer spin particles such as photons and the Higgs particle do not
follow this principle (they are called bosons). This is very important in statistical physics and
is responsible for the existence of white dwarfs and neutron stars.
What does this imply for the electron structure of atoms?
This goes under the name Aufbau principle: if we start with a nucleus that contains Z protons
we add electrons one by one systematically filling the allowed states starting with the lowest
energy states systematically going through all possible values of n, `, m, ms . Up to n = 3 this
goes as follows:

Principle qu. nr. AM qu. numbers spin number of electron states

n=1 ` = 0, m = 0 ms = ± 12 2
n=2 ` = 0, m = 0 ms = ± 12 2
` = 1, m = −1, 0, +1 ms = ± 12 6
n=3 ` = 0, m = 0 ms = ± 12 2
` = 1, m = 0, ±1 ms = ± 12 6
` = 2, m = 0, ±1, ±2 ms = ± 12 10
If all electron states for a fixed n are occupied we talk about a closed shell. Such a shell
contains 2n2 electrons (check the numbers in the table!). Such closed shells correspond to
particularly stable and chemically inert electron configurations:
Nobel gases: their "inertia" is explained by the fact that all electrons sit in shells that are
completely filled; e.g. Helium (He) has Z = 2(n = 1), while Neon (Ne) has Z = 10 = 2(n =
1) + 8(n = 2).
Alkali metals on the other hand are extremely reactive. This is because they have the
same electron configurations as Nobel gases plus one extra valence electron that has to
9.1 Periodic table and the missing quantum number 143

occupy a new shell all by itself, because all other "seats" are taken, e.g. Sodium (Na) has
Z = 11 = 2(n = 1) + 8(n = 2) + 1(n = 3).
This valence electrons played a special role in the discovery of spin, since the spin and
orbital angular momenta of all electrons in a closed shell add up to zero, and the total angular
momentum of the atom comes from the valence electron alone. This allows us to learn about
the spin property of a single electron although we look at an atom with many more electrons!
 144 Chapter 9. Spin Angular Momentum

9.2 Atoms in a magnetic field – Zeeman effect

A classical rotating charge produces a magnetic dipole moment ~µ:

Figure 9.1: A current due to an electron moving on a circle produces a magnetic moment!

(−e)v
I= and A = πr2 (9.2)
2πr
(−e)vr (−e) (−e) ~
|~µ| = IA = = × me vr = |L| (9.3)
2 2me |{z} 2me
angular momentum

where −e is the charge of the electron and me its mass.

Putting back the vectors we get for the magnetic momentum induced by the orbital angular
momentum of an electron
(−e)~ µB
~µ = L ≡ − ~L , (9.4)
2me h̄
where we have introduced the common Bohr magneton
eh̄
µB ≡ , (9.5)
2me
which is useful (in hindsight) since angular momentum is quantised in multiples of h̄.
Now I should stress it would be completely impossible to measure the minute angular
momentum of an electron or other microscopic particle. But since the angular momentum is
(linearly) proportional to the magnetic moment we can measure it indirectly because the ~µ
reacts to external magnetic fields ~B. From an earlier course on electromagnetism you know
that the magnetic dipole experiences a torque that tries to align ~µ with ~B and there is an
associated potential energy
µB ~ ~
Emag = −~B · ~µ = + B·L . (9.6)
h̄
9.2 Atoms in a magnetic field – Zeeman effect 145

Note the minus sign in the 2nd expression is due to the fact that ~µ aligned parallel to ~B is
energetically preferred. If the magnetic field is constant (homogeneous) there is no net force
on the magnetic dipole moment, but if magnetic field is not constant (inhomogenous) the
electron will also feel a force – we will come to that in the next section.
It is now natural to promote this to an additional term in the quantum mechanical Hamilto-
nian – this is not a proof, since we have used classical intuition, but it works.

bmag = + µB ~B ·~L
H b (9.7)
h̄

For now we will focus on a constant magnetic field ~B = (0, 0, Bz ). So we get an additional
piece to the Hamiltonian

bmag = µB Bz L
H bz (9.8)
h̄
Now if the electron is in an angular momentum eigenstate, then the Lbz eigenvalue = h̄m
with m an integer. Here we also see why people introduced the m for this quantum number,
because they called it the magnetic quantum number. So we see that the energy of the electron
gets shifted
µB b µB eh̄Bz
hH
bmag i = Bz hLz i = Bz h̄m = µB Bz m = m (9.9)
h̄ h̄ 2me
and this shift follows discreet steps if the magnetic momentum and, hence, angular momentum
is quantised!

m = −`, −` + 1, . . . 0 . . . , ` − 1, ` . (9.10)

This is called the Zeeman Effect which leads to a change of the line spectrum and requires
electrons in an excited state with n > 1 and ` ≥ 1.
Comments:
1. In general this leads to the lifting of the degeneracy of the energy eigenstates of the atom
– there are 2n2 eigenstates for the energy eigenvalue En .
2. In general we expect 2` + 1 different energy shifts if ~B 6= 0. This is the number of
different values of m, and note this is an odd number! In the Figure 9.2 with ` = 1 we
found three energy levels.
3. However, in practise in experiments with alkali metals also even numbers of lines and,
hence, energy levels were found. This included cases where despite ` = 0 (zero orbital
angular momentum) two lines were found.
They only way out is to allow half-integer spin since 2 j + 1 = 2 requires j = 12 . This
gave direct evidence for half-integer angular momentum and spin!
4. This doubling of line spectra was observed for alkali metals like Sodium (Na, Z = 11 =
2 + 8 + 1) and Lithium (Li, Z = 3 = 2 + 1), which both have a single valence electron
with ` = 0. Hence, the Zeeman effect had to come from the intrinsic spin angular
momentum and intrinsic magnetic moment of the valence electrons.
146 Chapter 9. Spin Angular Momentum

5. The three lines illustrated in Figure 9.2 correspond to the normal Zeeman effect. Spin
and other aspects of the magnetic interactions in an atom lead to complicated level
splitting patterns known as the anomalous Zeeman effect. The observation of the Zeeman
effect was an important milestone in atomic physics because it provided evidence for the
quantization of angular momentum and spin, and for the electron being in the atom and
directly involved in the production of spectral lines. The Zeeman effect is also widely
used in astronomy to measure large magnetic fields in stars.

Figure 9.2: Illustratation of the Zeeman effect in the case of transitions between n = 2, ` = 1
and n = 1: A non-zero magnetic field "lifts" the degeneracy of the E2 energy eigenvalue, and
instead of one sharp spectral line three sharp lines are seen. In classical physics we would have
expected a broadening or smearing out of the sharp line because m could be any real number.
 9.3 Stern Gerlach experiment 147

9.3 Stern Gerlach experiment

It was Otto Stern in 1921 who proposed to use the effect of an inhomogeneous magnetic
field on a magnetic dipole to measure the magnetic moments of atoms. If ~B = ~B(x, y, z) then,
in addition to a torque the field will exert a net force on the dipole because the little magnet’s
north and south poles are in slightly different fields. In our initial considerations we find it most
convenient to choose the z−axis to lie along the direction of the magnetic field, ~B = (0, 0, Bz ).
As usual, the force is obtained from the gradient of the potential:
~F = −~∇Emag (9.11)
= ~∇(~µ · ~B) (9.12)
= ~∇(µz Bz ) with ~B pointing in the z−direction. (9.13)
 
∂ Bz ∂ Bz ∂ Bz
= µz , , (9.14)
∂x ∂y ∂z
In the Stern-Gerlach experiment the magnet poles are shaped to produce a magnetic field that
varies rapidly in the z−direction only, Bz = Bz (z), 1 so the force also acts in the z−direction:

~F = (0, 0, µz ∂ Bz )
∂z
µB ∂ Bz
= (0, 0, − Lz ) classical orbiting e− (9.15)
h̄ ∂z
∂ Bz
= (0, 0, −µB m` ) quantum mechanical orbiting e− (9.16)
∂z
where, in the last step, we have introduced the QM quantization rule for the z−component
of orbital angular momentum. The 1922 Stern-Gerlach experiment was performed with a
beam of neutral silver atoms, the inhomogeneous magnetic field being produced by one of the
poles having a razor-sharp pointed face. The details are shown in Figure 9.4 which also show
the original results. Since the silver atoms’ magnetic moments are randomly oriented in the
incident beam, the classical expectation was that there be a continuous range of deflections,
reflecting the continuous distribution of Lz values in the beam. The QM prediction is that a
discrete set of deflections should result from the quantization of orbital angular momentum
with 2` + 1 possibilities corresponding to m` = −`, −` + 1, . . . 0, . . . `, including zero deflection
for m` = 0. In fact no zero deflection was seen and only 2 deflected beams, one above and
one below the expected zero deflection, inconsistent with integral values of ` but suggesting
instead the value 1/2 to give 2` + 1 = 2.
This experiment and many other hints from atomic spectra and the Zeeman effect led to
the Goudschmidt-Uhlenbeck suggestion in 1925 that in fact the electron has an intrinsic spin
angular momentum h̄/2. The Stern-Gerlach experiment on Ag and later ones on Cu, Au, Na,
1 We assume here that |Bx |, |By | << |Bz | so the angular momentum is quantised in the z-direction and the
force is predominantly in the same direction. In the actual experimental set-up there are small deviations from
this idealised picture. Also note that because ~∇ · ~B = 0 also the other components must be non-vanishing but it
can be shown that this has no significant effect on the outcome.
148 Chapter 9. Spin Angular Momentum

K, Cs and H finds its explanation in the fact that all these atoms have closed electron shells
plus one extra valence electron. The closed shell contributes a net zero angular momentum and
the valence electron has zero orbital angular momentum, ` = 0, but 1/2 unit of spin angular
momentum. The atom as a whole therefore has a total angular momentum h̄/2 due solely to
the spin of the valence electron, and it is its accompanying spin magnetic moment which is
being measured in the Stern-Gerlach experiment. Thus the experiment provided evidence both
for the existence of electron spin and for the azimuthal quantization of angular momentum.
However there was at least one more surprise: the electron also has twice the magnetic moment
expected from the analogy with orbital motion. This experimental result, that the electron has
intrinsic spin angular momentum which produces a magnetic moment even if ` = m = 0,
is expressed in terms of the g-factor, gs , defined by:

µB ~ µB
~µs = −gs S ' −2 ~S for an e− (9.17)
h̄ h̄

Comments:
• The modern value for the electron is gs = 2.002381304386 ± 0.000000000020 corre-
sponding to a spin magnetic moment of very nearly 1 Bohr magneton: the factor of 1/2
from the spin S is almost exactly cancelled by the 2 from the g-factor. This result has
no classical explanation, but the value gs = 2 is a consequence of Dirac’s relativistic
equation for the electron; the small correction is actually in agreement with modifica-
tions calculated from Quantum Electrodynamics, the relativistic extension of quantum
mechanics to the interaction of photons and electrons.
• Other particles in nature have been found with spin-1/2, including the proton and the
neutron. Because their masses are some 1800 times larger than the electron the nuclear
magneton is smaller by that factor:2
eh̄
1 Nuclear magneton ≡ µN = = 5.446 × 10−4 µB = 5.05 × 10−27 J T−1 (9.18)
2m p

with g p = 5.586 and gn = −3.826. The reason these particles have g-factors so different
from the canonical value of 2 given by the Dirac equation is because the strong interac-
tions play a powerful role in giving these particles a finite size and a complicated and
measurable structure - they are surrounded by clouds of so called virtual particles such
a pi mesons.
For a spinning particle such as an electron in an atom which also has orbital angular
momentum, the magnetic moment is the vector sum (using the approximation gs ' 2):

~µtot = ~µL + ~µs

µB µB (9.19)
= − (~L + gs~S) ' − (~L + 2~S) for an e− .
h̄ h̄
2 Beware signs: we have used an explicit minus sign up to this point for the electron’s magnetic moment, so
it’s gs is positive. To apply the above equations to the proton we have to remove this minus sign in appropriate
places. This doesn’t cause any confusion in practice.
9.3 Stern Gerlach experiment 149

This leads to an extra term in the Hamiltonian of the form

bmag = −~B · ~µtot = µB ~B · (~L

H b + 2~S) for an e− . (9.20)
h̄

We now come back to the actual set-up of the Stern-Gerlach experiment, in which a
collimated beam of neutral atoms was sent in the y-direction through an inhomogeneous
magnetic field (in the z-direction) as described above and finally landed on a screen to detect
where the atoms had gone.

In the original experiment by Otto Stern (he had the idea) and Walther Gerlach (he built
the experiment) neutral silver atoms (Ag) were used, but later this experiment was repeated
with alkali metals such as sodium (Na) and even hydrogen. All these atoms have in common
that they have one valence electron with ` = m = 0, so that the whole angular momentum
and hence magnetic moment of the atom comes from this single electron, because all other
electrons reside in closed shells.

So anticipating that the electron has spin 1/2 and the eigenvalues of the spin operator Sz
has eigenvalues h̄ms with two allowed values ms = ± 12 we can compute the force that these
atoms should experience in the experiment if our conjecture is correct:

∂ ∂
Fz = − Emag = (µz Bz )
∂z ∂z
∂ µB
= (−2 Sz Bz ) replacing the spin operator by the eigenvalues h̄ms (9.21)
∂z h̄ (
∂ Bz −µB ∂∂Bzz for ms = + 12 spin up
= −2µB ms =
∂z +µB ∂∂Bzz for ms = − 12 spin down

Since there are two discrete choices here (spin up or spin down) which lead to a force in
opposite directions we expect two dots on the screen while a classical physicist would expect
a smeared out deposit of atoms on the screen because ms would be allowed a continuous range
of values. Here is an illustration of the experimental set-up and outcome, and a picture of the
actual result.
150 Chapter 9. Spin Angular Momentum

Figure 9.3: Illustration of the Stern-Gerlach experiment: inhomogeneous magnetic field in

z-direction, atoms moving in y-direction.

Figure 9.4: Postcard sent by Stern and Gerlach to Bohr showing their result. Left without,
right with magnetic field. They wrongly congratulate Bohr to the confirmation of his model
and quantisation of angular momentum. They should have expected an odd number of dots
because of integer quantisation of orbital angular momentum. The correct interpretation –
confirmation of spin 1/2 of the electron came only a few years later!
 9.3 Stern Gerlach experiment 151

Comments:
• It is essential to use a neutral beam in this experiment otherwise the relatively weak
deflections due to the magnetic dipole interaction would be completely masked by the
large deflections caused by a charged beam passing through a strong magnetic field
(Lorentz force).
• Measurement Theorem: the Stern-Gerlach (SG) experiment measures the z-component
of the spin angular momentum of the atoms (actually the magnetic moment that is re-
lated to it) if the magnetic field points in the z-direction. The possible outcomes are the
two possible eigenvalues – in this case ±h̄/2. But notice we have to make an actual
measurement by putting the glass plates to see this – otherwise we do not know where
the atoms are!
• State Preparation: we can use the SG experiment to prepare two spatially beams
(ensembles) of atoms in specific spin states; in this case spin up Sz = +h̄/2 or spin down
Sz = −h̄/2 with respect to the z-axis. Because they experience different forces the two
beams are spatially separated in the z-direction!
• Later we will consider sequential combinations of SG experiments measuring spin
in different directions (e.g. we can have magnetic field pointing in the x-directions
measuring the x-componet Sx of the spin, which allows for interesting tests of the
measurement theorem, collapse of wave function and hidden variables.

9.3.1 Why Spin cannot be understood classically

Several aspects of spin angular momentum defy our attempts to understand it as the analogue
of a top spinning about its axis - despite the fact that one’s mental image of spin is precisely
this classical system! The key to the difficulty lies in the experimental fact, gleaned from
many high energy scattering experiments on electron and positron beams, that the electron is
as close to Newton’s idealised point particle as one could imagine:

Electron scattering experiments are consistent with: me = 0.51 MeV and re < 10−20 m.

Taking the electron as a sphere of mass me and radius re spinning about its centre with
average speed v, and ignoring factors of order 1, the angular momentum is,

h̄ h̄ h̄
J ≈ re me v ≈ , implying, re ≈ ≥ ≈ 2 × 10−13 m,
2 2me v 2me c
which is seven orders of magnitude greater than the experimental limit. To put this into the
context of 1921, compare this with the sizes of atoms (around 10−10 m) and of the proton
(10−15 m); yet at the time it was already known that the electron was very much smaller than
either atom or nucleus. We can put this another way: the speed of rotation needed to explain
the magnitude of the spin angular momentum for a particle as small as an electron exceeds the
speed of light by seven decades:

h̄
v≈ > 2 × 107 c.
2me re
152 Chapter 9. Spin Angular Momentum

Thus, like many quantum phenomena, spin has a classical analogue, but the analogue fails to
provide a quantitative explanation; it only gives us a convenient mental picture. Electron spin
appears to be an entirely quantum mechanical phenomenon. This statement is not, however,
quite correct: in 1926 Dirac presented his famous relativistic wave equation for the electron
- the Dirac equation - which not only gives the correct spin angular momentum, but also an
excellent approximation to its measured magnetic moment and a prediction that the electron
has an antiparticle, now called the positron. Thus relativity in concert with QM plays an
essential role in the explanation of spin.
Having concluded that we cannot picture spin angular momentum simply as matter rotating
in some sort of orbit around the electron’s centre whose shape we could hope to determine
- perhaps by some sort of generalisation of a spherical harmonic - we now face a dilemma:
what are the coordinates that play the role of (x, y, z) or (r, θ , ϕ) in describing orbital angular
momentum? Since we do not know we must resort to an abstract approach via the angular
momentum operators for spin. This has already been accomplished by our discovery that the
commutation rules for the angular momentum operators b J~ provide for the possibility of spin
angular momentum h̄/2. We now continue along that path by following Heisenberg’s matrix
approach to QM in which no reference need be made to the coordinates.
 9.4 Matrix mechanics and spin 153

9.4 Matrix mechanics and spin

In his matrix version of QM Heisenberg achieved a form of the theory which made reference
only to directly measurable physical observables, the expectation values. The formalism is
ideally suited to the quantum mechanical study of angular momentum, especially spin. The
most appropriate notation to use is Dirac’s bra/ket formulation.
An important property of intrinsic, spin j = s angular momentum (AM) of fundamental
particles that it is a fixed, unmovable quantitiy unlike orbital angular, which depends on the
p Hence, we have focus on states with fixed j = s or length-squared of
specific spatial motion.
the spin AM |J|~ = h̄ j( j + 1).
We seem to be hopping between j and s notation; recall b J~ are the abstract AM operator
which we will use to develop the QM spin and derive the spin AM operators. The spin AM
operators will eventually be called Sx , Sy , Sz , and j, m will be relabelled as s, ms .

Spin states:
Since j = s is fixed a general spin state will be a linear combination of (2 j + 1) states
(expansion theorem)

j j
|χi = ∑ cm |ψ j,m i or χ = ∑ cm ψ j,m . (9.22)
m=− j m=− j

where the expansion coefficients are interpretated as follows: |cm |2 is the probability to find
the eigenvalue h̄m in the measurement of Jbz . For example imagine sending particles in this
spin state through a Stern-Gerlach apparatus with the magnetic field in the z-direction (recall
that the force on the particles is proportional to −2µB m∂ Bz /∂ z:

Figure 9.5: Illustration of the Stern-Gerlach experiment with a spin j particles all in the
same general state χ. |cm |2 are the probabilities to find the Jz eigenvalue h̄m. We find
(2 j + 1)separate beams since different forces act on particles with different h̄m eigenvalue.

Converting to matrix form we can write the state χ as a (2 j + 1)-component column

154 Chapter 9. Spin Angular Momentum

vector:
         
cj 1 0 0 0
 c j−1  0 1 0 0
         
χ =  .  = c j  .  + c j−1  .  + · · · + c− j+1  .  + c− j  ...  .
 ..   ..   ..   .. 
(9.23)
 
         
c− j+1  0 0 1 0
c− j 0 0 0 1
 9.4 Matrix mechanics and spin 155

9.4.1 Working out the matrix form of Sx , Sy , Sz

The matrix form of the spin operators, Sx , Sy and Sz , follows from the known properties of
the abstract AM operators (this is really all we need to derive them):

[Jbx , Jby ] = ih̄Jbz , plus cyclic permutations

2
J~ | j, mi = h̄2 j( j + 1)| j, mi
b
(9.24)
Jbz | j, mi = h̄m| j, mi
p
Jb± | j, mi = h̄ j( j + 1) − m(m ± 1)| j, m ± 1i

Let us begin with the two simplest cases, Sz and ~S2 :

Sz = (Jz )m,m0 = h j, m| Jbz | j, m0 i = h̄mh j, m| j, m0 i = h̄mδm,m0

| {z }
=h̄m| j,m0 i
   
(Jz ) j, j (Jz ) j, j−1 . . . (Jz ) j,− j j 0 ... 0
 (Jz ) j−1, j (Jz ) j−1, j−1 . . . (Jz ) j−1,− j   0 j−1 ... 0  (9.25)
=  = h̄ 
   
.. .. .. .. .. .. .. .. 
 . . . .   . . . . 
(Jz )− j, j (Jz )− j, j−1 . . . (Jz )− j,− j 0 0 ... − j

  2
~S2 = J~2 = h j, m| J~ | j, m0 i
b = h̄2 j( j + 1)h j, m| j, m0 i = h̄2 j( j + 1)δm,m0
m,m0 | {z }
=h̄2 j( j+1)| j,m0 i
 
1 0 ... 0 (9.26)
 0 1 ... 0 
= h̄2 j( j + 1) 
 
.... . . .. 
 . . . . 
0 0 ... 1

How can we find Sx and Sy ? Simply use the raising and lowering operators Jb+ = Jbx + iJby
and Jb− = Jbx − iJby and find their matrix form. We then finally use that Sx = (S+ + S− )/2 and
Sy = (S+ − S− )/(2i).

Jb+ | j, m0 i
p
S+ = (J+ )m,m0 = h j, m| = h̄ j( j + 1) − m0 (m0 + 1)δm,m0 +1
√ | {z }
=h̄ j( j+1)−m0 (m0 +1)| j,m0 +1i (9.27)
S− = (S+ )†
 Example 9.1 Case j = s = 1/2:

!
(J+ ) 1 , 1 (J+ ) 1 ,− 1
   
0 h̄ † 0 0
S+ = 2 2 2 2 = and S− = (S+ ) = (9.28)
(J+ )− 1 , 1 (J+ )− 1 ,− 1 0 0 h̄ 0
2 2 2 2
156 Chapter 9. Spin Angular Momentum

With this input we find

 
1 h̄ 0 1
Sx = (S+ + S− ) =
2 2 1 0
  (9.29)
1 h̄ 0 −i
Sy = (S+ − S− ) =
2i 2 i 0

and recalling our earlier results, using j( j + 1) = s(s + 1) = 21 ( 12 + 1) = 3

4
   
h̄ 1 0 3 1 0
Sz = and ~S = h̄
2 2
(9.30)
2 0 −1 4 0 1


 Example 9.2 Case j = s = 1:

   √   
(J+ )1,1 (J+ )1,0 (J+ )1,−1 0 h̄ 2 √
0 0
√ 0 0
S+ =  (J+ )0,1 (J+ )0,0 (J+ )0,−1  = 0 0 h̄ 2 and S− = h̄ 2 √0 0
(J+ )−1,1 (J+ )−1,0 (J+ )−1,−1 0 0 0 0 h̄ 2 0
(9.31)

With this input we find

 
0 1 0
1 h̄ 
Sx = (S+ + S− ) = √ 1 0 1
2 2 0 1 0
  (9.32)
0 −i 0
1 h̄ 
Sy = (S+ − S− ) = √ i 0 −i
2i 2 0 i 0

and recalling our earlier results, using j( j + 1) = 1(1 + 1) = 2

   
1 0 0 1 0 0
Sz = h̄ 0 0 0  and ~S2 = 2h̄2 0 1 0 (9.33)
0 0 −1 0 0 1

 9.5 Pauli matrices and spin wave functions for j ≡ s = 1/2. 157

9.5 Pauli matrices and spin wave functions for j ≡ s = 1/2.

Since spin does not emerge from the quantum mechanics of the Schrödinger equation, it has to
be grafted on as an additional postulate: the wave function of a particle with spin is a product
of the spatial wave function, Ψ(~r,t) which satisfies the time dependent Schrödinger equation,
and a spin wave function, χspin which is a (2 j + 1)−component vector obtained from the
matrix formalism:3

Ψtotal = Ψ(~r,t)χspin (9.34)

This is not to say that QM is unable to incorporate spin into the theory in a unified way; the
missing ingredient is relativity: the Schrödinger equation is non-relativistic. Shortly after the
Schrödinger equation was first introduced Dirac published his relativistic wave equation for
the electron - the Dirac equation - which not only gives the correct spin angular momentum,
but also an excellent approximation to its measured magnetic moment and a prediction that
the electron has an antiparticle. Relativistic wave equations also exist for other values of spin.
In the Dirac equation the wave function contains all three ingredients in a unified way: the
space-time dependence, the spin dependence and the antiparticle wave function.
For spin-1/2, we have seen that the angular momentum operators b ~ now called ~S for ‘spin’
J,
and without the ‘hat’, are represented by 2 × 2 Hermitian matrices Si , given by the Pauli
matrices σi :
h̄
Si = σi , i = 1, 2, 3 (or, in an often used equivalent labelling, i = x, y, z), (9.35)
2

     
0 1 0 −i 1 0
where: σ1 = σx = σ2 = σy = σ3 = σz =
1 0 i 0 0 −1
(9.36)

We emphasise again that the matrices satisfy precisely the same commutation relations as
the quantum mechanical angular momentum operators themselves (please check this for the
s = 1/2 and s = 1 cases described in the previous Section!):

[Sx , Sy ] = ih̄ Sz , [Sy , Sz ] = ih̄ Sx , [Sz , Sx ] = ih̄ Sy (9.37)

Another useful way to look at the Pauli matrices is that any Hermitian 2 × 2 matrix can be
written as a linear combination of Pauli matrices and the unit matrix,
 
1 0
I = σ0 = (9.38)
0 1
3 Actually the situation is a little more complicated: when a spinning particle is subjected to a magnetic
interaction the spin part becomes separately time-dependent . This is achieved through having time-dependent
coefficients a = a(t) and b = b(t) in the expansion theorem for χspin in eqs. (9.47) & (9.48). We will see an
example of this in our discussion of spin precession of particles in a magnetic field.
158 Chapter 9. Spin Angular Momentum

The eigenvectors (or eigenstates), χ+ and χ− , of the diagonal matrix Sz = h̄/2 σz are interpreted
as quantum states with definite values of the z−component of spin; these values are the
eigenvalues h̄ms = +h̄/2 and h̄ms = −h̄/2. To find these eigenstates we apply a systematic
procedure, although the result is fairly obvious:
   
1 0
χ+ = and χ− = (9.39)
0 1

We are using the notation χ in the case of spin-1/2 for the column vectors denoted C in the
general discussion of section 2.

Proof: The requirement that χ be an eigenstate of Sz is:

h̄
Sz χ = λ χ, (9.40)
2
where, although we already know that the eigenvalues of Sz correspond to λ = ±1, we shall
find it convenient to actually confirm this by calculation. Now the most general form for any
normalised spin-1/2 wave function is
 
a
χ= with |a|2 + |b|2 = 1. (9.41)
b

Our task is to find a, b and λ . Using the known form for the Sz matrix, the requirement that χ
be an eigenstate of Sz becomes:
 
h̄ h̄ a
Sz χ = λχ= λ
2 2 b
  
h̄ 1 0 a
=
2 0 −1 b
 
h̄ a
= (9.42)
2 −b
Equating the elements of column vectors in the first and last lines and cancelling h̄/2, we find:

aλ = a (9.43)
bλ = −b (9.44)

These two equations are clearly inconsistent if both a and b are non-zero; but they can be
satified by taking one to be zero. This gives two possibilities,
either b = 0 a 6= 0 and therefore λ = +1
and normalising, a = 1
or a = 0 b 6= 0 and therefore λ = −1
and normalising, b = 1
Q.E.D.
9.5 Pauli matrices and spin wave functions for j ≡ s = 1/2. 159

As befits eigenstates of a Hermitian operator with different eigenvalues, they are orthonormal,
ie. both normalised and orthogonal:

† †
χ± χ± = 1 and χ± χ∓ = 0 (9.45)

Since ~S2 = Sx2 + Sy2 + Sz2 commutes with Sz they are also eigenstates of ~S2 with common
eigenvalue 3h̄2 /4:

h̄
Sz χ± = ± χ± and ~S2 χ± = 3 h̄2 χ± (9.46)
2 4

All these equations can be checked out explicitly by using the matrices for the spin operators,
and you should do this for yourself. Moreover you can see that the Expansion Theorem
holds because any 2-dimensional vector, χ, can be written as a linear combination of the two
unit vectors χ± given in eq. (9.39): 4

 
a
χ = a χ+ + b χ− = (9.47)
b

while χ † = ( a∗ b∗ ), with normalisation: ∑ms |cms |2 = |a|2 + |b|2 = 1 (9.48)

In general the state χ is not an eigenstate of any of the spin operators. This is just like the
case of states obtained from linear combinations of energy eigenstates: they, too, are not
eigenstates of the Hamiltonian operator. The case of spin is so much simpler because the
space is finite-dimensional (2-D for spin-1/2) and so there are always only a finite number
eigenstates and terms in the expansion theorem. However the measurement postulate is the
same.
Measurement Postulate: For a spin-1/2 particle in a general state χ given above, the possible
outcome of a measurement of the z−componenet of spin is one of the eigenvalues ±h̄/2 (ie.
ms = ±1/2), with respective probabilities |a|2 and |b|2 . If say, a measurement yields a value
ms = +1/2, then the wave function ‘collapses’ to the eigenstate χa f ter = χ+ . Since this is
an eigenstate of Sz , a subsequent measurement will yield the same value ms = +1/2 with
probability 1. We will shortly explore these concepts using the Stern-Gerlach experiment as
the method for preparing spin states and measuring spin components.

4 This is a special case of equation (9.22) for spin-1/2,

with spin eigenvalues j = s = 1/2 and m j = ms = ±1/2,
and the spinorial wave function χ becomes a two-component vector with c1/2 = a, c−1/2 = b.
160 Chapter 9. Spin Angular Momentum

Figure 9.6: Illustration of the Stern-Gerlach experiment with spin 1/2 particles. In I) they are
in different random spin states coming out of an oven, in II) they are all in the same general
state χ (see (9.47)).

 Example 9.3 QUBIT: Expectation values of spin operators for a general spin 12 state
 
a
Assume the spinning particle is in a general spin state χ = ; we will also need the
b
Hermitian conjugate χ † a∗ b∗ , where a and b are complex numbers.

Normalisation requires χ † χ = |a|2 + |b|2 = 1.

In general expectation values are computed in matrix QM as hAi = χ † Aχ where now A is
one of the three spin matrices:

  
† h̄ † ∗ ∗
0 1 a
hSx i = χ Sx χ = χ a b
2 1 0 b
 
h̄
b h̄
= χ † a∗ b∗ = (a∗ b + b∗ a) = h̄ Re(a∗ b)
2 a 2
hSy i = χ † Sy χ = . . . = h̄ Im(a∗ b) (9.49)
  
† h̄ † ∗ ∗
1 0 a
hSz i = χ Sz χ = χ a b
2 0 −1 b
 
h̄
a h̄
= χ † a∗ b∗ = (|a|2 − |b|2 )
2 −b 2
9.5 Pauli matrices and spin wave functions for j ≡ s = 1/2. 161

With some effort you can prove (using the normalisation condition) that
2 2 2 h̄2
hSx i + hSy i + hSz i = , (9.50)
4
which means that the vector of expectation values of all spin operators h~Si lives on a sphere of
radius h̄/2. This is also called the Bloch sphere and represents the state space of a spin 1/2
particle.
Spin 12 particles like electrons, protons,.. are prototypical examples of qubits, the smallest
units of information in quantum computers. Note that the smallest units of information in
usual, classical computers are called bits that can only take the value 0 or 1 (binary choice). A
q(uantum)bit allows  to use
 quantum mechanical superpositions of up and down states of the
a
spin 21 particles χ = = aχ+ + bχ− which contains much information in the coefficients
b
than the binary choices of the classical bits. 

Energy eigenstates including spin of electrons in atoms: We can now also describe the
inclusion of spin in the electron wavefunction. If we assume a Coulomb potential from the
nucleus and a constant magnetic field in the z-direction ~B = (0, 0, Bz ) then the combined
Hamiltonian is:
h̄2  2
∂ 2 ∂
 b2
~L Ze2 µB
b=−
H + + − + ~B · (~L b + 2~S)
2m ∂ r 2 r ∂r 2mr 2 4πε0 r h̄
2
(9.51)
2  2 
~ 2
h̄ ∂ 2 ∂ L
b Ze µB Bz b
=− 2
+ + 2
− + (Lz + 2Sz )
2m ∂ r r ∂r 2mr 4πε0 r h̄
Note this is a simplified version of the Pauli equation for spin 1/2 particles, which itself can be
obtained as a non-relativistic (slow velocity v << c) approximation of the relativistic Dirac
equation.
In this case the maximal set of commuting operators is enlarged to include Sz (it commutes
with everything else in the Hamiltonian) to give {H, b ~Lb2 , L
b , S }. So energy eigenstates are
z z
simultaneous eigenstates of these four operators and are labelled by the corresponding quantum
numbers {n, `, m, ms }. Energy eigenstates have the form
Rn` (r)Y`m (θ , φ )χ+ for ms = + 21 spin up
(
ψn`mms = (9.52)
Rn` (r)Y`m (θ , φ )χ− for ms = − 21 spin down
and the energy eigenvalues receive an additional contribution (Zeeman effect including spin)
from the last term in the Hamiltonian (9.51)
µB Bz E1
Entot = En + (h̄m + 2h̄ms ) = 2 + µB Bz (m + 2ms ) , (9.53)
h̄ n
where En and E1 are the usual eigenvalues without magnetic field. A general state of an
electron is then a linear combination of these ψn`mms
 
Ψ+ (~x,t) tot
Ψ(~x,t) = = ∑ cn`mms ψn`mms (r, θ , φ )e−iEn t/h̄ (9.54)
Ψ− (~x,t) n,`,m,ms
 162 Chapter 9. Spin Angular Momentum

9.6 Spin quantisation along the x-axis: eigenstates of Sx

Given the eigenfunctions, eq. (9.39), for spin quantised along the z−axis, we now wish to find
the eigenfunctions for spin quantised along the x−axis. This is equivalent to asking for the
eigenstates χ 0 of Sx ,

h̄
Sx χ 0 = λ χ 0 . (9.55)
2
where, although we already know that λ = ±1 (because we are describing a spin-1/2 particle),
we shall find it convenient to actually confirm this by calculation. Now the information we
begin with is that the eigenstates of Sz are given by eq. (9.39), so we can use the expansion
theorem to express our unknown wave functions χ 0 as a linear combination, eqs. (9.47), (9.48):

 
0 a
χ = aχ+ + bχ− = with |a|2 + |b|2 = 1. (9.56)
b

and then use the matrix representation Sx ,

 
h̄ 0 1
Sx = 2 (9.57)
1 0

Putting these all together, the requirement that χ 0 be an eigenstate of Sx reads:

 
0 h̄ 0 h̄ a
Sx χ = λ χ = λ
2 2 b
  
h̄ 0 1 a
=
2 1 0 b
 
h̄ b
= (9.58)
2 a

Equating the elements of column vectors in the first and last lines and cancelling h̄/2, we find:

aλ = b (9.59)
a = bλ (9.60)

Dividing the equations gives a consistency condition leading to the expected two eigenvalues:

λ 2 = 1 hence λ = ±1 (9.61)
0 be
Since a and b have the same magnitude for both eigenvalues, the condition that χ±
normalised requires:

1
|a|2 + |b|2 = 2|a|2 = 1, or, taking a real and positive, a= √ (9.62)
2
9.6 Spin quantisation along the x-axis: eigenstates of Sx 163

1 1
and therefore a = √ b = ± √ for λ = ±1 (9.63)
2 2
0 and χ 0 ,
Finally we obtain the two eigenstates χ+ −

0 1
χ± = √ (χ+ ± χ− ) (9.64)
2
or, in explicit form,
   
0 1 1 0 √1
1
χ+ = 2 √ and χ− = (9.65)
1 2 −1

where the ± denote eigenstates of Sx with eigenvalues ±h̄/2. It goes without saying that these
states are still also eigenstates of S2 with the same eigenvalue 3h̄2 /4 as the original eigenstates
χ± of Sz . We shall see the physical significance of all four of these states when we consider
the triple Stern-Gerlach experiment.
Going between Sz eigenstates and Sx eigenstates:
We summarise here the main result and the connections of the two sets of eigenvectors
(eigenstates) Sz χ± = ± h̄2 χ± and Sx χ± 0 = ± h̄ χ 0 . It is just a consequence of the generalised
2 ±
expansion theorem that we can expand one set of eigenstates in terms of the eigenstates of the
other operator!
 
0 1 1 1
χ+ = √ = √ (χ+ + χ− )
2 1 2
 
0 1 1 1
χ− =√ = √ (χ+ − χ− )
2 −1 2
  (9.66)
1 1 0 0
χ+ = = √ (χ+ + χ− )
0 2
 
0 1 0 0
χ− = = √ (χ+ − χ− )
1 2
 164 Chapter 9. Spin Angular Momentum

9.7 Rotated spin operators and their eigenfunctions

As a preparation for our discussion of the Stern-Gerlach experiment we now ask the following
question: Suppose we prepare a spin-1/2 system in an eigenstate of Sz and then subsequently
measure the component of spin in some other direction; what are the possible outcomes?
An equivalent way to pose this question is: What are the eigenstates of the spin operator
corresponding to a direction specified by the unit vector ~n? Because the spin operator is a
vector its component along ~n is simply the projection,

Sn =~n · ~S 0 for which S χ 0 = ± h̄ χ 0

and we seek the χ± (9.67)
n ±
2 ±

For simplicity we only consider the case where the vector ~n lies along the direction obtained
by rotating the z-axis by an angle α about the y-axis. If we call this the z0 -axis, and the rotated
x-axis perpendicular to it the x0 -axis, then we are just asking for the new spin operators in the
new coordinate system (x0 , y0 , z0 ):

z
z0 6
K
A
A
A

A
A


Aα
A
A -y = y0




α

AA
U





x

x0

Figure 9.7: Picturing the rotation by an angle α about the y-axis.

Since ~S transforms as a vector the figure shows that the relation between new and old
matrices is:

Sx0 = Sx cos α − Sz sin α

Sy0 = Sy (9.68)
Sz0 = Sz cos α + Sx sin α ≡ Sα
9.7 Rotated spin operators and their eigenfunctions 165

so that the new matrices become,

 
h̄ − sin α cos α
Sx 0 =
2 cos α sin α
 
h̄ 0 −i
Sy0 = Sy = (9.69)
2 i 0
 
h̄ cos α sin α
S z0 = S α =
2 sin α − cos α

As a check on our algebra we note that α = 0 or α = 360◦ gives back our original matrices,
while α = 90◦ just exchanges the x- and z-matrices, with a sign change on Sx0 , as we would
expect since the new x0 -axis is now playing the role of the original z-axis. This shows rather
clearly that it’s just a matter of our choice onto what axis we decide to measure the projection
of the spin.
Now we look for the eigenstates of rotated version of the Sz matrix, namely Sα = Sz0 . By the
expansion theorem, eq. (9.47), these eigenstates, χ 0 , must be normalised linear combinations
of the original eigenstates of Sz (eq. (9.39):
 
0 a
χ = aχ+ + bχ− = with |a|2 + |b|2 = 1. (9.70)
b

For this to be an eigenstate, with eigenvalue λ h̄/2 (where we already know λ = ±1 because
the particle has spin-1/2),
h̄
Sα χ 0 = λ χ 0 . (9.71)
2
Using our matrix representation for Sz0 and the expressions for the original eigenvectors χ±0 ,

 
0 h̄ 0 h̄ a
Sα χ = λχ = λ (9.72)
2 2 b
  
h̄ cos α sin α a
=
2 sin α − cos α b
 
h̄ a cos α + b sin α
= (9.73)
2 a sin α − b cos α
Equating the elements of column vectors in the first and last lines, cancelling h̄/2 and collecting
coefficients of a and b, we find:

a(λ − cos α) = b sin α (9.74)

a sin α = b(λ + cos α) (9.75)

Dividing the equations gives a consistency condition leading to the expected two eigenvalues:

(λ − cos α)(λ + cos α) = sin2 α, ie. λ 2 − cos2 α = sin2 α, or λ 2 = 1, hence λ = ±1

166 Chapter 9. Spin Angular Momentum

(9.76)
Considering each case in turn gives the two eigenstates χ+0 and χ 0 , where we use the trig.
−
2 2
identities (1 − cos α) = 2 sin α/2, (1 + cos α) = 2 cos α/2, sin α = 2 sin α/2 cos α/2 :

b (1 − cos α) 2 sin2 α/2, sin α/2,

For λ = +1 : = = = (9.77)
a sin α 2 sin α/2 cos α/2 cos α/2

b −(1 + cos α) 2 cos2 α/2, cos α/2,

For λ = −1 : = =− =− (9.78)
a sin α 2 sin α/2 cos α/2 sin α/2
0 be normalised:
The magnitude of a in each case is obtained from the condition that χ±
b
|a|2 + |b|2 = |a|2 (1 + | |2 ) = 1 (9.79)
a

For λ = +1 : a = cos α/2, b = sin α/2 (9.80)

For λ = −1 : a = sin α/2, b = − cos α/2 (9.81)

Finally we obtain the required eigenstates of Sz0 :
   
0 cos α/2 0 sin α/2
χ+ = and χ− = (9.82)
sin α/2 − cos α/2
We can check our algebra by noting that we recover the original Sz eigenstates χ ± for
α = 0. But by putting α = 2π we fail to recover the original eigenstates, even ’though this
corresponds to the same point in space as α = 0. This quite remarkable property is an entirely
new phenomenon: the spin-1/2 wave functions are not single-valued, but double-valued
or in other words we have to perform a 4π rotation to get back the original wave function,
while after a 2π rotation it comes back with a minus sign!
0 0
χ± (α + 2π) = −χ± (α)
0 0
but, χ± (α + 4π) = χ± (α) (9.83)
Note that physical quantities are not affected by this problem. For example the expectation
value
hAi = χ † Aχ −→ (−χ † )A(−χ) = hAi
is invariant under a 2π rotation.
Finally we note the result we will apply to the double Stern-Gerlach experiment:
   
π 0 1 1 0 1 1
For α = ; χ+ = √2 and χ− = √2 (9.84)
2 1 −1
A glance at Figure 9.7 shows that α = π/2 corresponds to the new z-axis pointing along
the old x-axis: thus Sz0 = Sx and so these eigenstates are eigenstates of Sx , which we already
derived earlier.
 9.8 The Triple Stern Gerlach Experiment 167

9.8 The Triple Stern Gerlach Experiment

Before describing the details of the Stern-Gerlach experiment, we shall first abstract its essence
in order to discuss the theory of measurement in QM. The experiment can be thought of as a
method for either measuring a component of the spin or of preparing an ensemble of states
having one component of spin determined, ie. of preparing an eigenstate of Sz , Sx or whatever.
We represent the apparatus as a ‘black box’ into which a beam of spin-1/2 particles is injected
and from which emerge two spatially separated beams: one with spin projection along the
measurement axis +h̄/2; the other with −h̄/2. Figure 2 illustrates two separate measurement
experiments in which an unpolarised beam of spin-1/2 particles enters the apparatus. We can
think of the incoming beam as having a random mixture of all possible spin orientations: one
experiment measures the z−component; the other the x−component. The two outgoing states
0 respectively.
are spatially separated and exit in the pure eigenstates χ± and χ±

- χ+ - 0
χ+
- Sz - Sx
- χ− - 0
χ−

Figure 9.8: Abstract Stern-Gerlach experiments, one measuring the z−component of spin, the
other the x−component. Note that the outgoing beams have equal intensity one-half that of
the incoming beam.

We also repeat here the relation between Sz and Sx spin eigenstates which will be needed
below:
 
0 1 1 1
χ+ = √ = √ (χ+ + χ− )
2 1 2
 
0 1 1 1
χ− =√ = √ (χ+ − χ− )
2 −1 2
  (9.85)
1 1 0 0
χ+ = = √ (χ+ + χ− )
0 2
 
0 1 0 0
χ− = = √ (χ+ − χ− )
1 2
These illustrate the angular momentum quantization rule that for a spin-1/2 particle (the
incoming particle) the only possible outcome of a measurement of a component of the spin
angular momentum is one of the eigenvalues, +h̄/2 or −h̄/2. This is the first stage of the
measurement postulate of QM. Note how radically different the outcomes of these experi-
ments are compared with our classical expectation: since the incoming beam is unpolarised
we expect the spins to be oriented randomly in all possible directions, with projections on
the z-axis having a continuum of values lying between +h̄/2 and −h̄/2; instead we only
168 Chapter 9. Spin Angular Momentum

measure two values, and any single particle passing through the apparatus will exit either
in one beam or the other, not anywhere between the beams. Even more at variance with
our classical experience is that the very same incoming beam going through the Sx measur-
ing apparatus now seems to have only the maximum possible x-axis projections, ±h̄/2, a
clear impossibility if they already have the maximum z-axis projections as evinced by the
Sz measuring experiment! From the quantum mechanical point of view we have to accept
that in the incoming beam the spin projection of an individual particle is not known, only its
probability. Thus it is only after a measurement or preparation that we can say for certain
what it is, and then only at the moment of measurement, not before. If we are tempted to
say, ‘Well, it really had the value we determined all along and we just didn’t know it until we
actually looked’, then the fact that a similar measurement of the x-component can only yield
±h̄/2 should cast doubt on this interpretation, since now it seems it could ‘really’ have had
a component inconsistent with the value already determined. Thus it seems that a particle’s
property depends on which experiment we set up. We shall see this in the triple Stern-Gerlach
experiment. Thus, just like the two-slit experiment, we have to think of an incoming particle
as being in neither one state χ+ nor the other χ− , but ‘in some sort of equal mixture of the
two’; the physical meaning of this statement is that a measurement of the z-component of spin
for an unpolarised beam will yield only one of the two eigenvalues of Sz , but with equal proba-
bility. Once the measurement is made, the wave function ‘collapses’ into the appropriate state.5

The double Stern-Gerlach experiment enables us to illustrate the next stage of the mea-
surement postulate. First consider two Stern-Gerlach experiments, both measuring the z-
component of spin. We now discard one of the beams emerging from the first stage. This can
be viewed as measuring the z-component of spin and using this to prepare the system in the
eigenstate χ+ , or as the ‘collapse’ of the wave function into the state χ+ as a result of the
measurement (which is the process of separating the incoming beam and discarding the lower
one). Since the state entering the second stage has ‘collapsed’ to the eigenstate χ+ (or has
been prepared in that eigenstate) the result of the second measurement is certain: only one
beam now exits the second, corresponding to spin up.

5A subtle point about the wave function representing the unpolarised incoming beam: we have not dealt with
such incoherent states in this course, and so I have avoided writing down a wave function for it; but such states
have the property that we can still predict the probabilities I am talking about in the text. The formalism used for
these so-called mixed states is the density matrix formalism.
9.8 The Triple Stern Gerlach Experiment 169

- χ+
χ+
- Sz
- Sz
- χ−
Discarded
beam
Figure 9.9: Double Stern-Gerlach experiment illustrating the measurement process in QM.
Both measure the z-component of spin, but only one beam emerging from the first is admitted
into the second.
6

The triple Stern-Gerlach experiment illustrates several aspects of the quantum measur-
ing process and dramatises the consequences of the expansion theorem in a most remarkable
way. We now interpose between the two Sz experiments of the double Stern-Gerlach apparatus
a third experiment which measures the x-component, Sx :

- χ+
0
χ+
- Sz
χ+ Sx
- - χ−
- Sz - 0
χ−
Discarded
- χ− beam
Discarded
beam
Figure 9.10: Triple Stern-Gerlach experiment measuring sequentially the z, x and z-component
of the spin.

The first and last stages measure the z-component of spin, but the beam selected from the
first is now submitted to an Sx measurement and only one emergent beam is in its turn
submitted to the final stage, a second Sz measurement. The intermediate measurement
has now influenced the outcome of the last measurement and we see the phenomenon of
regeneration: the down spin state discarded in the first stage reappears in the outgoing beam of
the whole experiment, as if the intermediate measurement of Sx had ‘disturbed’ the system and
reintroduced some down component. In equations, the sequence is as follows: the incoming
state is an unpolarised mixture of spin up and spin down with respect to any chosen axis;
the z-axis is the one chosen for the first stage measurement, so the spin up and down beams
emerging from the Sz measurement are of equal intensity. Choosing to inject only the spin
up beam into the Sx measuring apparatus means that the incoming state for this stage is, χ+ ,
170 Chapter 9. Spin Angular Momentum

which, using the expansion theorem we may write in terms of the Sx eigenstates,
1 0 1 0
χ+ = √ χ+ + √ χ− (9.86)
2 2
Again there are only two beams emerging from the Sx measurement with equal intensity, from
0 . This is the initial state for the final S measurement, which
which we select the one in state χ+ z
we again expand, but this time in terms of Sz eigenstates:

0 1 1
χ+ = √ χ+ + √ χ− (9.87)
2 2
Here we have the final outcome: there are two equal intensity beams emerging from the
final stage, with each Sz eigenstate being present, despite the fact that the first stage of the
experiment had discarded the beam in the χ− spin down state.
Now we come to a crucial point: one may wish to attribute the regeneration observed in the
triple Stern-Gerlach experiment to the so-called ‘disturbance’ introduced by the intermediate
Sx measurement; but in fact this is no disturbance at all. We can show this by altering the last
experiment by not excluding the Sx spin down beam, but allowing it also to enter the last Sz
measuring phase. The result is that only the spin up component emerges from the last stage:

χ0
-+ - χ+
@
- χ+ Sx @@- Sz
χ0
-−
- Sz
- χ−
Discarded
beam
Figure 9.11: Variation of the triple Stern-Gerlach experiment without discarding the down
component at the intermediate stage. This shows that the intermediate stage introduces no
‘disturbance’ when no detection occurs there.

In equations this is simply stated: the two beams exiting the Sz apparatus are recombined
when they enter the final stage, maintaining their coherence. Thus the beam entering the final
stage is a linear superposition of equal intensity beams, with wave function
1 0 1 0
χSz out = √ χ+ + √ χ−
2 2
= χ+ (9.88)

where, in the last line, we recognise the combination of Sx eigenstates as the spin up Sz
eigenstate χ+ . Hence we confirm that the Sx apparatus itself does not disturb the system.
Rather than the collapse of the wave function resulting from a disturbance inside the Sx
apparatus, it is a consequence of a measurement being registered afterwards - in the case of the
9.8 The Triple Stern Gerlach Experiment 171

triple Stern-Gerlach experiment the measurement is the process of completely excluding the
lower beam emerging from the Sx apparatus; the upper beam is not touched at all and therefore
suffers no disturbance. The analogy with the two-slit experiment should be obvious: if we
block the beam from one slit altogether we lose the interference pattern. Thus we see that the
act of measurement introduces a degree of decoherence - the triple Stern-Gerlach experiment
described is an extreme case where one beam is completely destroyed, but one can imagine
introducing a counter in the lower beam instead as one does in attempting to follow the path of
the beams in the two-slit experiment. Finally we note that the experiments described here are
very closely related to the famous Aspect et al experiments which tested the Bell inequalities
and confirmed this apparently strange behaviour of the QM measurement process.
 172 Chapter 9. Spin Angular Momentum

9.9 Spin precession in a magnetic field

A classical dipole in a magnetic field precesses with the Larmour frequency, and the same
applies to the QM case of a magnetic moment produced by orbital motion. It may not be so
obvious that the same holds for a magnetic moment produced by spin, especially as electron
spin is so difficult to understand classically. Here we show how to study the problem using
QM, while at the same time illustrating the role of the Schrödinger equation in spin dynamics
and showing that it is the quantum mechanical expectation values that precess. In an applied
magnetic field the spin state χ is altered by the fact that the energy changes. This adds a term
to the Hamiltonian which acts only on the spin part of the wave function:

bs = −~µs · ~B = gs µB Bz Sz = ωs h̄ σz
H (9.89)
| {zh̄ } 2
ωs

The spin part of the wave function χ(t) is obtained by demanding it satisfy an appropriate
TDSE which takes account of the above interaction, and which induces a time-dependence:
 
c1 (t) bs χ(t) = ih̄ ∂ χ(t)
χ(t) = where H (9.90)
c2 (t) ∂t

    
h̄ h̄ 1 0 c1 ċ1
ie. ωs σz χ(t) = ωs = ih̄ (9.91)
2 2 0 −1 c2 ċ2

   
h̄ c1 ċ1
Hence, ωs = ih̄ (9.92)
2 −c2 ċ2

ωs ωs
ie. ċ1 = −i c1 and ċ2 = +i c2 (9.93)
2 2

Hence the solutions, c1 (t) = ae−iωst/2 and, c2 (t) = be+iωst/2 |a|2 +|b|2 = 1 (9.94)

where the constants a and b are chosen to normalise the wave function and to satisfy the
initial (t = 0) conditions. Two cases illustrate the relation between the classical and quantum
pictures:
Case (1) Suppose the wave function is initially in the spin up eigenstate χ+ of Sz ,

e−iωst/2
 
c1 (t = 0) = 1, c2 (t = 0) = 0, hence a = 1, b = 0 and χ(t) = (9.95)
0
9.9 Spin precession in a magnetic field 173

Now, this remains an eigenstate of Sz , with eigenvalue +h̄/2, for all times; all its expectation
values are time-independent and so it is a stationary state with no precession:
h̄ †
hSz i = χ † (t)Sz χ(t) = χ σz χ
2    −iω t/2 
h̄ 1 0 e s
e+iωst/2 0


=
2 0 −1 0
 −iω t/2 
h̄
e s
= e+iωst/2 0
2 0
h̄
hSz i = (9.96)
2

h̄ †
hSx i = χ † (t)Sx χ(t) = χ σx χ
2    −iω t/2 
h̄ 0 1 e s
e+iωst/2 0


=
2 1 0 0
 
h̄ 0
e+iωst/2 0


= −iω
2 e st/2
hSx i = 0 (9.97)

Similarly hSy i = 0. Thus we have the perhaps surprising outcome that a pure spin eigenstate
does not precess at all, despite the fact that in the quasi-classical pictures of the quantum
rules the spin vector doesn’t point exactly in the z-direction. However the expectation values
correspond precisely to the classical values: if the classical dipole is exactly aligned along
the z-axis (allowed classically) then not only does the spin have zero projection in the x- and
y-directions, but there is no torque acting and therefore no precession.
Case (2) Suppose the wave function is not a pure eigenstate but initially an equal mixture of
spin up and spin down:
 −iω t/2 
1 1 1 e s
c1 (t = 0) = c2 (t = 0) = √ , hence a = b = √ , and χ(t) = √ (9.98)
2 2 2 e+iωst/2

Since this state is an equal mixture of spin up and spin down it is not surprising to discover
that its average projection on the z-axis is zero:
h̄ †
hSz i = χ † (t)Sz χ(t) = χ σz χ
2
e−iωst/2
  
h̄ +iω t/2 −iω t/2

1 0
= e s e s
2 0 −1 e+iωst/2
e−iωst/2
 
h̄ +iω t/2 −iω t/2


= e s e s
2 −e+iωst/2
hSz i = 0 (9.99)
174 Chapter 9. Spin Angular Momentum
h̄ †
hSx i = χ † (t)Sx χ(t) = χ σx χ
2
e−iωst/2
  
h̄ +iω t/2 −iω t/2

0 1
= e s e s
4 1 0 e+iωst/2
e+iωst/2
 
h̄ +iω t/2 −iω t/2


= e s e s
4 e−iωst/2
h̄  +2iωst/2 
= e + e−2iωst/2
4
h̄
hSx i = cos ωst (9.100)
2
Similarly,
h̄
hSy i = sin ωst (9.101)
2
Here we have the analogue of the classical situation where the dipole lies in the x-y plane
with zero z-component: the spin vector will then precess around the z-axis. The QM example
shows that indeed the expectation values have precisely these properties; moreover we now
see that the QM precession frequency is ωs ≈ 2ωL for an electron.