Lecture Notes

P re fa c e

Greetings to you, enthusiastic aspiring quantum scientist! A warm welcome to you and your friends.
This summer school is aimed at giving junior college and polytechnic students a glimpse into the
peculiar concepts and exciting developments in the quantum world. Our goal is to give you a brief
introduction to quantum physics and its applications. The main questions we focus on are:

• What is quantum physics and how different compared to classical physics?

• What is the simplest quantum system and crazy stuff can we do with it?

• How is quantum physics useful to other areas of science E.g. Chemistry, Biology, Computer
Science & Engineering, etc.?

As far as possible, we will try to keep the lectures self-contained, assuming readers to have no prior
background other than secondary school mathematics and physics. While we strive explain all
quantum concepts as intuitively as possible, however, we may sometimes intentionally omit com-
plicated derivations and technical details in order to to keep the lectures succinct. Nevertheless,
students are always encouraged approach their corresponding lecturers for more clarification. In
addition, to do further reading after this short introductory course.

Recommended list of introductory texts on quantum physics:

1. D. J. Griffiths & D. F. Schroeter, Introduction to Quantum Mechanics, Cambridge University Press

(2018)
2. J. J. Sakurai & J. Napolitano, Modern Quantum Mechanics, Cambridge University Press (2021)
3. N. Zettili, Quantum Mechanics Concepts and Applications, Springer Publication (2013)
4. M. A. Nielsen & I. L. Chuang, Quantum Computing and Quantum Information, Cambridge Univer-
sity Press (2013)

All the best for your future endeavors!

— QCamp 2023 Organising Team,

Centre of Quantum Technologies

i
A c k n o w le d g e m e n t s

C o n t r ib u t o r s

The organising team is eternally grateful to

Invited Career Panelists

Alexander Ling Director, Quantum Engineering Programme; NUS
Chune Yang Lum Co-founder and CEO, SpeQtral

Lecturers
Alexander Hue History and Applications of Quantum Physics
Elizaveta Maksimova Complex Numbers and Linear Algebra
Peter Sidajaya Coins, Cats, and Superposition
Wu Shuin Jian Optical Qubits, Polarisation, and Interference
Atharv Joshi Superconducting Qubits and Quantum Computing
Fernando Valadares Superconducting Qubits and Quantum Computing
Zaw Lin Htoo Schrödinger’s Equation
Frits Verhagen Entanglement and Nonlocality
Enrique Cervero Quantum Cryptography
Chee Chong Hian Quantum Chemistry

Tutors
Png Wen Han Justin Peh Yu Xiang Soe Moe Thar
Zhang Rongjie Nigel Benjamin Lee Junsheng Tanvirul Islam
Huang Ni-Ni Nyayabanta Swain Harshank Shrotriya

Journal Club Facilitators

Ayesha Reezwana Benjamin Tan Huang Ni-Ni
Jonathan Schwinger Atharv Joshi Vindhiya Prakash

Discussion Circle Facilitators

Andrew Tanggara Angelina Frank Aw Cenxin Clive
Clara Fontaine Fernando Valadares Kyle Chu

QQuest Game Masters

Elizaveta Maksimova Huang Ni-Ni Nigel Benjamin Lee Junsheng
Ayesha Reezwana Kyle Chu Elizabeth Lim Mei Ying
Andrea Duina

ii
 Experimental QKD Team
Du Jinyi Lab Executive
Justin Peh Yu Xiang Lab Assistant
Wu Shuin Jian Lab Assistant
Zhang Xingjian Lab Assistant

Excursion Personnel
A*STAR Scientists and Staff
SpeQtral Researchers
Celine Trieu Chaperone
Kai Xiang Lee Chaperone
Jinge Bao Chaperone
Elizaveta Maksimova Chaperone

without whom QCamp 2023 would not have been possible.

Q C a m p 2 0 2 3 O r g a n is in g C o m m it t e e

Committee Members
Clara Fontaine Angelina Frank Aw Cenxin Clive
Zaw Lin Htoo Wu Shuin Jian Jinge Bao
Elizaveta Maksimova Chee Chong Hian

Advisors
Jenny Hogan Associate Director, Outreach and Media Relations, CQT
Puah Xin Yi Outreach and Communications Executive, CQT

iii
T im e t a b le

Monday Tuesday Wednesday Thursday Friday

5th June 2023 6th June 2023 7th June 2023 8th June 2023 9th June 2023

9.30am Group A Group B

Activity Qubits Getting Weirder Application

10.00am
Welcome to QCamp! Coins, Cats, and Schrödinger’s Equation Quantum Chemistry
Superposition
10.30am
Excursion Lab

Break Break Break A*STAR and Experimental Quantum Break

11.00am SpeQtral Key Distribution

Onboarding Qubits Getting Weirder Location: A*STAR Location: CQT Application

11.30am
History and Optical Qubits, Entanglement and Journal Club
Applications of Polarisation, and Nonlocality
Quantum Physics Interference
12.00pm

12.30pm
Lunch and
Lunch Lunch Shuttle back to CQT Lunch
Group Photo
1.00pm Lunch

1.30pm Lunch
Onboarding Discussion
Shuttle to A*STAR
Mathematics: Tutorial Tutorial Career Advice and
2.00pm Complex Numbers Experience Sharing
and Linear Algebra Superposition and Schrödinger’s Equation,
Optical Qubits Entanglement, and
Nonlocality
2.30pm
Break Break

3.00pm
Break Break
Lab Excursion
Tutorial
3.30pm Activity
Qubits Application Experimental Quantum A*STAR and
Mathematics:
Key Distribution SpeQtral QQuest
Complex Numbers
and Linear Algebra Superconducting Quantum
4.00pm Qubits and Cryptography
Quantum Computing

4.30pm
Break Break Break

Activity
5.00pm
Discussion Circle Discussion Circle Discussion Circle
Farewell Event and
Certificate Handout
5.30pm

Meeting Point: CQT Level 3 Seminar Room on Monday, Tuesday, Wednesday, and Friday;
A*STAR (Group A) or CQT Level 3 Seminar Room (Group B) on Thursday.
Dress code: Smart causal, long pants and covered shoes, with QCamp t-shirt on Tuesday.
Social Event: There will be an optional social event and dinner on Wednesday evening.

iv
C o n te n ts

Monday 2

1 History and Applications of Quantum Physics 2

1.1 The World Before Quantum . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2

1.2 The Rise of Quantum Mechanics . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3

1.3 The Ongoing Revolution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8

2 Complex Numbers 10

2.1 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16

3 Linear Algebra 18

3.1 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27

Tuesday 33

4 Qubits & Superposition 33

4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33

4.2 The Qubit State . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34

4.3 Measurement on a Qubit . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36

4.4 Some Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39

4.5 Quantum Operators . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42

4.6 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47

4.7 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48

v
C O N T E N T S

4.8 Exercise Solutions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51

5 From Light to Optical Qubits 56

5.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56

5.2 Ray Optics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56

5.3 Wave Optics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59

5.4 Electromagentic Optics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63

5.5 Appendix . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77

6 Superconducting Circuits for Quantum Computing 81

6.1 From Superconductors to Quantum Computers . . . . . . . . . . . . . . . . . . . . 81

6.2 Superconductivity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83

6.3 Superconducting Circuits . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86

Wednesday 96

7 The Schrödinger Equation 96

7.1 Historical Development of Quantum Mechanics . . . . . . . . . . . . . . . . . . . . 97

7.2 Fully Describing a Physical System . . . . . . . . . . . . . . . . . . . . . . . . . . . 100

7.3 Guessing the Quantum Time Evolution . . . . . . . . . . . . . . . . . . . . . . . . 104

7.4 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111

7.5 Continuous Case of the Schrödinger Equation . . . . . . . . . . . . . . . . . . . . . 113

8 Entanglement and Nonlocality 127

8.1 Entanglement and Teleportation . . . . . . . . . . . . . . . . . . . . . . . . . . . . 127

8.2 Bell Nonlocality . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 158

9 Cryptography 163

9.1 What is Cryptography and Why is it Important? . . . . . . . . . . . . . . . . . . . 163

9.2 Probability and Entropy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 166

9.3 Classical Cryptography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 169

9.4 Quantum Cryptography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 176

vi Q C a m p 2 0 2 3
C O N T E N T S

Friday 182

10 Quantum Chemistry 182

11 Quantum Journal Club 183

11.1 List of papers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 183

vii Q C a m p 2 0 2 3
Monday

1
C h a p te r 1

H is t o r y a n d A p p lic a t io n s

o f Q u a n tu m P h y s ic s

Lecture notes by Alexander Hue

Nearly a century ago, quantum mechanics was discovered. Since then, the word “quantum” has
been “abused” everywhere. For example, in films and television series, “quantum” is used as the
magical solution to every problem, ranging from enabling time travel to explaining magic. In
comics and games, it is the excuse used to explain time controlling powers or multiverses.

Meanwhile, this fascination of “quantum” does not stop in the realm of fiction. In popular media,
“quantum” is hailed as the solution to everything. Sometimes, it is the unimaginable force that
defies logic and common sense. Sometimes, it is this mysterious force that helps you to access your
inner peace and serenity. Sometimes, it is just used as a prefix on words to make it sound profound.
It is then not surprising that some people think that quantum mechanics is “just a theory,” and
that its practical purpose is very limited in “the real world.”

That cannot be further away from the truth. In reality, the discovery of quantum mechanics is
what makes the modern world. Its impact is still transforming our lives as we speak. While the
theory itself is counter-intuitive, it is the result of years of research and real-life experiments. To
understand the importance of quantum mechanics, let us start from the very beginning, before the
word “quantum” was even known.

1.1 T h e W o r ld B e fo re Q u a n tu m

“There is nothing new to be discovered in physics now. All that remains is more and more
precise measurement,” said Lord Kelvin in 1900. At the time, there were three main subfields of

2
C H A P T E R 1. H IS T O R Y A N D A P P L IC A T IO N S O F Q U A N T U M P H Y S IC S

physics that were known: classical mechanics (then called, simply, mechanics), thermodynamics,
and electrodynamics. Since these three managed to explain almost every physical phenomena
known then, it was thought that the whole of physics was solved. This turned out to be the
pinnacle of human hubris. Beyond this façade of success, there lies a few problems that were
deemed minor at the time.
The first problem deals with fire. We can tell the temper-
ature of a fire based on the colour of its flames. This is be-
cause fire emits different spectrum of light depending on its
temperature. The same goes for the human body, though
we emit light mainly in the infrared region. In an ideal case,
to keep the temperature constant, a body has to absorb
and emit light perfectly. We call this ideal object a “black-
body”. Like fire, it gives off radiation depending on the
The black-body radiation distribution and
temperature, and its spectrum was detected in experiments.
the “ultraviolet catastrophe,” which says
However, based on the knowledge of electromagnetism and
that we will be dead next to a fire.
thermodynamics then, its spectrum could not be explained.
In fact, according to the classical theories, the object emits so much radiation that people will be
vaporised standing next to it. This is called the “ultraviolet catastrophe.” This was one of the
cracks of classical physics.

The second problem concerns getting electricity from light. In 1887, Heinrich Hertz discovered
that when light shines on a coil, it helps transmit a spark across a gap. This is the photoelectric
effect, whereby light causes electricity. While it is easy to understand that light is the cause for
the electricity, it is difficult to explain away certain properties of the photoelectric effect based
on the idea of light as a wave. These two seemingly minor questions eventually changed the very
foundation of physics as was known. (For more technical details, refer to Schrödinger Equation)

1.2 T h e R is e o f Q u a n tu m M e c h a n ic s

T h e S o lu t io n

In 1900, Max Planck came up with a solution to the black-body radiation problem. He proposed a
guess solution to the distribution that matched well with the experimental data. The implication
of his solution is that light behaves like particles or small packets of energy, called “quanta”, which
we now know as a photon. However, since the wave picture of light was so well understood and
already proven in experiments multiple times, Planck treated it as a theoretical convenience to
avoid serious backlash from other physicists. However, Albert Einstein took this result seriously.
In 1905, based on the hypothesis that light behaves like particles, Einstein managed to explain all

3 Q C a m p 2 0 2 3
C H A P T E R 1. H IS T O R Y A N D A P P L IC A T IO N S O F Q U A N T U M P H Y S IC S

the strange properties found in the photoelectric effect. This marked the start of quantum theory.

T h e A to m

At the same time, our knowledge of atoms grew as we speak. In 1911, Ernest Rutherford discovered
that most of the mass of an atom concentrates in its centre, called the atomic nucleus. He proposed
an atomic model that electrons orbit around the atomic nucleus like planetary motion. However,
this model is unstable based on the known laws of electrodynamics. In 1913, inspired by the
discreteness nature started by the discovery of the light quanta, Niels Bohr proposed a different
model. He proposed that only certain energy levels are permitted in the electron orbits, and
these energy levels are discrete. This model successfully predicted the spectrum of hydrogen gases.
However, this story is far from over.

T h e F o r m a lis m s

Since the energy levels are discrete in nature, in 1925, Werner Heisenberg discovered a mathematical
structure that describes the laws of quantum mechanics. It was later realised that this has the same
mathematical structure as matrices, and so this approach is called the matrix mechanics (refer to
Complex Numbers & Linear Algebra). Heisenberg’s development of matrix mechanics eventually
led to the strange conclusion that certain properties cannot be measured to an infinite precision.
In particular, one cannot know the position and momentum of an object to an infinite precision
simultaneously. This is called the Heisenberg uncertainty principle, which has a huge philosophical
impact on reality, and is directly opposite to the deterministic viewpoint of classical mechanics.

Meanwhile, since we can see that light behaves as particles sometimes, is it not possible that
particles behave like waves sometimes too? This led to the idea of matter waves in 1924, proposed
by Louis de Broglie. He proposed an equation to calculate the wavelength of “normal” particles
and found that the wavelength is so small that it is not observable in normal conditions. On the
other hand, de Broglie’s wavelength allows the wave-like behaviour of smaller particles, like the
electron, to be observed in experiments. Electron diffraction has since been verified experimentally.
With this, the concept of wave–particle duality is complete. The description we call “wave” and
“particle” are just an incomplete ideal picture of reality that seemed contradictory because we only
observe the macroscopic world.

Inspired by the concept of matter waves, Schrödinger believed that everything can be described as
waves and began his journey to find the wave equation to describe it. In 1926, he discovered the
wave equation, the Schrödinger’s equation. One question remained: what is the meaning of these
waves? Eventually, Max Born proposed that these waves encode the probability of a measurement
outcome. Using this idea, they managed to predict the hydrogen spectrum with good agreement

4 Q C a m p 2 0 2 3
C H A P T E R 1. H IS T O R Y A N D A P P L IC A T IO N S O F Q U A N T U M P H Y S IC S

with experimental result. From then on, the unquestionable deterministic world was shattered,
replaced by a possibly probabilistic world (refer to Schrödinger Equation).

Q u a n tu m M e c h a n ic s

At this point, there were two very different viewpoints of quantum mechanics, Heisenberg’s matrix
mechanics and Schödinger’s wave mechanics. Eventually, while they began from different starting
points, it was found that they are in fact the same theory, the two sides of the same coin. So, by
the end of the roaring twenties, the basic rules of quantum mechanics was established. The self-
pleasure that is theoretical physics is over, and quantum mechanics was never heard from again...
Of course that is not true, because that was only the beginning. Soon, many began to apply the
strange ideas of quantum mechanics to everything else, creating new technologies along the way.

T h e Q u a n tu m R e v o lu t io n

With an entire new territory in physics to explore, people began to explore the applications of
quantum mechanics for the rest of the century. Countless discoveries and inventions were made,
and our lives have been changed because of quantum mechanics. Here is a nonexhaustive list of
all the technologies that has since appeared because of quantum mechanics.

5 Q C a m p 2 0 2 3
C H A P T E R 1. H IS T O R Y A N D A P P L IC A T IO N S O F Q U A N T U M P H Y S IC S

M ic r o s c o p y

When we want to look at very small stuff (say, a cell), we use the
microscope. However, there is a limit to what a microscope can
enlarge due to the wavelength of light. If the object is too small,
light begins to diffract just like water waves do when it encounters
an obstruction. The shorter the wavelength, the smaller we can
see. Unfortunately, there is a limit to how short the wavelength
of light can be. Thanks to wave–particle duality, we know that
electrons can behave like waves too, with a wavelength that is
smaller than visible light. Using the wave behaviour of electrons,
we have the electron microscope which allows us to observe even
smaller things than the optical microscope allows.

It gets even better. When we are dealing with probability waves,

things can decay through a barrier where they normally cannot Image of a pollen (top) and image

access. This is called quantum tunnelling, and has an analogue of copper (bottom) using scanning

phenomenon in classical electrodynamics. Using this property, the electron microscope and scanning

scanning tunnelling microscope (STM) was invented. This even tunnelling microscope respectively.

allows us to see and manipulate individual atoms.

T e c h n o lo g y fro m A to m s

The Bohr model began our quest to know more about atoms. While it was rudimentary, the Bohr
model allowed us to predict the spectrum of hydrogen. It was then natural to apply quantum
mechanics to atoms. This explains the discrete spectrum of atoms and molecules beyond hydrogen
gas, allowing us to use them as a fingerprint of different chemical species. This is how we know
the elements in stars and galaxies far far away. This is also why flame test in chemistry works.

As more is known about the atom, we can manipulate the

atoms to a finer degree. The knowledge of transition be-
tween different energy levels in the atom eventually leads
to the advent of lasers, which stands for “Light Amplifica-
tion by Stimulated Emission of Radiation.” Even though the
first functioning laser appeared in 1960, the basic govern-
ing equation was already written down by Einstein in 1917.
Nowadays, lasers have become an indispensable part of our lives, be it in entertainment (laser tag,
laser mouse), educational (laser pointers), and many more.

6 Q C a m p 2 0 2 3
C H A P T E R 1. H IS T O R Y A N D A P P L IC A T IO N S O F Q U A N T U M P H Y S IC S

The knowledge of transitions between energy levels in an atom

also allows us to invent the most accurate clock ever. Under
suitable conditions, the electrons in an atom can be induced to
oscillate between two energy levels. These oscillations are highly
regular and periodic, and it occurs very fast, which allows very
accurate timekeeping. It is so accurate that it is used to define the standard unit of time. Accurate
timekeeping is important because without it, the global positioning system (GPS) cannot work
properly.

Last but not least, quantum mechanics also found its use in nuclear physics. While the discovery of
radiation is experimental, the framework of quantum mechanics allows calculations and estimations
needed to design an atom bomb. This is the key weapon that helped the Allies in World War II and
it is what terrorises everyone nowadays. Of course, a better control over the radioactive materials
eventually led to nuclear power plants, which provides an enormous amount of power.

S o lid S ta te P h y s ic s

If quantum mechanics can be applied to study an atom, natu-

rally they can be used to study a lot of atoms. And what is
a solid if it is not a lot of atoms? So began the field of solid
state physics. Solid state physics is important because by un-
derstanding the principles behind how electrons flow in solids,
we can build smaller electronic components.

In the old days, amplifiers and switches were implemented using vacuum tubes. These vacuum
tubes are big and difficult to work with. Moreover, they need a lot of power and produces a lot of
heat. In 1947, with their understanding of solid state physics, William Shockley, John Bardeen, and
Walter Brattain designed a solid-state equivalent of vacuum tubes. This is the famous transistor.
It is much smaller in size and requires way less power to operate. This began the ever decreasing
size of electronics. Not only did it allowed the invention of portable radios, hearing aids, personal
computers, and countless other electronics, it also spawned an entire industry, the semiconductor
industry. This is how the Silicon Valley obtained its namesake.

Another resultant invention of solid state physics is the super-

conductor. In 1911, Heike Onnes discovered that when mer-
cury was cooled to 4.2 K, its resistance vanished. This is
called “superconductivity.” The explanation of superconductiv-
ity only came in 1957 with the help of quantum mechanics.
The knowledge of its working principle allowed scientists to de-
sign better superconductors that operate at higher temperature.

7 Q C a m p 2 0 2 3
C H A P T E R 1. H IS T O R Y A N D A P P L IC A T IO N S O F Q U A N T U M P H Y S IC S

With no resistance, superconductors allow us to create strong magnetic fields, like the ones needed
in maglevs. However, there are still an entire class of high-temperature superconductors whose
working principle is not known as of today, so better inventions are yet to come.

M e d ic a l P h y s ic s

Many of the inventions above have been used for medical

purposes. For example, lasers are used for laser surgery that
allows operations with better precision. Such surgeries in-
clude LASIK and neurosurgery. On the other hand, in 1925,
electrons were discovered to have an inherent magnetic prop-
erties in them called spin. Using superconductors, we can
create strong magnetic fields that probe these spins in our
bodies, allowing a non-intrusive peek inside our body. This
is the magnetic resonance imaging (MRI).

In 1928, Paul Dirac incorporated ideas of special relativity into quantum mechanics. The Schrödinger
equation was replaced by the Dirac equation. This predicted the existence of an antiparticle of the
electron, called the positron, before its eventual discovery in 1932. Nowadays, positrons are used
in medical physics in positron-emission tomography (PET) scans.

Q u a n tu m D o ts

Quantum dots are nanoparticles, made up of up to 50 atoms, that can emit a very specific wave-
length of light. The colour emitted by quantum dots can be fine-tuned since the colour of the
quantum dot depends on the size of the nanoparticle. Currently, quantum dots have been used in
television displays by companies such as Samsung. It provides a more accurate colour spectrum
in our electronic displays.

1.3 T h e O n g o in g R e v o lu t io n

As we can see, the advent of quantum mechanics has changed the landscape of technology entirely.
However, the story is not over yet. Every day, new attempts are made to achieve better quantum
technologies in laboratories all over the world. Here are a few new technology you can look out for
in the near future.

8 Q C a m p 2 0 2 3
C H A P T E R 1. H IS T O R Y A N D A P P L IC A T IO N S O F Q U A N T U M P H Y S IC S

Q u a n tu m C ry p to g ra p h y

The philosophical implication of quantum mechanics caused a huge debate between Einstein and
Bohr. One important argument of Einstein is that quantum mechanics is incomplete, resulting
in the “EPR Paradox” proposed in 1935. The paradox was not solved until 1964 by John Bell,
where quantum mechanics emerged victorious. This was done by using long-distance correlations of
particles in quantum mechanics, called quantum entanglement. Since then, quantum entanglement
has been applied to quantum cryptography, a method of encryption that cannot be hacked if
applied correctly (refer to Quantum Key Distribution, Entanglement & Teleportation, and Bell
Nonlocality).

Q u a n tu m M e t r o lo g y

Quantum metrology is a field that aims to increase measurement accuracy using properties of
quantum mechanics, such as quantum entanglement. This paves the way towards high-precision
sensors. We can apply knowledge from quantum metrology to the measurement of magnetic fields,
or even photography to get a ultra-high-definition picture.

Q u a n tu m C o m p u te rs

In condensed matter physics, numerical computation of properties of materials is difficult because

it requires huge computational power. Richard Feynman proposed that it would be simpler to
use quantum systems to compute the behaviour of quantum systems. This is the first notion of
quantum simulators, where quantum systems are used to simulate quantum systems.

Later, this idea was expanded. Using the properties of superposition, one uses a quantum bit
(qubit) instead of bit as their programming unit (refer to Qubits and Optical Qubits: Polarisation
& Interference). This can potentially perform certain calculations faster than a classical computer.
Several efficient quantum algorithms have been discovered, and many more are being discovered
still.

C o n c lu s io n

Quantum mechanics have changed the world. It has made many modern technologies possible.
All these technological advancements are only possible when we study physics of the small scale.
However, things do not stop at what we already have now. Researches into utilising quantum
theories for technological advancements are in progress. Soon, we may see yet another technological
revolution, where our lives continue to be changed by the strange theory that is quantum mechanics.

9 Q C a m p 2 0 2 3
C h a p te r 2

C o m p le x N u m b e rs

Lecture notes by Jamie Sikora

Edited by Tommy Tai, Angelina Frank and Krishna Chaitanya

Motivation

In physics, especially in quantum mechanics, it is convenient to express sinusoidal waves as

complex numbers. Complex numbers consist of two parts—the real part and the imaginary
part. Often, the phase of the wave is encoded in the imaginary component.

Complex Numbers

We define the square root of -1 as

√
i= −1.

Then a complex number can be written as

z = a + bi = Re(z) + Im(z)i,

where a, b ∈ R with R being the set of real numbers (−∞, ∞). Also, Re(z) denotes the real
part and Im(z) denotes the imaginary part of the complex number z.

Example 2.1. Solve the quadratic equation x2 + 5x + 10 = 0 for x.

Solution: Using the quadratic formula:

√
For ax2 + bx + c = 0, the roots are x = −b± b2 −4ac
2a :

√ √ √ √ √
−5 ± 25 − 4 · 10 −5 ± −15 −5 ± 15 −1 −5 15
x= = = = ± i.
2 2 2 2 2

10
C H A P T E R 2 . C O M P L E X N U M B E R S

Argand Diagram

We geometrically represent complex numbers on an Argand Diagram. The x axis represents

the real part of the complex number z while the y axis represents the imaginary part of the
complex number.

In Figure 2.1, we have illustrated three complex numbers, namely 3 + 2i, 2 − i, and −1 − 3i, on
the Argand Diagram. Notice that the complex number z = x + iy is represented like a 2D vector
starting from the origin and ending at the coordinate (x, y).

Figure 2.1: Argand Diagram.

Complex Number Arithmetic

Consider the complex numbers z = a + bi and w = c + di. Then we can define

• z + w = (a + bi) + (c + di) = (a + c) + (b + d)i.

• zw = (a + bi)(c + di) = ac + adi + bci + bdi2 = (ac − bd) + (ad + bc)i.

For the addition of two complex numbers, you can see that we add the real and imaginary numbers
separately.

Example 2.2. Consider the following examples.

1. (1 + 0.5i) + (2 + 3i) = 3 + 3.5i.

2. (1 + 3i)(−2 + i) = −2 + i − 6i + 3i2 = −2 − 5i − 3 = −5 − 5i.

11 Q C a m p 2 0 2 3
C H A P T E R 2 . C O M P L E X N U M B E R S

Figure 2.2: Addition of Complex Numbers (2 + 3i) + (1 + 0.5i).

Consider Example 2.2.1. We illustrated the addition of two complex numbers in Figure 2.2. Here,
we want to add 2 + 3i represented by the orange vector and 1 + 0.5i represented by the blue vector.
The final result is 3 + 3.5i which is represented by the green vector. The dashed orange vector is
obtained by translating the orange vector such that its starting point coincides with the end point
of the blue vector. Clearly, the green vector is akin to the resultant vector after a vector addition.

The multiplication of two complex numbers is better illustrated using the polar representation of
complex numbers, which we will introduce later.

P r o p e r t ie s o f C o m p le x N u m b e rs

Complex Conjugate

We define the complex conjugate of a complex number z = a + bi as

z ∗ = a − bi.

Notice that we simply flip the sign of the imaginary component of z. Complex conjugation
of a complex number z is denoted as z ∗ . On the Argand Diagram, the conjugate is achieved
by taking the reflection of the vector about the Re(z) axis.

12 Q C a m p 2 0 2 3
C H A P T E R 2 . C O M P L E X N U M B E R S

Example 2.3. What are the complex conjugates of the following complex numbers?

• z1 = 5 + 10i.

• z2 = 3 − 2i.

• z3 = −3 = −3 + 0i.

• z4 = 2i = 0 + 2i.

Solution:

Conjugating gives

• z1∗ = 5 − 10i.

• z2∗ = 3 + 2i.

• z3∗ = −3 − 0i = −3 = z3 . (unchanged, why?)

• z4∗ = 0 − 2i = −2i = −z4 . (totally changed, why?)

Figure 2.3: Conjugation of Complex Numbers.

Modulus

The modulus of a complex number z = a + bi is given by

p
|z| = a2 + b2 .

The modulus of z is the length of the vector which represents the complex number z.

Note: The modulus of a complex number is like the absolute value of a real number, as |z| ≥ 0.
Think about the distance between a real number and the origin on the number line! Example

13 Q C a m p 2 0 2 3
C H A P T E R 2 . C O M P L E X N U M B E R S

2.4. What is the modulus of the following complex numbers?

• 5 + 10i.

• 3 − 2i.

• −3 = −3 + 0i.

• 2i = 0 + 2i.

Solution:

We can use the formula to calculate the modulus of each complex number.
√ √
• |5 + 10i| = 52 + 102 = 125.
√
• |3 − 2i| = 32 + (−2)2 = 13.
p

√
• |−3 + 0i| = (−3)2 + 02 = 9 = 3. (This is the absolute value – distance between |3|
p

and 0 on the number line!)

√ √
• |0 + 2i| = 02 + 22 = 4 = 2. (This is like the absolute value!)

Property 2.1. Below is a list of properties for complex numbers z and w.

• zw = wz.

• zz ∗ = |z| .
2

1 z∗
• z −1 = = 2 . when z 6= 0
z |z|
• (zw) = z ∗ w∗ .
∗

• (z + w)∗ = z ∗ + w∗ .

• (z ∗ )∗ = z.

• |z| = |z ∗ |.

• |zw| = |z||w|.

• |z + w| ≤ |z| + |w|.

1+i
Example 2.5. What if we are given the fraction
, does this make sense? Yes, as long as
2−i
the denominator is not 0 (or 0 + 0i). We can clean it up as
1+i (1 + i)(2 + i) 1 + 3i 1 3
= = = + i.
2−i (2 − i)(2 + i) 5 5 5
Notice how we multiplied the top and bottom by the conjugate of the denominator.
c + di
More generally, to get the form of a + bi out of we can always do
p + qi
c + di p − qi cp + dpi − cqi + dq (cp + dq) + (dp − cq)i
= 2 2
= .
p + qi p − qi p +q p2 + q 2
cp + dq dp − cq
So, a = and b = 2 .
p2 + q 2 p + q2

14 Q C a m p 2 0 2 3
C H A P T E R 2 . C O M P L E X N U M B E R S

So far, we have limited our discussion of complex numbers to the rectangular coordinate represen-
tation of complex numbers known as Cartesian plane. However, such a representation is limited
in conveying the true richness of complex numbers.

Polar Representation of Complex Numbers

A complex number z on the Argand Diagram can also be specified in polar coordinates
(r, θ) where r is the distance from the origin and the θ is the angle from the Re(z) axis.

We can freely convert a complex number’s representation from rectangular form z = x+iy to polar
form z = reiθ .
  
y
r = x2 + y 2 , θ = tan−1
p
(x, y) → .
x
(r, θ) → (x = r cos(θ) , y = r sin(θ)) .

Figure 2.4: Polar Representation of Complex Numbers.

Euler’s Formula

Euler’s formula establishes an alternative way of expressing the polar representation of

complex numbers (r, θ) using the complex exponential

re±iθ = r(cos(θ) ± i sin(θ)). (2.1)

The formula can be proved by adding the Taylor (Maclaurin works too) series expansion of the
cosine function and the sine function, multiplied by i. In polar coordinates, complex multiplication
is given by (r1 , θ1 ) · (r2 , θ2 ) = (r1 r2 , θ1 + θ2 ). This is true because r1 eiθ1 · r2 eiθ2 = r1 r2 ei(θ1 +θ2 ) =
r1 r2 (cos(θ1 + θ2 ) + i sin(θ1 + θ2 )).

15 Q C a m p 2 0 2 3
C H A P T E R 2 . C O M P L E X N U M B E R S

2 .1 E x e r c is e s

Exercise 2.1. Calculate the following to the form of a + bi:

(a) (1 + i) + (2 − 2i) (b) (1 − 3i) − (4 + 7i) (c) (3 + 5i) + (π + 2πi)

(d) (1 + i)(4 − i) (e) (1 + 9i)(−1 + i) (f ) (1 − i)(2 + 0i)

(g) (1 + 2i)∗ (h) ((1 + 2i)∗ )∗ (i) (((1 + 2i)∗ )∗ )∗

(j) |1 + 2i| (k) |4 + 10i| (l) |2 + i|

(m) (1 + 2i)−1 (n) (4 + 10i)−1 (o) (2 − i)−1

Exercise 2.2. Write the following in polar coordinates:

(a) (1 + i) (b) (1 − 3i) (c) (3 + 5i)

(d) (4 − i) (e) (1 + 9i) (f ) (2 + 0i)

Exercise 2.3. Write the following in rectangular coordinates:

√
(a) (1, 2π/3) (b) ( 2, π/4) (c) (5, tan−1 (3/4))

(d) (2, 0) (e) (1, π/2) (f ) (3, tan−1 (3))

Exercise 2.4. Using what we have learn so far about complex numbers, prove that

(a2 + b2 )(c2 + d2 ) = (ac − bd)2 + (ad + bc)2 ,

for any complex numbers a, b, c, d.

16 Q C a m p 2 0 2 3
C H A P T E R 2 . C O M P L E X N U M B E R S

For the interested reader:

de Moivre’s Theorem

de Moivre’s Theorem states that

(r cos(θ) + i sin(θ))n = rn (cos(nθ) + i sin(nθ)),

which is a consequence of Euler’s Formula, as can be shown below

(r cos(θ) + i sin(θ))n = (reiθ )n = rn einθ = rn (cos(nθ) + i sin(nθ)),

where we simply used (2.1) to get the first and last equality.

Consider the complex number z = 1 + i, how will z · z and z · z · z look like? As seen in Figure 2.5,
the angle of z, z · z and z · z · z are 4, 2
π π
and 3π
4 respectively. Clearly, from Figure 2.5, the angle of
z raised to the power n will simply be 4 .
On the other hand, the modulus of z, z · z and z · z · z
nπ
√ √
are 2, 2 and 2 2. The modulus of z raised to the power n will be (2)n/2 .

Figure 2.5: Powers of Complex Numbers.

√
Using Euler’s formula, z = 1 + i = 2ei 4 .Thus, z 2 = (1 + i)2 which upon evaluation using de
π

√ 3π
Moivre’s Theorem will yield z 2 = 2ei 2 . Likewise, z 3 = (1 + i)3 = 2 2ei 4 .
π

17 Q C a m p 2 0 2 3
C h a p te r 3

L in e a r A lg e b r a

Lecture notes by Jamie Sikora

Edited by Tommy Tai, Angelina Frank and Krishna Chaitanya

Motivation

In quantum mechanics, we represent the quantum state of a system as a wavefunction,

denoted as ψ. Multiple wavefunctions can physically superpose together like waves to
form a single resultant wavefunction. Mathematically, such resultant wavefunction can
be calculated by adding these multiple wavefunctions together, or as commonly said, by
taking a linear combination of wavefunctions. A collection of all possible superpositions of
wavefunctions or quantum states, is known as the Hilbert space.

The mathematical formalism of quantum mechanics treats wavefunctions as complex vectors

and the Hilbert space as a complex vector space. The branch of mathematics, that deals
with vectors and linear systems, relevant to quantum mechanics is linear algebra. In the
subsequent part of the notes we will use the standard Dirac notation to denote quantum
state vectors using kets |i and bras h|.

18
C H A P T E R 3 . L IN E A R A L G E B R A

V e c t o r A r it h m e t ic

Vector

A ket vector |vi is represented by a column of numbers. For example, a 2-dimensional

vector looks like  
v1
|vi = .
v2
b

Note that the symbol =

b denotes that the column of numbers is a representation for the
vector |vi in a particular basis. This will become more clear as we move along. You can
think of a vector with n numbers as a point in n-dimensional space. The set of all n-
dimensional vectors is denoted by Rn if the numbers are real or by Cn if the numbers are
complex.

   
v1 w1
Just like numbers, we can add vectors. Given two vectors |vi = and |wi = , adding
v2 w2
b b
them gives  
v1 + w1
|vi + |wi = .
v2 + w2
b

We can also add vectors of higher dimensions in a similar way.

   
1 3
Example 3.1. Add the vectors |vi = and |wi = .
2 −100
b b

Solution:        
1 3 1+3 4
+ = = .
2 −100 2 − 100 −98

We can also scale vectors by just  multiplying

 each entry by the same number. For example
v1
scaling the vector represented by ∈ C2 by some complex number c ∈ C gives the new vector
v2
represented by  
c · v1
c · |vi = ∈ C2 .
c · v2
b

Note that the new vector is of the same dimension.

 
i
Example 3.2. Scale the vector |vi = ∈ C2 by c = 7.
−2
b
  

i 7i
Solution: 7 · = .
−2 −14

Note that “vector multiplication” is not a well defined concept (as

 opposed
 to multiplication be-
1 3
tween two numbers, such as 3 × 5i = 15i). Consider two vectors and . Think about in
2 −100
     
1×3 1×3 2×3
what way do we “multiply” them? ? Or, + ? However, instead
2 × −100 1 × −100 2 × −100
of “vector multiplication” there are inner products!

19 Q C a m p 2 0 2 3
C H A P T E R 3 . L IN E A R A L G E B R A

In n e r P r o d u c t s a n d N o rm s

Inner Products

Before definingthe inner product of 2 vectors, let us introduce the conjugate transpose of
v
a vector |vi =
b 1 ∈ C2 . The conjugate transpose of |vi, denoted by hv| (called bra v), is
v2
represented by a row of numbers:

b v1∗ v2∗ (3.1)

 
hv| =

   
v1 w1
The inner product of two complex vectors |vi = ∈ C and |wi =
2
∈ C2 is then
v2 w2
b b
defined as
hv|wi = v1∗ · w1 + v2∗ · w2 .

Recall that z ∗ is the complex conjugate of the complex number z. We can take the inner
product of larger dimensional vectors in a similar way.

Note that in linear algebra, “inner product” is a more general family of functions that maps two
vectors to a number that satisfies certain properties. How we define the inner product above is one
of its instance called the “dot product”. For the purpose of this lecture, we will define the inner
product as the dot product as above.

Orthogonal Vectors

We say the two vectors |vi and |wi are orthogonal if hv|wi = 0. Geometrically in a plane
(i.e: two dimensional vector space), these vectors make a 90◦ angle with each other.

   
i 3
Example 3.3. What is the inner product of |vi = and |wi = ? Are they orthogonal?
2 i
b b

Solution: No, they are not orthogonal

 
∗ ∗
 3
i 2 = (−i)(3) + (2)(i) = −i.
i

 √   √ 
1/√2 1/ √2
Example 3.4. What is the inner product of |vi = and |wi = ? Are they
−1/ 2
b b
1/ 2
orthogonal?

Solution: Yes, they are orthogonal

" #
h ∗ ∗
i √1 1 1
√1 √1 2 = − = 0.
2 2 −1
√
2
2 2

20 Q C a m p 2 0 2 3
C H A P T E R 3 . L IN E A R A L G E B R A

Norm

The norm of the vector |vi is defined as

p
hv|vi. kvk =
 
v1
q
In particular, the norm of the vector |vi = ∈ C2 is kvk = |v1 | + |v2 | , where we
2 2
v2
b
can see in this two-dimensional case that the norm of |vi is its length on the plane.

 
i
Example 3.5. What is the length of the vector |vi = ?
−2
b

Solution: q
2 2
√
kvk = |i| + |−2| = 5.

Unit Vectors, Normalizing

Unit vectors, or normal vectors, have norm equal to 1. Normalizing a non-zero vector
|vi means to scale it by kvk
1
to make its norm equal to 1, because if |zi = 1
kvk |vi then
q
kzk = kvk2 = kvk = 1 for any non-zero |vi.
hv|vi kvk

Basically, normalizing a vector stretches it if the norm is less than 1 or shrinks it if the norm is
greater than 1. Note it still points in the same direction afterwards.
 
0
What happens when you try to normalize the vector |vi = ? The norm is 0 so kvk
1
is not
0
b
   
0 0
defined. Also c · = for all c ∈ C, so we cannot scale it to make it have norm 1 anyway.
0 0
Thus, it is important we are normalizing a vector that is not all 0’s.
 
1
Example 3.6. Normalize the vector |vi = to make it a unit vector.
−2
b
√
Solution: Since kvk = 12 + (−2)2 = 5 we can scale it by c = √15 to get the unit vector
p

   √ 
1 1 1 1/ √5
√ |vi =
b √ = .
5 5 −2 −2/ 5

21 Q C a m p 2 0 2 3
C H A P T E R 3 . L IN E A R A L G E B R A

Born’s Interpretation

Max Born, a German physicist, suggested that the wavefunction ψ itself has no physical
meaning. However, the modulus square of the inner product between the wavefunction and
the position vector | hψ|xi |2 is directly proportional to the probability of finding a particle
at position x.

Since the probability of finding the particle throughout the entire space is 1 (assuming the
particle does exists), then the norm of the wavefunction kψk has to be 1. The wavefunction
is said to be normalized.

L in e a r C o m b in a t io n , In d e p e n d e n c e a n d B a s is

Linear Combination

A vector |vi is said to be a linear combination of a set of vectors if the vector |vi can be
decomposed as a linear sum of vectors (these vectors may be scaled by numbers) that belong
to this set. For example, for vectors |v1 i , |v2 i , . . . , |vn i ∈ Cm and numbers c1 , c2 , . . . , cn ∈
C; a linear combination looks like

c1 |v1 i + c2 |v2 i + · · · + cn |vn i .

Linear Independence

A set of vectors is linearly independent if any vector of this set cannot be written as a linear
combination of the remaining vectors in this set. Conversely, the set is linearly dependent
if a vector in this set can be written as a linear combination of the remaining vectors in
this set.

     
3 1 0
Example 3.7. Consider the vectors |vi = , |ai = and |bi = such that |vi =3 |ai +5 |bi.
5 0 1
If we consider the set S = {|vi , |ai , |bi}, then S is linearly dependent since |vi can be written as a
linear combination of |ai and |bi. The mathematical way to show that set S is linearly dependent
is to show that there exists a non-zero solution to the equation c1 |ai + c2 |bi + c3 |vi = 0 which is
c1 =3, c2 =5, c3 =−1. On the other hand, set T = {|ai , |bi} is linear independent since |ai cannot be
written in terms of |bi. Mathematically, the linear independence of set T can be shown by showing
that the only solution to the equation c1 |ai +c2 |bi = 0 is c1 =c2 =0.

22 Q C a m p 2 0 2 3
C H A P T E R 3 . L IN E A R A L G E B R A

Example 3.8. Are the vectors

      
 1 1 3 
|v1 i =
b 1 , |v2 i =
b 0 , |v3 i =
b 2 .
1 1 3
 

linearly dependent or linearly independent?

Solution: They are linearly dependent since we can write |v3 i = 2 |v1 i+|v2 i as a linear combination
of |v1 i and |v2 i.

Basis

A basis of Cn is a set of n linearly independent vectors in Cn , {|v1 i , . . . , |vn i}. Any vector
in Cn can be written as a linear combination of vectors in the basis. That is, for any vector
|vi in Cn there exists n complex numbers c1 , . . . , cn ∈ C such that

|vi = c1 |v1 i + c2 |v2 i + · · · + cn |vn i .

The same holds for Rn too, but then we only use real numbers and real vectors, that is,
vectors in Rn .

A simple basis of R2 , called the standard basis consists of the two elements ê1 = [ 10 ] and ê2 = [ 01 ],
since, any vector |vi =
b [ ab ] of R2 may be uniquely written as |vi =
b aê1 + bê2 . Recall that while
introducing vectors we learned that the column of numbers is a representation for the vector |vi
in a particular basis. This becomes clear now since we can represent the same vector |vi in a
different basis giving a different column of numbers i.e. consider a different basis with elements
ê3 = (1, 1) and ê4 = (1, −1), then |vi =
b a0 ê3 + b0 ê4 with a0 = a+b
2 and b0 = 2 .
a−b
In this new basis
representation: |vi =
b (a , b ). 0 0

Example 3.9. Can I write any vector in C3 as a linear combination of these three vectors
      
 1 1 3 
|v1 i =
b 1 , |v2 i =
b 0 , |v3 i =b 2 ?
1 1 3
 

Solution: No. There are 3 vectors but they are not linearly independent. Thus they do not form
a basis.

23 Q C a m p 2 0 2 3
C H A P T E R 3 . L IN E A R A L G E B R A

Orthonormal Basis

We say a basis is orthonormal if each element of the basis has norm 1 and each pair of
vectors are orthogonal. We can equivalently define an orthonormal basis as any set of n
vectors {|v1 i , . . . , |vn i}, each in Cn (or in Rn ), where:

1 if i = j,

hvi |vj i =
0 if i 6= j.

Example 3.10. Are the following vectors an orthonormal basis

  √   √ 
1/√2 1/ √2
|w1 i = , |w2 i = ?
−1/ 2
b b
1/ 2

Solution: Yes, as hw1 |w1 i = hw2 |w2 i =1 and hw1 |w2 i =0, both have norm 1 and are orthogonal.

M a t r ic e s

Matrix

A matrix is represented by a box of numbers with respect to a particular basis, for example
 
a b
M = .
c d
b

This is a matrix with two rows and two columns. These map vectors to vectors. For
example, given the vector  
x
|vi = ,
y
b

we have that M maps |vi to

    
a b x ax + by
M |vi = = .
c d y cx + dy
b

In general, a matrix with n rows and m columns maps a vector in Cm to a vector in Cn .

The rule for matrix-vector multiplication in other dimensions can be extended from the
above definition.

Complex conjugate of a matrix M, denoted by M† , can be similarly defined as complex

conjugate of a vector (3.1)

a∗ c∗
 
M = ∗
†
.
b d∗

24 Q C a m p 2 0 2 3
C H A P T E R 3 . L IN E A R A L G E B R A

Example 3.11. Compute M |vi where

  

1 3 1
M= and |vi = .
4 7 −4
b b

Solution:       
1 3 1 1 − 12 −11
M |vi = = = .
4 7 −4 4 − 28 −24
b

Unitary Matrices

A matrix U is unitary if and only if U† U = UU† = I, where I is the identity matrix (matrix
with ones in its diagonal and zero everywhere else).

An important property of a U is that it conserves the inner product of 2 vectors i.e. if

|v10 i = U |v1 i and |v20 i = U |v2 i

hv10 |v20 i = hv1 | U† U |v2 i = hv1 |v2 i

for all vectors |v1 i and |v2 i. Moreover, the columns and rows of the unitary matrix form
two different orthonormal bases B1 and B2 respectively.

Unitary matrices preserve the norm (or length) of vectors. This is important in Quantum Mechanics
since we want our wavefunctions to always be normalized.
q
kUvk = hv| U† U |vi = hv|vi = kvk.
p

The following matrices are unitary (in fact M2 , M3 , M4 are known as the Pauli spin matrices)
       
1 0 0 1 0 −i 1 0
M1 = M2 = M3 = M4 = .
0 1 1 0 i 0 0 −1
b b b b

25 Q C a m p 2 0 2 3
C H A P T E R 3 . L IN E A R A L G E B R A

K ro n e c k e r P ro d u c t

Kronecker Product

The Kronecker product is a vector operation to construct bigger vectors. The Kronecker
product is defined as  
    v1 w1
v w v1 w2 
b 1 ⊗ 1 =
|vi ⊗ |wi = v2 w1  .

v2 w2
v2 w2
The Kronecker product operation can be extended to matrices. Consider the Kronecker
product of two general matrices:
     
e f e f

ae af be bf
    a b
a b e f  g h g h  ag ah bg bh 
⊗ =
 e f    =  
c d g h e f   ce cf de df 
c d
g h g h cg ch dg dh

As we going to see later, the Kronecker product plays an important role in the representation
of multiple quantum systems that are coupled to one another.

Example 3.12. Compute |vi ⊗ |wi where

   
2 1
|vi = and |wi = .
−3 2
b b

Solution: Taking the Kronecker product yields

 
    2
2 1 4
|vi ⊗ |wi = ⊗ =
−3 .

−3 2
b
−6

Example 3.13. Compute Z ⊗ I where

   
1 0 1 0
Z = and I =
0 −1 0 1
b b

Solution: Taking the Kronecker product yields

 
1 0 0 0
0 1 0 0
Z⊗I = b 
0 0 −1 0
0 0 0 −1

26 Q C a m p 2 0 2 3
C H A P T E R 3 . L IN E A R A L G E B R A

3 .1 E x e r c is e s

Consider the vectors

     √   √ 
1 0 1/√2 1/ √2
|v1 i := , |v2 i := , |v3 i := , |v4 i :=
0 1 1/ 2 −1/ 2

and the matrices

       
1 0 0 1 1 0 0 −i
I := , X := , Z := , Y := ,
0 1 1 0 0 −1 i 0
 √ √ 
1/√2 1/ √2
H := .
1/ 2 −1/ 2

Exercise 3.1. Carry out the following vector calculations to give the explicit column vector
representation
(a) |v1 i + |v1 i (b) |v1 i + |v4 i (c) |v3 i + |v1 i

(d) 2 |v1 i (e) − π |v4 i (f ) 2i |v2 i

Compare your answer for part (a) and part (d). What do you notice?

Exercise 3.2. Carry out the following matrix-vector multiplications

(a) I |v1 i (b) X |v2 i

(c) Z |v3 i (d) Y |v4 i

Exercise 3.3. Define |v5 i := Y |v4 i. Calculate the following inner products and norms

(a) hv1 |v1 i (b) hv1 |v4 i (c) hv3 |v1 i

(d) hv1 |v5 i (e) hv5 |v1 i (f) hv5 |v5 i
(g) Hv1 (h) Hv1 + Hv1 (i) Hv1 + Hv2

(j) Are there any orthogonal pairs above?

(k) Consider the vectors H |v1 i, H |v1 i + H |v1 i, and H |v1 i + H |v2 i. Which vector is the longest?
Which vector is the shortest?

Exercise 3.4. Which of the following sets of vectors are orthonormal bases for R2 ?

(a) {|v3 i , |v4 i} (b) {|v1 i , |v1 i} (c) {|v3 i , |v2 i} (d) {|v3 i , |v4 i , |v3 i}

27 Q C a m p 2 0 2 3
C H A P T E R 3 . L IN E A R A L G E B R A

Exercise 3.5. Which of the following sets of vectors are orthonormal bases for R3 ?
         
 1 1 2   1 1 
(a) 1 , 0 , 0 (b) 1 , 0
1 0 0 1 0
   

Exercise 3.6. Define the following matrices

     
1 0 2 0 3 0
A1 = , A2 = , A3 = .
0 1 0 2 0 3

Calculate the following:

(a) A1 |v3 i (b) A2 |v1 i (c) A3 |v4 i

What do you notice?

Exercise 3.7. Calculate the following.

(a) I |v4 i (b) X |v3 i (c) Z |v2 i

What do you notice?

Exercise 3.8. We denote the following quantum states as vectors:

     √   √ 
1 0 1/√2 1/ √2
|0i := , |1i := , |+i := , |−i := .
0 1 1/ 2 −1/ 2

Carry out the following vector calculations:

(a) |0i ⊗ |0i (b) |0i ⊗ |1i

(c) |1i ⊗ |0i (d) |1i ⊗ |1i

(e) |+i ⊗ |+i (f) |−i ⊗ |−i

(g) (|+i ⊗ |+i ⊗ |+i ⊗ |+i ⊗ |1i) · (|+i ⊗ |1i ⊗ |+i ⊗ |1i ⊗ |0i)

28 Q C a m p 2 0 2 3
C H A P T E R 3 . L IN E A R A L G E B R A

For the interested reader:

The mathematics of quantum mechanics has a connection with the field of abstract algebra in
mathematics which includes rigorously defining spaces in which vectors “live”. We define a vector
space and a vector in this context as given below here:

V e c to r S p a c e s

Vector Space

A vector space is a collection of vectors that fulfills a certain list of criteria:

• it must contain the zero vector.

• (closed under summation) the sum of vectors within the vector space is in the vector
space.

• (closed under scalar multiplication) any vector resulting from the multiplication of a
vector in the vector space by a scalar is in the vector space.

Example: The 2-Dimensional version of a vector space is a plane.

−5
z 0
5

−5

|v 1⟩ 0 x
⟩ 1⟩
|v 2 −|v
⟩
|v 2 5

5 0 −5
y

Figure 3.1: 2D Vector Space - Plane

Geometrically, a two-dimensional vector space represents a plane. In Figure 3.1, we plot the plane
5x − 4y + z = 0 together with its normal vector (−5, 4, −1)> (grey). This plane is spanned by two

29 Q C a m p 2 0 2 3
C H A P T E R 3 . L IN E A R A L G E B R A

vectors |v1 i =
b (1, 2, 3)> (green) and |v2 i =
b (4, 5, 0)> (orange). The blue vector is obtained from
|v2 i − |v1 i, which is a linear combination of the vectors in this vector space (represented by the
plane), clearly lies in the plane. However, the red vector (1, −5, 3)> is not a linear combination of
|v1 i and |v2 i, hence it does not lie in the plane.

E ig e n v e c t o r s

Eigenvectors

A matrix A has an eigenvector |vi with a corresponding eigenvalue λ if and only if the
following equation is true:
A |vi = λ |vi

If |vi is not an eigenvector, then A |vi is not parallel to |vi.

Finding the Eigenvalues and Eigenvectors

Suppose we have the eigenvector equation A |vi = λ |vi and you wish to find all the eigen-
vectors of A and their corresponding eigenvalues. For this linear system to have non-trivial
solutions (i.e: non-zero λ), the determinant of A − λI must be zero where I is the identity
matrix.

The determinant of a general 2 by 2 matrix is

a b
det(M ) = = ad − bc
c d

For matrices of larger dimensions, the method to compute the determinant will be more
complicated and will not be covered here. However, readers are encouraged to pursue more.

Example 3.14. Consider the matrix:

 
3 1
A =
0 2
b

The eigenvectors of A are given: |a1 i = 1

and |a2 i =
b [ 10 ]. In Figure 3.2, the eigenvectors
 
b −1

|a1 i and |a2 i are coloured green and orange respectively. The resultant vectors after multiplying
matrix A to the eigenvectors are A |a1 i and A |a2 i respectively. Observe that A |ai i is parallel to
|ai i (where i = 1 or i = 2) such that A |ai i is obtained from a scalar multiplication of |ai i. This
scalar multiple is known as the corresponding eigenvalue.

30 Q C a m p 2 0 2 3
C H A P T E R 3 . L IN E A R A L G E B R A

Before transformation A After transformation A

|v⟩ A|v⟩
|a2 ⟩
A|a
|a 1

2⟩
⟩

A|
a 1⟩
Figure 3.2: Eigenvectors

To show the effect of multiplying A with an arbitrary vector |vi that is not an eigenvector, we will
b [ 12 ] and illustrate it with a blue vector. The resultant vector A |vi is clearly not parallel
let |vi =
to |vi.

Exercise 3.9. Find the eigenvalues and eigenvectors for the following matrix:
 
5 −1
B=
2 3

Note that the eigenvalues are expected to be complex numbers and the corresponding eigenvectors
will likely have complex entries.

Application of Eigenvectors to Quantum Mechanics

In Quantum Mechanics, we represent a physical observable, such as position, momentum,

and energy by a matrix (or more generally, an operator). The eigenvectors of the matrix
represents the possible physical states of the system while their corresponding eigenvalues
represents the observed values of these states.

31 Q C a m p 2 0 2 3
Tuesday

32
C h a p te r 4

Q u b it s & S u p e r p o s it io n

Lecture notes by Aw Cenxin Clive

Revised by Peter Sidajaya

4 .1 In t r o d u c t io n

It doesn’t take long for readers of popular science and science fiction to encounter really touch-
and-go explanations of “entanglement”, “superposition” and “uncertainty principle”. Throw in a
couple of over-simplifications of quantum “paradoxes” like Schrödinger’s Cat and we often end up
with a perfect storm of what might be called “quantum fake news”: viral misinformation about
quantum theory and its implications. If we want to avoid this, it is crucial to go beyond one-liner
takes on quantum phenomena. In other words, understanding and using quantum mechanics well
means we need to do our homework.

This is precisely what we will kickstart in this chapter, building from the bottom up by examining
the simplest case of quantum phenomena: the humble qubit.

Ta k e a w a y s

By the end of this chapter, the reader should:

• Gain first intuitions about what it means to say some given quantum object is a qubit and
what it means for qubits to be in superposition.

• Know how to perform “Born Rule” probability calculations, a foundational principle in quan-
tum theory.

• Gain first intuitions about quantum operations acting on a qubit

33
C H A P T E R 4 . Q U B IT S & S U P E R P O S IT IO N

• Know some big differences between quantum and classical physics.

In this chapter, we will stick to the most basic category of qubit states: pure states. These are
states perfectly constructed and known. So when we speak of qubits here, we mean specifically
pure qubits. We will also stick to focusing on single-party states. These are states that involve
only one system, and so entanglement is impossible. We will explore two-party states in Chapter
4.

4 .2 T h e Q u b it S t a t e

4 .2 .1 T h e B a s is o f a Q u b it

The qubit (standing in for quantum bit) is the simplest case because it is a quantum system with
a “two-levelled degree of freedom.” This is sometimes expressed as a “ket” |ψi:

|ψi ∈ C2

Which reads that the quantum state (|ψi) lives in (∈) a complex space1 of two levels/dimensions
(C2 ). This two-level nature of qubits brings us to the computational basis:

|0i , |1i

These are the two possible outcomes if we make a measurement “along the 0/1 basis”. We will get
to what exactly this means in a bit. The main point is that given a certain kind of measurement,
a qubit can only give one of two outcomes specific to that kind of measurement.2 In this way,
you can think of a qubit as a quantum object that can only give one of two possible “answers”
(outcomes) for a given “question” (measurement). Note that every ket (in a complex space like
C2 ) can be represented by vectors. For instance |0i and |1i can be represented as vectors as such:
   
1 0
|0i = , |1i = (4.1)
0 1
b b

Exercise 4.1. How do we know the above vector representation gives a valid orthonormal basis
for C2 ?

|0i and |1i are examples of qubit states. However, to get more interesting phenomena, we will need
to discuss superpositions.

1 Called a Hilbert Space.

2 Notice that a thing with just one level in a degree of freedom would not very interesting! Whenever you make
a measurement it will always give you the same result no matter how much you do stuff to it or let it evolve through
time. Indeed, it should be plain that a one-levelled degree of freedom is a really a contradiction in terms.

34 Q C a m p 2 0 2 3
C H A P T E R 4 . Q U B IT S & S U P E R P O S IT IO N

The Significance of Qubits in Quantum Science

It might surprise some that the qubit’s homely and simple character is also what makes
it indispensable in both understanding quantum theory and in forming the backbone of
most technologies of the quantum frontier.

There are three big reasons for this: (1) By understanding the simple qubit, many of
the counter-intuitive ideas in quantum theory can be seen clearly in their most distilled
form, which can be recontextualized in higher dimensional systems (such as a qutrit).
(2) Its simplicity and kindred connection with the classical bit makes it a cornerstone
of quantum information, quantum communication and quantum computation. (3) There
are also very important phenomena (such as the polarization of light and the intrinsic
magnetic “spin” of certain particles) that have a qubit nature. All this to say that looking
at the qubit is crucial for everyone interested in quantum science and technologies.

4 .2 .2 W r it in g Q u b it s a s G e n e r a l S u p e r p o s it io n s

We have already met the two members of the computational basis (|0i , |1i). We now meet two
important superpositions of these states:
|0i + |1i |0i − |1i
√ , |−i :=|+i :=√
2 2
These are equally weighted superpositions of |0i and |1i. We would find a |0i state half the time we
measured such a state (along the computational basis), and |1i otherwise. If we adopt the vector
representation from (4.1), we have:
   
1 1 1 1
b √
|+i = , b √
|−i = (4.2)
2 1 2 −1

Exercise 4.2. Verify that this is the correct vector representation for |+i and |−i. Verify also
that they form an orthonormal basis.

States of the |+i, |−i basis are definitely not the only kind of qubit superposition we can speak
of. The general superposition of the computational basis is not equally weighted. This is how it
can be expressed:
s.t. |a| + |b| = 1 (4.3)
2 2
|ψi := a |0i + b |1i ,

This leads us to a very important point: each unique assignment of the complex numbers a and b
give us a unique qubit |ψi.3 Any (pure) state that is a qubit can be written in this (4.3) form (no

3 Up to a “phase”. This will be touched on in the next chapter.

35 Q C a m p 2 0 2 3
C H A P T E R 4 . Q U B IT S & S U P E R P O S IT IO N

matter what this qubit may physically represent). Notice also that given the constraint on a and
b any |ψi lies on the circle, never going below or beyond it. Following our choice in (4.1), we have:
 
a
, s.t. |a| + |b| = 1
2 2
|ψi =
b
b

O n ly s u p e r p o s it io n s o f b a s e s .

Moving on, let’s have one last clarifier. After what we have learnt so far, could we say that “|+i is
a superposition. Meanwhile, |0i is not.”? Well, we can’t. This is because there are no such things
as “superpositions” unless we say speak of “superpositions of” a given bases (of which, clearly,
there must be at least two). Notice then, that just as |+i and |−i are a superpositions of |0i and
|1i, |0i and |1i are superpositions of |+i and |−i as well!

|+i + |−i |+i − |−i

|0i = √ , |1i = √
2 2

Exercise 4.3. Verify the above.

4 .3 M e a s u re m e n t o n a Q u b it

4 .3 .1 T h e B o rn R u le

The |a|2 + |b|2 = 1 constraint alludes to one of the most fundamental principles in quantum theory:
The Born Rule. This principle gives us the statistics of measurement results given quantum
systems. To understand this, first notice how:
2 2
|hψ|0i| = |(a∗ h0| + b∗ h1|) |0i|
2
= |a∗ h0|0i + b∗ h1|0i|
2
= |a∗ + b∗ (0)| = a2
2
|hψ|1i| = b2

What are the significance of these results? Well for such |ψi, |a|2 is the probability of getting a |0i
outcome from a measurement of |ψi. |b|2 is similarly understood as that of a |1i outcome. This is
why a and b are called “probability amplitudes” - complex numbers that can help us produce the
statistics of possible measurement outcomes for quantum states.

This applies more generally as follows: the statistics for outcomes for quantum states are given
by the square of the inner product of the ket of the wavefunction representing the state |ψi and

36 Q C a m p 2 0 2 3
C H A P T E R 4 . Q U B IT S & S U P E R P O S IT IO N

the ket of the target outcomes (say, |0i , |1i , |+i , |−i). Let’s say we prepare a quantum system in
state |ψi and want to see how probable a measurement would produce the outcome |φi. Then this
probability is given by:

P(φ|ψ) := |hφ|ψi| (4.4)

2

Where hφ|ψi is the dot product of |φi and |ψi. Having said all this, it becomes clear why “|a|2 +
|b|2 = 1” must hold for qubits. It’s just due to “normalization”: the sum of the probabilities for
all possible outcomes must always be 1.4 For qubits, this can be expressed as such:

P(0|ψ) + P(1|ψ) = 1 (4.5)

Naturally, this applies to the |+i, |−i basis too: P(+|ψ) + P(−|ψ) = 1

Some Examples

Let’s do some examples here. Consider the calculations for P(+|0) and P(−|1).

 1 2 2
 
1  1 1
P(+|0) = |h+|0i| = √ 1∗
2
1∗ = √ =
2 0 2 2
 2 2
1   0 −1 1
P(−|1) = |h+|0i| = √ 1∗
2
−1∗ = √ =
2 1 2 2

After this, it is obvious what P(−|0) and P(+|1) are. We use the principle of normalization
to get the following:

1 1
P(−|0) = 1 − P(+|0) = 1 − =
2 2
1 1
P(+|1) = 1 − P(−|1) = 1 − =
2 2

This is why |+i and |−i can be called unbiased superpositions of |0i and |1i. Neither
outcome (|0i , |1i) is more likely than the other.

4 This is also sometimes called the law of total probability.

37 Q C a m p 2 0 2 3
C H A P T E R 4 . Q U B IT S & S U P E R P O S IT IO N

Exercise 4.4. Consider another arbitrary qubit

|φi = c |0i + d |1i

Verify that P(φ|ψ) has no imaginary component. Why would it be a problem if P(φ|ψ) was
complex?

Before we finish off this section, there are some notes that we must take into account.

4 .3 .2 P h a s e s

Consider the states |+i at − |+i. Are they different? In terms of their matrix representations, they
differ only by a minus sign in all of the entries. You can verify that the probability of measuring
any state |ψi would be the same for both. Therefore, physically, |+i and − |+i are really the same
state. We call the minus sign in front a global phase, and global phases do not really matter. In
general, |ψi and eiθ |ψi represent the same state.

The states |+i and |−i, on the other hand, also only differs by a minus sign, but only in one of its
entry in the matrix representation. However, it is clear that the probability of measuring |+i from
|+i and |+i from |−i are different (1 and 0, respectively). We call the minus sign in this case a
relative phase, and relative phases do actually represent something physical.

Exercise 4.5. Verify that the probability of measuring |ψi from eiθ |+i are the same for all θ.

4 .3 .3 T h e B a s is o f a M e a s u re m e n t

When we perform a measurement, we must perform it in some orthonormal basis. For example, we
can perform a measurement on a spin qubit in the z axis, which is equivalent to saying measuring
on the {|0i , |1i} basis, using conventional notation. This means that we will get either 0 or 1 as
a result. We can also instead measure it in the x axis, which is the {|+i , |−i} basis. However,
we can’t measure it in the {|+i , |0i} basis. A simple calculation shows that using the Born rule,
P (+|ψ) + P (0|ψ) > 1 for any |ψi, which violates normalisation of probability.

A corollary of this statement is that there is no way to perfectly discriminate two non orthogonal
states. Suppose we have a state |ψi that we know to be in either |+i or |0i. If we measure it in
the {|0i , |1i} basis, we will always get 0 if it is indeed in the state |0i, but there is a 50% chance
of getting a false positive if the state is actually in |+i. The same is true if we measure it in the
{|+i , |−i} basis instead.

38 Q C a m p 2 0 2 3
C H A P T E R 4 . Q U B IT S & S U P E R P O S IT IO N

4 .3 .4 M e a s u re m e n t c h a n g e s a S ta te

One of the more peculiar phenomenon in quantum mechanics that you might’ve heard is how
measurement change the state of the system being observed. Mathematically, what this means is
that once we measure |ψi in some basis {|ii}i and found it to be in, say, |ji, the state of the system
is now definitively |ji. If we then measure it in the same basis again, the probability of getting |ki
is given by
P (k|j) = δkj ,

as to be expected from an orthonormal basis.

4 .4 S o m e D is c u s s io n

Before we go on deeper and deeper into the math, let’s first tackle some concerns about qubits
that may have already popped up.

4 .4 .1 U n d e r s t a n d in g S u p e r p o s it io n

“Bits = Yes or No. Qubits = Yes or No or Maybe?” Well, not really.

This notion (that qubits can only give one of two outcomes given a measurement) could
trip up some of us who have heard of this catchy popular science saying that quantum
bits are different from classical bits because they can answer not only “yes” or “no” but
also “maybe”. This seems to be pointing to the idea of superposition. If so, it is well
taken, but only with a (sizeable) grain of salt. Later we will see more closely why this is
probably not the best way of putting it.

For now, just note that the main issue is that the analogy of a qubit answering seems
to stands in for a measurement. This seems to imply that a qubit can give one of three
outcomes. This is false. Given a measurement, there are only two outcomes a qubit can
“choose” from, not three (as indicated by C2 . If it were three it would be a “qutrit”,
where |τ i ∈ C3 ).

Most of us already know that classical bits, of everyday personal computers, also only give one
of two possible outcomes (namely, “0” or “1”). Well, if this is the case, then how are qubits any
different from a classical bits? The answer for this needs to be explained in steps. Firstly, it is in
how the outcomes are connected to the kinds of measurements that can be performed. Secondly, it
is in how these outcomes are determined (or rather not determined), given these measurements. It

39 Q C a m p 2 0 2 3
C H A P T E R 4 . Q U B IT S & S U P E R P O S IT IO N

is in how the outcomes are given, not how many outcomes can be given. It’s about what is “under
the hood” for these outcomes. For a proper understanding of what this means we must turn to
the very important concept of superposition in quantum mechanics.

Superpositions do not represent randomness due to ignorance.

When I hear about Schrodinger’s cat, I reach for my gun.

- Stephen Hawking

Now, some conceptual clarifiers about superposition before moving forward. Sometimes states like
|+i and |−i are likened to a coin toss. Half of the time we get heads and the other half of the
time we get tails. But this analogy is only good up to a point. Coin tosses are not quantum
superpositions. This is because the randomness of a superposition and the “randomness” of a coin
toss are very different. Coin tosses seem random to us because we do not know how the coin
was flipped and all the various factors that would affect its trajectory and its upward face. Such
apparent randomness is really randomness due to ignorance, from lack of knowledge. If we were
an expert physicist armed with a supercomputer and all the information relevant to the coin toss,
the process would be shown as completely deterministic to us. We would know the outcome once
the coin is tossed and win every bet thereafter!

Your Crush is not a Qubit

Likewise, consider the analogy of saying that superpositions are like not knowing whether
your crush likes you or not. You have to ask to find out. While this is cute, this is not
a good parallel for the same reason: whether your crush will say they like you or not is
predetermined before measurement.a This is not the case for quantum superpositions.
These superpositions are pure states: under quantum theory, they are the most pristine
description of system - there is nothing more to know. And so the randomness is not
from ignorance (as with the crush situation) but intrinsic to the universe.

a Though, it is true that a very badly executed “measurement” could perturb your crush’s “state” so
much that you get a unfavourable outcome.

The expert physicist, however, cannot gamble against the quantum randomness of the humble
qubit. The randomness we find in quantum superposition is not randomness by ignorance at all.
It is genuine, intrinsic randomness. Here, there are no results pre-established by initial conditions
and classical laws. Under quantum mechanics, we have actual indeterminism in the universe.

40 Q C a m p 2 0 2 3
C H A P T E R 4 . Q U B IT S & S U P E R P O S IT IO N

This was what the Schrödinger’s Cat thought experiment tries to illustrate. In this thought
experiment, we put a cat into a large sealed box that contains a qubit system and a device.
Preparing the qubit in the |+i state, we connect it to the device, such that if the device measures
the system to be in |0i state it would dispense a delicious cat treat. Meanwhile if the device
measures the system to be |1i it would dispense a poisonous gas, which kills the cat. It can be
described as follows:

|0i ⊗ |No poisoni ⊗ |Alivei + |1i ⊗ |Poisoni ⊗ |Deadi

|Cati = √
2
Why Schrodinger came up with this thought experiment is not to say that we will not know if
the cat is dead or alive until we open the box. This happens classically (imagine a coin toss that
determines the cat being dead or alive). In a classical setting we also don’t know if the cat is dead
or alive until we open the box but, note, it had always been alive or dead before that. The result
was predetermined prior to measurement.

Rather, with this illustration, Schrödinger was raising the the bizarre conclusion that since we
are dealing with a quantum scenario (no coin tosses but a qubit in a relevant superposition), the
randomness is genuine and therefore, prior to measurement, the cat was neither dead nor alive but
a superposition of both.

But how can life and death be put into superposition? This is the absurd conclusion that
Schrödinger was getting us to consider to see the wild implications of the indeterminism found
in quantum theory.

What if we put a spin on it?

Notice then that the spinning coin illustration sometimes used by science communicators
are only helpful to a point. While it at least tries to acknowledge that the state wasn’t
originally |0i or |1i originally before measurement (since the coin keeps spinning), it’s not
clear what good parallel there is for coins being in a “heads” state being a superposition
of spinning coins too. So whether we are talking about “maybe” answers, coins tossed,
coins spun or your crush, it should be obvious by now that there are no classical parallels
to even the simplest of quantum system.

4 .4 .2 T h e M e a s u r e m e n t P r o b le m

Other than superposition, another quirk of quantum mechanics that we have encountered is the
mechanism of measurements. In our day-to-day classical world, measurement does not change a
state. A table won’t get longer just because you measure it with a ruler. However, the same does

41 Q C a m p 2 0 2 3
C H A P T E R 4 . Q U B IT S & S U P E R P O S IT IO N

not apply in quantum theory. As mentioned, when we measure a system, we change the state of
the system. Similar to the discussion on superposition, it must be noted that this change is an
intrinsic property of quantum theory, and not due to limitations on our measurement apparatus or
our ignorance. No matter how precise or how careful we are with our measurement, we will always
change the state of the system when we do a measurement. This phenomenon is sometimes called
the wavefunction collapse.

The wavefunction here refers to the coefficients of the basis states. Suppose before a measurement
we have the state
|ψi = a |0i + b |1i .

Then, we measure it in the measurement basis and get 0. The state after the measurement is then

|ψi = |0i .

Since the coefficients of |1i have vanished, we say that the wavefunction has collapsed to |0i. The
question of what causes the collapse of the wavefunction, or whether the wavefunction actually
collapses5 is known as the measurement problem. The answer to this question really depends on
one’s interpretation and understanding of quantum theory, and is way beyond the scope of the
lecture, but suffice it to say, this is a question that will get you different answers from different
quantum physicists.

4 .5 Q u a n tu m O p e ra to rs

So far we have been finding the kind of statistics quantum states will give when we check if they
belong to one state or another. But what if, before measurement, we want to perform some trans-
formations on the quantum system? That we want to describe the changes applied on or happening
to quantum system over time? Or what if we are not just checking a state but also some property
the state has? These are where “operators” must come in. We will look at the two most physically
important classes of operators: (1) “unitary operators” and (2) “Hermitian operators”. The former
correspond to valid evolutions in quantum states. Unitary operators transform quantum states
without making them lose their quantum character. Meanwhile, Hermitian operators correspond
to “observables”. That is, they represent measurements of a property (eg. like its energy) with
respect to the system.

But before we get down to looking at these two classes, let’s first introduce some math. Just
as every quantum ket |ψi can be represented with a vector, every quantum operator O can be

5 In some interpretations, the collapse of the wavefunction (or even the wavefunction itself) is just an approxi-
mation or redundant. The discussion of whether the collapse of the wavefunction (and the wavefunction itself) is
something ontically physical or not is beyond the scope of this lecture.

42 Q C a m p 2 0 2 3
C H A P T E R 4 . Q U B IT S & S U P E R P O S IT IO N

represented with a matrix.6 For the qubit space C2 , this would be a 2 by 2 matrix, like this:
 
a b
O= , s.t. a, b, c, d ∈ C
c d

There’s also the “hermitian conjugate” O† of the operator O. This is just the transpose and
complex conjugate of O. The superscript of O here is referred to as “dagger”†.
T ∗  ∗  ∗
c∗
 
T ∗ a b a c a
O = (O)
†


= = = ∗
c d b d b d∗

4 .5 .1 U n it a r y O p e ra to rs

Now, what makes an operator unitary U? Well, it must “preserve the norm” of the states it acts
on. The mathematical meaning of this can be seen as follows. A operator is unitary if and only if:

When
|Uψi = U |ψi , (a unitary operator U acts on a quantum state |ψi)
|Uφi = U |φi , (a unitary operator U acts on a quantum state |ψi)

Then
hUψ|Uφi = hψ| U† U |φi
:= hψ|φi

Note then that this simply means that an operator U is unitary if and only if:

U† U = I (4.6)

Where I is the “identity” operator. For qubits,

 
1 0
I= (4.7)
0 1

With this in mind, let’s introduce some very important examples. The first three would be the
“Pauli operators”

6 That is for discrete spaces like the qubit space.

43 Q C a m p 2 0 2 3
C H A P T E R 4 . Q U B IT S & S U P E R P O S IT IO N

     
0 1 0 −i 1 0
X = , Y = , Z = (4.8)
1 0 i 0 0 −1
b b b

Then, the “Hadamard” operator:

 
1 1
H = √1 (4.9)
1 −1
b 2

Finally, this particular operator (more on this in QM5: The Schrödinger Equation):

 −iEt/h̄ 
e 0
UE = , s.t. E, t, h̄ ∈ R (4.10)
0 eiEt/h̄
b

We can verify that all of these are unitary. Here we just do this for Y and H, while the rest are
quite simple and can be done as an exercise.

 †  T ∗  ∗  
0 −i 0 −i 0 i 0 −i
Y = †
= = =
i 0 i 0 −i 0 i 0
      
0 −i 0 −i 0 · 0 + (−i) · i 0 · (−i) + (−i) · 0 1 0
Y† Y = = = =I
i 0 i 0 i·0+0·i i · (−i) + 0 · 0 0 1
 †  
1 1 1 1 1 1
H† = √ =√
2 1 −1 2 1 −1
    
1 1 1 1 1 1 2 2
H H=
†
= =I
2 1 −1 1 −1 2 2 −2

Exercise 4.6. Verify that I, X, Z and UE are all unitary.

4 .5 .2 O p e ra to rs o n Q u b it S t a t e s

Let’s apply some of these unitary operators on qubit states! Consider the following examples.

       
1 0 1 1 1 1·1+0·1 1 1
I |+i = √ √ =√ = |+i
0 1 2 1 2 0·1+1·1 2 1
b
       
1 0 1 1 1 1·1+0·1 1 1
Z |+i = √ √ =√ = |−i
0 −1 2 1 2 0 · 1 + −1 · 1 2 −1
b

44 Q C a m p 2 0 2 3
C H A P T E R 4 . Q U B IT S & S U P E R P O S IT IO N

    
0 1 1 0
X |0i = = = |1i
1 0 0 1
b
     
0 1 1 1 1 −1
X |−i = √ =√ = − |−i
1 0 2 −1 2 1
b

    
1 1 1 1 1 1
H |0i =
b √ =√ = |+i
2 1 −1 0 2 1
     
1 1 1 1 1 1 2
H |+i =
b √ √ = = |0i
2 1 −1 2 1 2 0

We can bring this together with what we know from the Born Rule. If we measure a |0i state
for |1i, we know we will never get a positive result because according to the Born Rule, we have
P(1|0) = 0. But if we input the |0i through a X transformation, we get a very different result:

P(1|X on 0) = |h1| (X |0i)|

2

2
= |h1|1i| = 1

X acts like a kind of rotation in the Hilbert space, that sends 0 to 1. From this we have a very
small glimpse of how unitary operators can be used for quantum information and communication:
evolving one state to another. But what this all physically looks like will have to wait to later
chapters. Having covered evolutions, we now look at so-called observables.

4 .5 .3 H e r m it ia n O p e ra to rs

Hermitian operators are our second and final class of important quantum operators. These cor-
respond to measurements of state (like those discussed previously with respect to the Born Rule)
and, more broadly, measurements of properties that the states may have. The key definition for a
Hermitian operator is the following: an operator A is Hermitiian if and only if:

A† = A (4.11)

Importantly, we also introduce the concept of an “expectation value” hAiψ . This translates to the
average of the outcomes of performing (a very large number of) measurements of A upon some
state |ψi:

hAiψ = hψ| A |ψi (4.12)

Take for instance this operator:  

h̄ 1 0
Sz = (4.13)
2 0 −1
b

45 Q C a m p 2 0 2 3
C H A P T E R 4 . Q U B IT S & S U P E R P O S IT IO N

Notice how it is Hermitian, since h̄ is a real number.

 T !∗
h̄ 1 0
S†z =
2 0 −1
h̄∗ 1 0
   
h̄ 1 0
= = = Sz
2 0 −1 2 0 −1

We can calculate the expectation value of Sz for |0i and |1i in the following way:
  
h̄   1 0 1 h̄
hSz i0 = h0| Sz |0i = 1 0 = (4.14)
2 0 −1 0 2
  
h̄ 1 0 0 h̄
hSz i1 = h1| Sz |1i = (4.15)
 
0 1 =−
2 0 −1 1 2
(4.16)

The physical interpretation of this is that we perform a measurement corresponding to Sz on these

particular states, and we acquire h̄
2 and − h̄2 as values when acting on 0 and 1 states respectively.
This particular operator is often called the spin-half operator, which corresponds to a measurement
of the so-called “spin angular momentum” of so-called spin-half particles (like electrons). The
strange values that it returns for each measurement simply correspond to the angular momentum
this property (called “spin”) contributes with respect to the particle.

Exercise 4.7. Show that X, Y and H are all Hermitian operators. Show that UE isn’t.

Exercise 4.8. Calculate hSz i+ . Interpret this result.

Besides Sz there are plenty of operators that are Hermitian. Beyond qubits, and probably most
famously, we have the position operator, often denoted X, and momentum operator P . For the
concerns of this particular lecture we will not get into this right now. Just know that thus far we
have sufficiently covered the foundations of describing, measuring (both of state and properties)
and evolving (that is, changing) the simplest and most fundamental form of quantum state there
is - the qubit. With this, we summarize.

46 Q C a m p 2 0 2 3
C H A P T E R 4 . Q U B IT S & S U P E R P O S IT IO N

4 .6 S u m m a ry

Qubit Bases
   
1 0
|0i = , |1i =
0 1
b b
   
1 1 1 1
b √
|+i = , b √
|−i =
2 1 2 −1

Arbitrary Qubit |ψi

 
a
s.t. |a| + |b| = 1,
2 2
|ψi := a |0i + b |1i = , a, b ∈ C
b
b

The Born Rule

P(φ|ψ) := |hφ|ψi|
2

P(φ|ψ) refers to the probability of getting an outcome |φi from a quantum system prepped
in |ψi. Also, P(·) ∈ R, P(·) ∈ [0, 1]. That is, any probability P(·) must be real and
between 0 and 1.
P(0|ψ) + P(1|ψ) = 1

Quantum Operators
Unitaries: For some operator U to be a valid quantum transform, it must be unitary:

U† U = I

Observables: For some operator A to be a valid quantum observable, it must be hermitian:

A† = A

Where O† = (O)T

∗
The average (“over an infinite number of measurements”) of the
observable A is called its “expectation value”:

hAiψ = hψ| A |ψi (4.17)

Some important operators:

         
1 0 1 1 1 0 1 0 −i 1 0
I = , H = b √ , X = , Y = , Z =
0 1 2 1 −1 1 0 i 0 0 −1
b b b

47 Q C a m p 2 0 2 3
C H A P T E R 4 . Q U B IT S & S U P E R P O S IT IO N

4 .7 E x e r c is e s

Exercise 4.9. Recall the expression for an arbitrary qubit |ψi in the computational basis. Ex-
press this in the |+i, |−i basis. What’s the physical difference between expressing |ψi in the
computational basis and expressing it in the |+i, |−i basis? What physical transformation could
have occurred?

Exercise 4.10. Consider the following state:

|0i + |1i
|Ψn1 i =
2
Notice the difference between this state and |+i. Does |Ψn1 i represent a valid qubit? Explain.

Exercise 4.11. Consider the following state:

|0i + |+i
|Ψn2 i = √
2
Does |Ψn2 i represent a valid qubit? Explain.

Exercise 4.12. I want a “quantum coin” that is doubly-biased on |1i. Specifically, I want a qubit
to be such that I have twice the probability of getting |1i than a |0i from a 0/1 measurement. Give
an expression of a state |ΨBC i that can act as my biased quantum coin. (There are many possible
expressions!)

Exercise 4.13. How would an arbitrary state of a pure qutrit (|τ i ∈ C3 ) look like? How might
you express |τ i in vector representation?

Exercise 4.14. Are these valid quantum transformations?

   
0 i 1 2
M1 = , M2 =
i 0 3 4
b b

[More on next page]

48 Q C a m p 2 0 2 3
C H A P T E R 4 . Q U B IT S & S U P E R P O S IT IO N

Exercise 4.15. Find the output qubit states of these transformations. (Some are these are copied
over from the examples given before and so can be skipped.)

1. For X

(a) X |0i = |1i

(b) X |1i
(c) X |+i
(d) X |−i = − |−i

2. For Z

(a) Z |0i
(b) Z |1i
(c) Z |+i = |−i
(d) Z |−i

3. For I

(a) I |0i
(b) I |+i = |+i
(c) I |ψi

Recall what we’ve learnt about eigenvalues and eigenvectors. From the above one can identify
what are eigenvalues and eigenvectors of X, Z and I. List them. (Remark: For I, the total number
of eigenvectors is notable!)

Exercise 4.16. For an arbitrary qubit ket |ψi = a |0i + b |1i, find XZ |ψi and ZX |ψi. Are these
equivalent quantum states? Does the order of our quantum transformation make a difference to
our output quantum states?

[More on next page]

49 Q C a m p 2 0 2 3
C H A P T E R 4 . Q U B IT S & S U P E R P O S IT IO N

Exercise 4.17.

1. Consider the following qubit states

 
−1
(a) |ψ10 i =
0
b
 √ 
−1/√2
(b) |ψ20 i =
−1/ 2
b
 √ 
1/ 3
(c) |ψ30 i =
b p
2/3
2. Write down a valid partner (i.e |ψ11 i , |ψ21 i , |ψ31 i) to each of the states so as to form three
sets of orthonormal bases.

3. Find X |ψ30 i and X |ψ31 i. What do you notice about the new kets? (You can repeat this
question with Z or H in place of the X for a deeper understanding of unitary operators.)

50 Q C a m p 2 0 2 3
C H A P T E R 4 . Q U B IT S & S U P E R P O S IT IO N

4 .8 E x e r c is e S o lu t io n s

Solution 4.1. Both of the have a norm of 1.

  2
2  ∗  1
|h0|0i| = 1 0∗ =1
0
  2
2  ∗  0
|h1|1i| = 0 1∗ =1
1

And they are orthogonal:

  2
 0
h0|1i = 1∗ ∗

0 =0
1
And since we are dealing with a C2 space, just two of these orthonormal vectors form a basis.

   
1 0
Solution 4.2. Recall that |0i = and |1i = . Hence:
0 1
b b

     
|0i + |1i 1 1 1 0 1 1
|+i := √ b √
= + √ = √
2 2 0 2 1 2 1
     
|0i − |1i 1 1 1 0 1 1
|−i := √ b √
= − √ = √
2 2 0 2 1 2 −1

Same as before. |h+|+i| = 1, |h−|−i| = 1, h+|−i = 0. Thus, they form an orthonormal basis.
2 2

Solution 4.3.

|+i + |−i 1 |0i + |1i 1 |0i − |1i 1

√ =√ √ +√ √ = |0i + |1i + |0i − |1i = |0i
2 2 2 2 2 2
|+i − |−i 1 |0i + |1i 1 |0i − |1i 1

√ =√ √ −√ √ = |0i + |1i − |0i + |1i = |1i
2 2 2 2 2 2

Solution 4.4.

P(φ|ψ) = |hφ|ψi|
2

2
= |(c∗ h0| + d∗ h1|)(a |0i + b |1i)|
2
= |c∗ a h0|0i + c∗ b h0|1i + d∗ a h1|0i + d∗ b h1|1i|
2
= |c∗ a + 0 + 0 + d∗ b|
= (c∗ a)(ca∗ ) + (c∗ a)(db∗ ) + (d∗ b)(ca∗ ) + (d∗ b)(db∗ )
= c2 a2 + d2 b2 + c∗ adb∗ + (c∗ adb∗ )∗
= c2 a2 + d2 b2 + 2 · Re(c∗ adb∗ )

51 Q C a m p 2 0 2 3
C H A P T E R 4 . Q U B IT S & S U P E R P O S IT IO N

The final equality invokes that for any complex number z: z +z ∗ = Re(z)+Im(z)+Re(z)−Im(z) =
2Re(z). It would be a problem if P(φ|ψ) could have an imaginary part since probabilities have to
be real!

Solution 4.5. P (ψ|eiθ |+i) = |eiθ hψ|+i |2 = |eiθ |2 | hψ|+i |2 = | hψ|+i |2

Solution 4.6. For the first three operators, just do it. As for UE , we have:

 −iEt/h̄ T ∗  −iEt/h̄ ∗  iEt/h̄ 

e 0 e 0 e 0
U†E = = = and
0 eiEt/h̄ 0 eiEt/h̄ 0 e−iEt/h̄
 iEt/h̄   −iEt/h̄   iEt/h̄−iEt/h̄   0 
e 0 e 0 e 0 e 0
UE UE =
†
= = .
0 e−iEt/h̄ 0 eiEt/h̄ 0 e−iEt/h̄+iEt/h̄ 0 e0
 
1 0
Which is just, =I
0 1

Solution 4.7. Verifying that X† = X and H† = H is very simple since all their entries are
real and the operators
 iEt/h̄  are symmetric. Meanwhile, as shown in the previous solution, U†E =
e 0
6= UE .
0 e−iEt/h̄

All this means that all the Pauli matrices and the Hadamard operator are valid observables.
They corresponds to measurements that will return +1 and -1 for the first and second states of
their respective basis. For instance hZi0 = 1, hZi1 = −1, hXi+ = 1, hXi− = −1. Notice then that
some transformations can double as hermitian operators also. Meanwhile, UE can only act as an
quantum transform and is not a valid observable.

    
 1 0 1  1
Solution 4.8. hSz i+ = h̄ 1
= 0 This means that if we are
 
1 22= 1 −11
0 −1 1 1
measuring |+i with the observable Sz , the average result we get would be 0. The fact that, as
shown in the notes, hSz i0 = h̄2 , hSz i1 = − h̄2 and that |+i is equally weighted between |0i and |1i, we
can see it as these two measurements cancelling each other out with 12 hSz i0 + 12 hSz i1 = 0 = hSz i+ .
This emphasizes the fact that hAiψ is an expectation value - an average over infinite measurements
of the observable A on the quantum state |ψi.

Solution 4.9. Noting how |0i and |1i can be written as superpositions of |+i and |−i, we have:

|+i + |−i |+i − |−i a+b a−b

|ψi = a |0i + b |1i = a · √ +b· √ = √ |+i + √ |−i
2 2 2 2
There is no physical difference at all! There is no physical change when we change the basis
representing the quantum state just as a actual position of a physical car has no change when we
rotate the y and x axes.

52 Q C a m p 2 0 2 3
C H A P T E R 4 . Q U B IT S & S U P E R P O S IT IO N

Solution 4.10.
2 2
1 1 1 1
P(0|Ψn1 ) = |h0|Ψn1 i| = h0|0i
2
+ h0|1i = =
2 2 2 4
2 2
1 1 1 1
P(1|Ψn1 ) = |h1|Ψn1 i| = h1|0i
2
+ h1|1i = =
2 2 2 4
1 1 1
⇒ P(0|Ψn1 ) + P(1|Ψn1 ) = + = = 6 1
4 4 2
This violates normalization. Hence, |Ψn1 i is not a valid state.

Solution 4.11. Firstly,

!
|0i + |+i |0i 1 |0i + |1i 1 1 1
|Ψn2 i = √ =√ +√ √ = √ + |0i + √ |1i
2 2 2 2 2 2 2

As in the previous exercise, we will find:

2
1 1 1 1 1
P(0|Ψn2 ) = √ + =√ + +
2 2 2 2 4
2
1 1
P(1|Ψn2 ) = √ =
2 2
1 1 1 1 1 1
⇒ P(0|Ψn2 ) + P(1|Ψn2 ) = √ + + + = √ + + 1 6= 1
2 2 4 2 2 4
This also violates normalization. Hence, |Ψn2 i is not a valid state.

Solution 4.12. For doubly biased on “1”:

P(1|ΨBC ) = 2 · P(0|ΨBC )
P(1|ΨBC ) + P(0|ΨBC ) = 1
⇒ 3 · P(0|ΨBC ) = 1
⇒ P(0|ΨBC ) = 1/3
P(1|ΨBC ) = 2/3

Now, applying the Born rule:

P(1|ΨBC ) = |h1|ΨBC i| = 2/3

2

r r
2 2
⇒ h1|ΨBC i = ± or ± i or other suitable complex combinations
3 3
P(0|ΨBC ) = |h0|ΨBC i| = 1/3
2

r r
1 1
⇒ h0|ΨBC i = ± or ± i or other suitable complex combinations
3 3
So a suitable state could be written as:
r r
2 1
|ΨBC i = |1i + |0i
3 3

53 Q C a m p 2 0 2 3
C H A P T E R 4 . Q U B IT S & S U P E R P O S IT IO N

Solution 4.13.
|τ i = a |0i + b |1i + c |2i

With constraints of a, b, c ∈ C and a2 + b2 + c2 = 1. We can represent the qutrit computational

basis as:      
1 0 0
|0i =
b 0 , |1i =
b 1 , |2i =
b 0
0 0 1

Solution 4.14.
   
0 i 0 −i
M1 = , M†1 =
i 0 −i 0
    
0 i 0 i 1 0
M†1 M1 = = =I
i 0 i 0 0 1

Since M1 is unitary, it is a valid quantum transformation.

   
1 2 1 3
M2 = , M2 =
†
3 4 2 4
    
1 2 1 3 10 14
M†2 M2 = = 6= I
3 4 2 4 14 20

Since M2 is not unitary, it is not a valid quantum transformation.

Solution 4.15.

1. For X

(a) X |0i = |1i

    
0 1 0 1
(b) X |1i = = = |0i
1 0 1 0
b
√ 1
     
0 1 1 1
(c) X |+i = √ = 2 = |+i
1 0 2 1 1
b

(d) X |−i = − |−i

2. For Z
    
1 0 1 1
(a) Z |0i = = = |0i
0 −1 0 0
b
    
1 0 0 0
(b) Z |1i = = = − |1i
0 −1 −1 −1
b

(c) Z |+i = |−i

     
1 0 1 1
(d) Z |−i = √1 = √1 = |+i
0 −1 2 −1 1
b 2

3. For I

54 Q C a m p 2 0 2 3
C H A P T E R 4 . Q U B IT S & S U P E R P O S IT IO N

    
1 0 0 0
(a) I |0i = = = |0i
0 1 1 1
b

(b) I |+i = |+i

    
1 0 a a
(c) I |ψi = = = |ψi
0 1 b b
b

X’s eigenvectors are |+i and |−i with eigenvalues of +1 and −1 respectively. Meanwhile Z’s
eigenvectors are |0i and |1i with eigenvalues +1 and −1 respectively. Finally, I’s eigenvectors are
every vector in the space with eigenvalues of +1 always.

Solution 4.16.

XZ |ψi = aXZ |0i + bXZ |1i = aX |0i − bX |1i = a |1i − b |0i

ZX |ψi = aZX |0i + bZX |1i = aZ |1i + bZ |0i = −a |1i + b |0i
⇒ XZ |ψi = −ZX |ψi 6= ZX |ψi

The order of the transformations does matter, they give different outputs!

Solution 4.17.
 
0
|ψ11 i =
1
p 
1/2
|ψ21 i = p
− 1/2
p 
|ψ31 i = p2/3
− 1/3
p 
2/3
For |ψ31 i = p
− 1/3
  p  p    p  p 
X |ψ30 i =
0 1 p2/3
= p 1/3
, X |ψ31 i =
0 1 p2/3 = p−1/3
1 0 1/3 2/3 1 0 − 1/3 2/3

From the above illustration, we see how unitary operators constitute a change in basis. In this
case, a flip about the |+i axes.

55 Q C a m p 2 0 2 3
C h a p te r 5

F ro m L ig h t t o O p t ic a l Q u b it s

Lecture notes by Wu Shuin Jian

5 .1 In t r o d u c t io n

For now, we have a mathematical description of qubits, but that would be pointless if there was
no way to implement them (we are not pure mathematicians after all). There are a number of
qubit candidates, and we will begin by considering one of the oldest phenomenon in the history of
science: light!

Ta k e a w a y s

By the end of this chapter, the reader should:

• Be able to explain how ideas about light developed.

• Explain what polarization is, the various ways to describe it and how they relate to each
other.

• See that the Poincare Sphere can be mapped exactly onto the Bloch Sphere.

• Understand how polarization can be used as qubits in theory and in the lab.

5 .2 R a y O p t ic s

We like to keep things simple, and only introduce additional complexity where necessary. As you
were likely first taught, light is made up of rays travelling in straight lines. It does account for how

56
C H A P T E R 5 . F R O M L IG H T T O O P T IC A L Q U B IT S

light gets from 1 place to another, but light seems to change directions in some cases. To explain
this, we turn to an old idea in physics that has proven effective even to this day.

5 .2 .1 T h e P r in c ip le o f L e a s t A c t io n

Lets consider the path you took to get from your home to CQT. Why did you choose that particular
path? Were you trying to get there in the shortest possible time? Or perhaps trying to minimise
the amount of money you had to spend? Such was the idea about how Nature operated, that
there was a kind of “Natural Economy” where Nature would behave in certain ways to minimise
or maximise some quantities. This would give rise to Fermat’s Principle, that light would travel
between 2 points along the path that would take the least time.

We shall now apply the this principle to see what Fermat was getting at. Basically, we will search
for a path of minimum time in the same way you did in travelling to CQT. Trivially, if 2 points A
and B were on a plane, you’d go with a straight line, but lets complicate things a bit.

You are now a lifeguard, and you see a drowning victim as in Figure 5.1. You need to reach the
victim as soon as possible to maximise the chance of survival. However, even the fastest swimmer
can run faster on land, would you pick a straight line path knowing this?

Figure 5.1: The lifeguard needs to get to the victim in the shortest time possible, but running on
the beach is faster than swimming in the sea, ie vrun > vswim , so would a straight line be the best
option?

Let’s rephrase the question: where along the boundary between beach and sea should be the
lifeguard aim to hit? As in Figure 5.2, we can consider 2 extremes: the point that gives the shortest
path on the beach, and the point that gives the shortest path in the sea. The direct straight line
path can be seen as the in-between option, also the option we would choose if vrun = vswim .

Now let’s use our intuition. Since vrun > vswim , we would want to cover more ground on the beach
than in the sea, so the lifeguard should hit the point on the boundary between beach and sea to

57 Q C a m p 2 0 2 3
C H A P T E R 5 . F R O M L IG H T T O O P T IC A L Q U B IT S

Figure 5.2: The set of all feasible paths to consider can be placed in between 2 extremes: minimum
time spent on the beach, and minimum time spent in the sea.

the right of the straight line path. The greater the value of vrun is over vswim , the further right
of the straight line we would want to be on, and as vrun goes to ∞, we would get the path that
completely minimises time spent in the sea.

We shall skip some mathematical derivations, but the end result will look something like Figure
5.3. Observe the angles θincident and θref raction taken with respect to the normal (the line at a
right angle to the boundary between beach and sea). It can be shown that:

sin(θincident ) vrun
= (5.1)
sin(θref raction ) vswim

and this should start looking very familiar.

Figure 5.3: The path of least time given some known values of vrun and vswim , and that vrun >
vswim . Oh hey! That’s just Snell’s Law.

To make the equation more explicit, we can define some reference speed, c that is greater than
both vrun and vswim . We can think of c as the speed of a fast robot (as measured in a lab and

58 Q C a m p 2 0 2 3
C H A P T E R 5 . F R O M L IG H T T O O P T IC A L Q U B IT S

taken to be constant) that will be taking the lifeguard’s job in the near future. We multiply c on
both sides. With some rearranging, we get:

c c
sin(θincident ) ∗ = sin(θref raction ) ∗ (5.2)
vrun vswim
The ratio of speeds just tells us how much faster the robot could go in the lab than the lifeguard
on that particular terrain. We can call them the “performance index on sand”, nsand , and the
“performance index in water”, nwater . And we end up with:

sin(θincident ) ∗ nsand = sin(θref raction ) ∗ nwater (5.3)

By now, I hope you have realised that we just derived Snell’s law without talking about light, and
all it took was finding the path of least time given some constraints. Replacing the lifeguard with
a ray of light and “performance index” with refractive index will lead to the same result, and with
just this simple idea we have arrived at a profound discovery. But things are about to get more
interesting.

5 .3 W a v e O p t ic s

While Ray Optics has been useful, it does run into some problems. There’s a weird thing that
light does, where if you try to send it through a small hole, it spreads out instead of staying in a
straight line as in Figure 5.4. Rays shouldn’t be abruptly changing direction just from entering a
hole, so we need a new theory.

Figure 5.4: The unexpected behaviour of diffraction that Ray Optics can’t explain.

59 Q C a m p 2 0 2 3
C H A P T E R 5 . F R O M L IG H T T O O P T IC A L Q U B IT S

5 .3 .1 W h a t a re W a v e s ?

A wave can be simply thought of as a propagating disturbance, such as a ripples of water radiating
from the spot where a stone was dropped. If we look at those water ripples, we can see some key
features of waves as in Figure 5.5.

Figure 5.5: Some key features of waves. This won’t be tested but what comes after depends on
this. Be sure to understand what amplitude, wavelength and frequency are.

Some key terms to note are as follows:

• Amplitude: maximum displacement in the wave

• Wavelength: distance between successive crests (or troughs)

• Frequency: the no. of crests that pass by a set point in 1 second

• Phase: a more subtle point; a way to specify the position of a wave (eg, is it at a crest
or trough) at a particular point in time, usually going from 0 to 2π. Of greater concern is
usually the phase difference between 2 waves, ie how much one is ahead or lagging behind
the other

A very important relation between wavelength λ and frequency f (where c is the speed of the
wave) that allows us to go back and forth between them is:

fλ = c (5.4)

The disturbances in waves can propagate in 2 ways, as seen in Figure 5.6:

• Longitudinal: where the direction of oscillation is parallel to the direction of wave propaga-
tion

60 Q C a m p 2 0 2 3
C H A P T E R 5 . F R O M L IG H T T O O P T IC A L Q U B IT S

• Transverse: where the direction of oscillation is perpendicular to the direction of wave

propagation

Figure 5.6: Longitudinal and Transverse waves.

Note that waves were thought to be mechanical in nature, ie, there has to some physical medium
for the disturbance to propagate in, be it air or water. Huygens thought that light was longitudinal,
and propagated through the mysterious 5th element aether that is invisible and is all around us.
That idea of aether had its origins with Aristotle.

5 .3 .2 W h a t d o W a v e s d o ?

A special property of waves is their ability to interfere with each other. Interference can be seen
as a key distinguishing feature of waves as compared to particles. If we have 2 waves that overlap
with each other (assuming they have the same wavelength/frequency and amplitude), they can
combine in some interesting ways as in Figure 5.7, they key factor being the phase difference
between the 2 waves. To combine 2 waves, simply add the displacements at every point of the
waves together to get the resultant wave. Should the waves perfectly overlap with each other, ie
crests overlap with crests, troughs with troughs, we say the phase difference is 0 or some multiple of
2π. The resultant wave will have double the amplitude compared to the initial waves. But should
the waves have a phase difference of π or some odd multiple of it, crests of 1 wave will overlap
with the troughs of the other, and the amplitudes will cancel out, yielding a resultant wave with
0 amplitude.

When waves propagate in more than 1 dimension, it is convenient to describe them using wave-
fronts. A wavefront is simply a surface of constant phase, as seen in Figure 5.8. The direction of
propagation can be inferred by drawing lines perpendicular to each wavefront. They help provide
a quick overview of how a wave is spreading out, and lay the groundwork for a key idea in Wave
Optics.

61 Q C a m p 2 0 2 3
C H A P T E R 5 . F R O M L IG H T T O O P T IC A L Q U B IT S

Figure 5.7: Intereference of 2 waves. Constructive interference occurs when the phase difference
is 0 or some multiple of 2π, when crests of 1 wave overlap the crests of the other. Destructive
interference occurs when the phase difference is an odd multiple of π, when crests of 1 wave overlap
with the troughs of the other.

5 .3 .3 H u y g e n s - F r e s n e l P r in c ip le

To explain how waves propagate from 1 point to another, we now have a strange idea to contend
with. The Huygens-Fresnel Principle basically states that every point on a wavefront acts as a
source of a new (spherical, or in 2D it’s circular) wave. To find the state of the wave at some future
point in time, we just need to apply the principle of interference and add up the infinite number
of new waves to get the resultant wave, as seen in Figure 5.9. Easy!

Now you have a way to account for diffraction. The smaller the slit, the more wave sources you’re
blocking, and so you’re lacking waves at the sides to interfere and form a resultant wave propagating
in a straight line. Thus, the waves that get through the slit end up spreading out more. The main
factor that determines the amount of spreading is the size of the slit compared to the wavelength,
as seen in Figure 5.10.

However, this principle had a problem that Huygens himself could not adequately address: since
the new waves would spread out in all directions, why should we only consider the parts that
propagate in the forward direction for interference to form the resultant wave? That is not to say
that Huygens was completely wrong, his idea could clearly account for phenomena Ray Optics
could not, but do note that all ideas will have their limitations.

Another important point that Figure 5.10 hints at is the relation between Wave and Ray Optics.

62 Q C a m p 2 0 2 3
C H A P T E R 5 . F R O M L IG H T T O O P T IC A L Q U B IT S

Figure 5.8: Wavefronts. Direction of propagation can be found by connecting wavefronts via arrows
that are perpendicular to each wavefront.

Should λ tend toward 0, diffraction becomes less and less significant, and we effectively are back
to dealing with rays. Thus, we can say that Ray Optics is an approximation of Wave Optics in the
limit of λ going to 0. But, note that Ray Optics is unable to account for interference. As such, we
can describe more phenomena with more complex ideas, but for simpler problems like refraction, it
saves time and effort to fall back on Ray Optics should the additional complexity not be required.
And now for 1 more leap in complexity to get to where we need to go!

Figure 5.9: The Huygens-Fresnel Principle in action.

5 .4 E le c t r o m a g e n t ic O p t ic s

It’s amazing how the idea of light as an electromagnetic wave came about. We won’t be solving
Maxwell’s equations, but we can get an intuitive understanding of them. Those 4 equations
summarise the then available knowledge of electricity and magnetism. Of note are Faraday and
Ampere’s Laws, as shown in Figure 5.11. With some mathematical manipulations, those could be
combined to form a wave equation, where a changing electric field created a changing magnetic

63 Q C a m p 2 0 2 3
C H A P T E R 5 . F R O M L IG H T T O O P T IC A L Q U B IT S

Figure 5.10: The amount of diffraction depends heavily on how large the slit is compared to the
wavelength. For slits much larger than the wavelength, diffraction is minimal. But once the slit
gets comparable to the wavelength or smaller diffraction becomes significant.

field and a changing magnetic field created a changing electric field. Sounds paradoxical, but at
least we don’t need aether anymore! The magnetic field is weaker than the electric field, so the
convention is to focus on the electric field since we know the magnetic field is just orthogonal to
it. We have now arrived at an important property provided by this description of light that is
extremely useful for quantum technologies.
(a) (b)

Figure 5.11: (a) Faraday Induction, moving a magnet through coiled wire creates a voltage and
(b) Ampere’s Law, a current running through a wire creates a magnetic field.

5 .4 .1 P o la r iz a t io n

This will be a long section as it’s the major point I’d like to get across.

64 Q C a m p 2 0 2 3
C H A P T E R 5 . F R O M L IG H T T O O P T IC A L Q U B IT S

The electromagnetic wave description makes it necessary that light is a transverse wave. Focusing
on the electric field component, it is intuitive that it can oscillate in a number of directions as seen
in Figure 5.12. I should be waving a rope around at this point in my lecture. The convention used
in most texts is that the wave is travelling in the z-direction, so oscillations of the electric field can
confined to the x and y directions.

The way the electric field oscillates is what we call polarization, a property found in transverse
waves but not in longitudinal ones. It is not something we can directly see, but it can have very real
physical effects. Recall that matter is made of protons, neutrons and electrons, and that charged
particles can interact with oscillating electric fields. In light of that, we need some way to describe
polarizations more precisely.
(a) (b)

Figure 5.12: (a) Electric field oscillating in the x-z plane, we shall call this Horizontal (H) polar-
ization (b) Electric field oscillating in the y-z plane, we shall call this Vertical (V) polarization

Let’s recall the Principle of Superposition. Suppose as in Figure 5.12, we have 2 light waves, 1
with H polarization and the other with V. We can describe the oscillation of the electric field along
the x-axis for H polarization as:

z
Ex = ax cos[ω(t − ) + φx ] (5.5)
c

and for the electric field along the y-axis for V polarization:

z
Ey = ay cos[ω(t − ) + φy ] (5.6)
c

where ax anday are the respective amplitudes, c the speed of light, and φx andφy the respective
phases.

We should be able to combine these waves together, but what happens to their polarizations?
Well, turns out we can just combine them too using simple vector addition. As can be seen in
Figure 5.13, the phase difference between the 2 waves becomes very important, as it determines
the resultant polarization of the wave. Phase differences of 0, π/2, π, 3π/2, ... result in Diagonal
(D) or Antidiagonal (A) polarizations, whereas phase differences of π/4, 3π/4, 5π/4, ... lead to Left

65 Q C a m p 2 0 2 3
C H A P T E R 5 . F R O M L IG H T T O O P T IC A L Q U B IT S

(L) and Right (R) Circular polarizations. Ironically, if we combine a large number of polarizations
together, such that the electric field is practically oscillating in every direction, we end up with
unpolarized light, which is how sunlight usually is.

Since the various polarizations can be combined to create interesting patterns, perhaps a way to
describe them all would be by simply plotting the path that the electric field traces out over 1
period. It is to that idea that we will now turn.
(a) (b)

Figure 5.13: Here we use the perspective of the recipient of the wave to decide what to call the
resultant polarizations. (a) H and V polarizations combine to form Antidiagonal (A) polarization
as indicated by the red dotted line. If their phase difference was shifted by another π radians
we would get Diagonal (D) polarization instead. (b) H and V polarizations combine to form Left
Circular (L) polarization as indicated by the red dotted line. If you were standing where the X
and Y axes are labelled, the dotted red line would appear to be rotating left. Again, should the
phase difference betwen H and V was shifted by another π radians we would get Right Circular
(R) polarization instead.

T h e p o la r iz a t io n e llip s e

Imagine that a wave of polarized light is coming towards you and you trace the displacement of
the electric field in the x-y plane as it arrives at your position. If the light is linearly polarized
(eg. H, V, D or A), you will find yourself tracing a straight line. If we have circularly polarized
light (R or L), you will end up tracing a circle. But are those the only possibilities? What if H
and V polarized waves with a phase difference not equal to any of the values discussed before were
to combine? Then, we end up with an ellipse as seen in Figure 5.14. Using the expressions for Ex
and Ey in equations 1.5 and 1.6, the ellipse is described by the equation:

Ex2 Ey2 Ex Ey
+ −2 cos(φ) = sin2 (φ) (5.7)
a2x a2y ax ay

where φ = φy − φx is the phase difference between Ex and Ey . A quick look at it tells us that the
rotation of the ellipse is due to the term with Ex Ey .

66 Q C a m p 2 0 2 3
C H A P T E R 5 . F R O M L IG H T T O O P T IC A L Q U B IT S

Figure 5.14: The Polarization Ellipse. ψ is the orientation angle, which describes how rotated the
ellipse is. χ is the ellipticity, or how stretched the ellipse is as compared to a circle. By defining
and φ = φy −φx , we can get the relations tan(2ψ) = and sin(2χ) =
ay 2r 2r
r= ax 1−r 2 cos(φ) 1+r 2 sin(φ).

To better understand the ellipse, we shall look at some special cases. Should we have Ey = 0, we
just end up with an expression with Ex alone, ie, H polarization. Likewise, Ex = 0 will end us up
with V polarization. Next, let’s have equal amplitudes ax = ay = a and phase difference φ = 0 or
π. This will yield Ex = ±Ey , which simply means D or A polarization. Finally, we shall keep the
equal amplitude condition ax = ay = a but change the phase difference to φ = π
2 or 2 .
3π
Solving
2 Ey2
the ellipse equation now yields a familiar equation (hopefully): = 1, which is simply R or
Ex
a2 + a2
L polarization. Many more possibilities are illustrated in Figure 5.15.

Figure 5.15: Various possibilities of the polarization ellipse for with polarization components of
equal amplitude ax = ay = a. As seen on the ThorLabs website; sometimes it’s good to educate
your customers about what you’re selling them.

The polarization ellipse is helpful, but it is a bit complicated. To specify a polarization state, we
need to provide the orientation and ellipticity, and those are hard to measure. Plus, we have cool
optical components that can change polarization states, how can we describe those polarization
changing operations in this formalism?. We now move to a description of polarization that allows

67 Q C a m p 2 0 2 3
C H A P T E R 5 . F R O M L IG H T T O O P T IC A L Q U B IT S

us to describe such operations, and those interested in deriving the equation of the polarization
ellipse from the electric fields can refer to the appendix (optional, you do NOT need to do this!).

J o n e s f o r m a lis m

Another way to specify a polarization state is to just list the x and y components in a vector, and
we use eiφ instead of cos(φ) to make life easier:

ax eiφx
     
Ex 1 ax
= →q (5.8)
Ey ay eiφy 2 2 ay eiφ
|ax | + |ay eiφ |

where φ = φy − φx as before. The final step in equation 1.8 is what we call normalisation, so
we can avoid dealing with numbers outside the range from -1 to 1 (this can really make your life
easier!). We can see how the 6 special polarization cases look in Jones Formalism:

H (Ey = 0) V (Ex = 0) D (ax = ay , φ = 0) A(ax = ay , φ = π)

       
1 0 1 1
√1 √1
0 1 2 1 2 −1

R (ax = ay , φ = 2)
π
L (ax = ay , φ = 2 )
3π
   
1 1
√1 √1
2 i 2 −i

and Figure 5.16 has more examples of Jones vectors with the equivalent representation as polar-
ization ellipses.

Figure 5.16: Other examples of Jones vectors and their corresponding ellipses where δ is the phase
difference, this time from fiberoptics4sale.com. Education can be a great sales pitch.

A point on normalizaing Jones vectors: we do not care about the global phase, only the relative
phase difference between Ex and
 Ey . As an example of what I mean,
  consider the Jones vector for
1 i
D polarization written as √12 . We could also write it as √12 , where we have just multiplied
1 i

68 Q C a m p 2 0 2 3
C H A P T E R 5 . F R O M L IG H T T O O P T IC A L Q U B IT S

the vector by a constant i. That would be equivalent to introducing a constant phase shift to both
Ex and Ey , which ultimately has no effect on the final polarization state once combined. So while
it is an abuse of notation, when working in Jones vector logic, we can say:

    
1 1 1 i 1 −1
√ =√ =√ (5.9)
2 1 2 i 2 −1

just don’t tell the mathematicians.

A great thing about vectors is that changes to them can be described by a matrix, so if a wave
with polarization specified by Jones vector E encounters some optical component, we can describe
the transformation performed on E by the component by a matrix T, and the new polarization
can be found by some simple linear algebra:

    
T11 T12 Ex T11 Ex + T12 Ey
TE = = (5.10)
T21 T22 Ey T21 Ex + T22 Ey

and if you have more optical components, just keep applying more matrices. Easy!

Figure 5.17: Examples of Jones matrices for common optical components.

By now you might be noticing something similar to the previous lecture on qubits, and you should!
There’s an easy mapping to bra-ket notation. For instance, we can choose |Hi → |0i , |V i →
|1i , |Di → |+i , |Ai → |−i. We now have another way to visualise superposition. Were we to work
in the HV basis, we could say that |Di and |Ai are superpositions of them:

69 Q C a m p 2 0 2 3
C H A P T E R 5 . F R O M L IG H T T O O P T IC A L Q U B IT S

     
1 1 1 1 0
= |Di = √ (|Hi + |V i) = √ ( + ) (5.11)
1 2 2 0 1

     
1 1 1 1 0
= |Ai = √ (|Hi − |V i) = √ ( − ) (5.12)
−1 2 2 0 1

But, if we choose to work in the DA basis instead, then it is |Hi and |V i that are in a state of
superposition:

     
1 1 1 1 1 1 1
= |Hi = √ (|Di + |Ai) = √ ( √ +√ ) (5.13)
0 2 2 2 1 2 −1

     
0 1 1 1 1 1 1
= |V i = √ (|Di − |Ai) = √ ( √ −√ ) (5.14)
1 2 2 2 1 2 −1

and let’s not forget we still have the RL basis as an option too:

     
1 1 1 1 1 1 1
= |Hi = √ (|Ri + |Li) = √ ( √ +√ ) (5.15)
0 2 2 2 i 2 −i

     
0 1 1 1 1 1 1
= |V i = √ (|Ri − |Li) = √ ( √ −√ ) (5.16)
1 2 2 2 i 2 −i

where we can neglect the global phase factor when it comes to √1 (|Ri
2
− |Li).

Hopefully this should reinforce the point that whether you’re dealing with a superposition of states
depends on what choice of basis you’re working in. With polarization, we have 3 main basis options:
HV, AD, RL, pick whatever’s easiest to work with for the problem. As practice, you can pick a
basis and write the other states as a superposition in that choice of basis that we have not covered.

So now we have a more convenient way to express polarization states and operations to transform
them, which is great, except that we have no way to measure electric fields. What we can measure is
power, which is proportional to the squared amplitude of the electric field. So, for experimentalists
who need to measure polarization, we need something a bit more practical. Let’s first see what
tools are available in the lab.

P o la r iz a t io n c o n t r o llin g o p t ic s

Remember that we have a description of light as an electromagnetic wave, and matter just so
happens to have charged particles that can interact with it. Of interest are crystals, which have
an extremely ordered structure that allows useful interactions with light. The one you might
have heard about most often is a polarizer, where the material has a strongly preferred axis of
transmission, such that any electric field component orthogonal to this axis gets absorbed. For

70 Q C a m p 2 0 2 3
C H A P T E R 5 . F R O M L IG H T T O O P T IC A L Q U B IT S

polarized light hitting a polarizer at an angle θ to the transmission axis, the power transmitted
follows Malus’ Law, ie, if the power before the polarizer was I0 , then the transmitted power is
I0 cos2 θ, as visualised in Figure 5.18.
(a) (b)

Figure 5.18: (a) The operation of a linear polarizer, only oscillations of the electric field parallel to
the transmission axis go through (b) if polarized light hits a polarizer at an angle with respect to the
transmission axis, then the power (square of the amplitude) that gets transmitted is proportional
to cos2 θ

But that’s not all! It was long known that certain crystals, like calcite, displayed what was called
double refraction, as seen in Figure 5.19. This long-running mystery could only be accounted
for with the electromagnetic understanding of light, and opened up the world of birefringent
materials. Birefringence simply means that a material has 2 different refractive indices for 2
different polarizations depending on the crystal axes, ie, the speed of light in such a material is
different depending on the polarization. This leads to 2 divergent beams that create the double
image. The convention is to call the ray that obeys Snell’s law the ordinary ray (o-ray) and the
the disobedient one the extraordinary ray (e-ray). A helpful graphic is in Figure 5.20

This birefringence is the key to polarization control beyond the transmission of 1 linear polarization.
Recall that in adding polarizations the phase difference between the 2 waves played a major role
in determining the final state. Now you have materials that can transmit orthogonal polarizations
at different speeds, ie, you can control the phase difference between the 2 polarization
components by adjusting the thickness of the material!. Basically, one polarization com-
ponent will be slowed down, or retarded, relative to the other. A half-wave plate means that the
retardation has been tuned to half a wavelength, and a quarter-wave plate refers to retardation of
a quarter of a wavelength. The transformations these wave plates can do is illustrated in Figure
5.21. And now that we have our tools, we can find ways to measure polarization the lab.

71 Q C a m p 2 0 2 3
C H A P T E R 5 . F R O M L IG H T T O O P T IC A L Q U B IT S

Figure 5.19: Calcite displaying double refraction. It was only with an understanding of polarization
that this strange phenomenon could be satisfactorily accounted for.

S to k e s v e c to rs a n d th e P o in c a r e s p h e re

How many measurements do we need to determine the polarization of some unknown beam of
light? Spoilers without proof: just 4. Each component of the vector S0 , S1 , S2 , S3 has a value
between 1 and -1. We are usually more interested in the last 3, as S0 is just the total power with
which we will normalise the other values with. In other words, we like working with S0 = 1.

Poincare had a fun idea: instead of representing polarization as an ellipse, why not a sphere?
The orientation angle and ellipticity are still there, but this sphere is located in 3D space with
S1 , S2 , S3 as the x,y and z axes as shown in Figure 5.24. Although the Stokes vector is usually
specified in terms of measured power, it can also be written in terms of the electric fields we have
gotten familiar with. The proof of this can be found in the appendix.

S0 = Ex∗ Ex + Ey∗ Ey = a2x + a2y (5.17)

S1 = Ex∗ Ex − Ey∗ Ey = a2x − a2y (5.18)

S2 = Ey∗ Ex + Ex∗ Ey = 2ax ay cosφ (5.19)

S3 = i(Ey∗ Ex − Ex∗ Ey ) = 2ax ay sinφ (5.20)

With that said, you might be wondering how the Jones and Stokes vectors are related given that
they can both be specified in terms of the electric field. To spoil things early, Jones vectors only

72 Q C a m p 2 0 2 3
C H A P T E R 5 . F R O M L IG H T T O O P T IC A L Q U B IT S

Figure 5.20: Birefringence in action. Note the typo: the polarized extraordinary ray should have
e-ray in the brackets instead of o-ray.

describe states on the surface of the Poincare sphere, or what we call pure states. Stokes vectors
on the other hand can describe states on the surface and within the Poincare sphere, what are
generally referred to as mixed states. The length of a Stokes vector (ie a 3D vector) can be found
by Pythagoras’ Theorem, and we call it the Degree of Polarization (DOP):

q
DOP = S12 + S22 + S32 (5.21)

The DOP is calculated from normalised values of S1 , S2 , S3 and varies from 0 to 1, where completely
unpolarized light has DOP = 0 and completely polarized light has DOP = 1. As you might have
guessed, Jones vectors only describe states with DOP = 1, whereas Stokes vectors can describe all
states with all possible values of DOP .

T h e c o h e re n c y m a t r ix

There is more to say about the relationship between Jones and Stokes vectors, and that goes into
the question of how to describe partial polarization (ie DOP < 1). So here comes the idea of the
Coherency Matrix which we shall denote as G:

Ex∗ Ex Ex∗ Ey
 
G= (5.22)
Ey∗ Ex Ey∗ Ey

The diagonal elements Ex∗ Ex and Ey∗ Ey are just the powers of x and y components, and the off-
diagonal terms are the cross-correlations between the x and y terms. Note that (Ex∗ Ey )∗ = Ey∗ Ex ,

73 Q C a m p 2 0 2 3
C H A P T E R 5 . F R O M L IG H T T O O P T IC A L Q U B IT S

Figure 5.21: Half-wave plates can be used to convert H to V, R to L and vice versa, a sort of
flipping operation. Quarter-wave plates are how to go back and forth between linear and circular
polarizations, ie, with the correct orientation you can convert H, V, A or D to R or L and vice
versa.

or simply complex conjugates, so they contain the same information. You can think of cross-
correlation as how in-step the x and y components are. The cases we have seen thus far have
maximal cross-correlation, as Ex and Ey differ by only a constant phase difference, eg whenever
Ex has a crest, Ey always has a trough or whatever other point in the oscillation. In the case of
unpolarized light then Ex∗ Ey = Ey∗ Ex = 0, ie when Ex is at a crest, Ey can be at any part of its
oscillation, meaning there are not oscillating in step with each other.

To get the Coherency Matrix from the Jones vectors, we need to do some multiplying:

 ∗
Ex∗ Ex Ex∗ Ey
 
Ex
G = E∗ ET = (5.23)


Ex Ey =
Ey∗ Ey∗ Ex Ey∗ Ey

and recalling how the Stokes vectors can be expressed in terms of electric fields, there’s a simple
way to relate the Stokes components to the Coherency Matrix that you can verify yourself:

 ∗
Ex∗ Ey
  
S0 + S1 S2 − iS3 Ex Ex
G= = (5.24)
S2 + iS3 S0 − S1 Ey∗ Ex Ey∗ Ey

Now for another fun question. If you start with a Coherency Matrix, you can see that it’s possible
to find a corresponding Stokes vector. You’ll need to solve a series of linear equations, but hey,
you have 4 unknowns and 4 equations, so it can definitely be done. But what if you want to find
a Jones vector instead?

Let’s consider the example of completely unpolarized light, which has a Coherency Matrix of:

 
1 0
Gunpolarized = (5.25)
0 1

74 Q C a m p 2 0 2 3
C H A P T E R 5 . F R O M L IG H T T O O P T IC A L Q U B IT S

Figure 5.22: How to get a Stokes vector from 4 power measurements (the yellow box is a power-
meter). P1 has a linear polairzer set to 0◦ , P2 has it set to 90◦ , P3 has it set to 45◦ and P4 has a
quarter-wave plate to convert circular polarization back to linear before sending ir though a linear
polarizer set at 0◦ or 90◦ (you can choose whether you want R or L polarization to have maximum
transmission). DOP stands for degree of polarization, if DOP = 1 the Stokes vector reaches the
surface of the Poincare sphere.

If you solve the linear equations to get the Stokes components, you will end up with:

   
S0 1
S1  0
Sunpolarized =
S2  = 0
   (5.26)
S3 0

But if you try to solve for the Jones vector, you end up with some contradictions. For instance,
we have Ex∗ Ex = 1 and Ey∗ Ey = 1, but Ex∗ Ey = Ey∗ Ex = 0. No matter which term you pick to be
0, there is no way to satisfy all the equations. Thus, it is impossible to get a Jones vector from
this Coherency Matrix.

We turn our attention back to the Stokes components in the Coherency Matrix. Notice that we
can break it up into a sum of 4 parts:

         
S0 + S1 S2 − iS3 1 1 0 1 0 0 1 0 −i
G= = [S0 + S1 + S2 + S3 ] (5.27)
S2 + iS3 S0 − S1 2 0 1 0 −1 1 0 i 0

and hopefully those 4 matrices are recognisable as the spin matrices:

       
1 0 1 0 0 1 0 −i
σ0 = , σ1 = , σ2 = , σ3 = (5.28)
0 1 0 −1 1 0 i 0

The factor of 1
2 can be part of the spin matrices or not, just a matter of convention. But with

75 Q C a m p 2 0 2 3
C H A P T E R 5 . F R O M L IG H T T O O P T IC A L Q U B IT S

Figure 5.23: The meaning of each Stokes vector component. Remember that normalized values for
S1, S2, S3 go from −1 to 1
(a) (b)

Figure 5.24: (a) the familiar polarization ellipse (b) the Poincare sphere is basically a different
representation of the ellipse, but now it’s located in a 3D space specified by the Stokes vector
components S1 , S2 , S3 .

that, I hope it is now clear to you that the Poincare sphere used to describe polarizations can be
mapped perfectly onto the Bloch Sphere used to describe qubits. Hence polarization can function
as a 2 level or spin- 12 system! Amazing isn’t it? It is often such surprising connections that lead
to unexpected developments in science, and perhaps you may discover more in the days to come.

76 Q C a m p 2 0 2 3
C H A P T E R 5 . F R O M L IG H T T O O P T IC A L Q U B IT S

5 .5 A p p e n d ix

5 .5 .1 D e r iv in g th e P o la r iz a t io n E llip s e ( S k ip p a b le )

For those curious, I show how to derive the equation of the polarization ellipse from Ex and Ey
here. This is not essential, so skip it if you don’t want to get bogged down in math.

First, I define t0 = t − z
c for convenience, which start us off with:

z
Ex = ax cos[ω(t − ) + φx ] = ax cos[ωt0 + φx ] (5.29)
c

z
Ey = ay cos[ω(t − ) + φy ] = ay cos[ωt0 + φy ] (5.30)
c

The overarching goal is to get rid of t0 in the equations, and we will try to do so by forcing out
something of the form: sin2 (ωt0 ) + cos2 (ωt0 ) = 1.

Next, we apply the cosine sum angle formula:

cos(α + β) = cos(α)cos(β) − sin(α)sin(β) (5.31)

which converts our starting equations to:

Ex
= cos(ωt0 )cos(φx ) − sin(ωt0 )sin(φx ) (5.32)
ax

Ey
= cos(ωt0 )cos(φy ) − sin(ωt0 )sin(φy ) (5.33)
ay

Now with some foresight, we will multiply by sin(φy ), by sin(φx ), and subtract the latter
Ex Ey
ax ay
from the former:

Ex Ey
sin(φy ) − sin(φx ) = cos(ωt0 )sin(φy )cos(φx ) − sin(ωt0 )sin(φy )sin(φx )
ax ay
−cos(ωt0 )sin(φx )cos(φy ) + sin(ωt0 )sin(φy )sin(φx )
= cos(ωt0 )[sin(φy )cos(φx ) − sin(φx )cos(φy )]

Note that we can use the sine difference angle identity:

sin(α − β) = sin(α)cos(β) − cos(α)sin(β) (5.34)

77 Q C a m p 2 0 2 3
C H A P T E R 5 . F R O M L IG H T T O O P T IC A L Q U B IT S

which gets us:

Ex Ey
sin(φy ) − sin(φx ) = cos(ωt0 )sin(φy − φx ) = cos(ωt0 )sin(φ) (5.35)
ax ay

where φ = φy − φx .

We now repeat this trick by multiplying by cos(φy ), by cos(φx ), and subtract the latter
Ex Ey
ax ay
from the former as before:

Ex Ey
cos(φy ) − cos(φx ) = cos(ωt0 )cos(φy )cos(φx ) − sin(ωt0 )cos(φy )sin(φx )
ax ay
−cos(ωt0 )cos(φy )cos(φx ) + sin(ωt0 )cos(φx )sin(φy )
= sin(ωt0 )[sin(φy )cos(φx ) − cos(φy )sin(φx )]

once again we can apply the sine difference angle identity, and we end up with:

Ex Ey
cos(φy ) − cos(φx ) = sin(ωt0 )sin(φy − φx ) = sin(ωt0 )sin(φ) (5.36)
ax ay

We can now see how to get sin2 (ωt0 ) + cos2 (ωt0 ) = 1. We now square the 2 equations we’ve worked
so hard to get:

Ex2 Ey2 Ex Ey
2
sin 2
(φ y ) − 2
sin2 (φx ) − 2 sin(φx )sin(φy ) = cos2 (ωt0 )sin2 (φ) (5.37)
ax ay ax ay

Ex2 Ey2 Ex Ey
cos2
(φ y ) − cos2 (φx ) − 2 cos(φx )cos(φy ) = sin2 (ωt0 )sin2 (φ) (5.38)
a2x a2y ax ay

and we add both of them up:

Ex2 Ey2 Ex Ey
+ −2 (sin(φx )sin(φy ) + cos(φx )cos(φy )) = sin2 (φ)
a2x a2y ax ay

and we apply one final trigonometric identity:

cos(α − β) = sin(α)sin(β) + cos(α)cos(β) (5.39)

and at long last we arrive at our desired equation:

Ex2 Ey2 Ex Ey
+ −2 cos(φ) = sin2 (φ)
a2x a2y ax ay

78 Q C a m p 2 0 2 3
C H A P T E R 5 . F R O M L IG H T T O O P T IC A L Q U B IT S

5 .5 .2 S t o k e s V e c t o r in T e r m s o f E le c t r ic F ie ld s ( A ls o S k ip p a b le )

We note that the power transmitted by an electric field is proportional to the square of its ampli-
tude, ie for the x component of the electric field Ex the power Px can be expressed as:

Px ∝ |Ex |2 (5.40)

and same holds for the y-component. We consider H to be 0◦ or equivalently along the x-axis,
and V to be 90◦ and along the y-axis. So, looking back at Figure 5.23, we can easily derive the
expressions for S0 and S1 :

S0 = |Ex |2 + |Ey |2 = Ex Ex∗ + Ey Ey∗ = a2x + a2y (5.41)

S0 = |Ex |2 − |Ey |2 = Ex Ex∗ − Ey Ey∗ = a2x − a2y (5.42)

where ax , ay are the amplitudes of the electric field components as we defined at the beginning of
the Polarization section.

To get S2 = P+45◦ − P−45◦ , we’ll need to apply a coordinate transformation to the Jones vector
specified in x and y to rotate it 45◦ :

cos(45◦ ) sin(−45◦ )
      
E−45◦ Ex 1 Ex − Ey
= = √ (5.43)
E+45◦ −sin(−45◦ ) cos(45◦ ) Ey 2 Ex + Ey
And referring back to how S2 is measured as shown in Figure 5.23, it’s basically the same logic
that led us to S1 :

∗ 1
S2 = E+45◦ E−45 ◦ − E−45◦ E+45◦ = [(Ex + Ey )(Ex∗ + Ey∗ ) − (Ex − Ey )(Ex∗ − Ey∗ )] = Ex Ey∗ + Ex∗ Ey
2
(5.44)

We can take this a step further to get the expression with the electric field amplitudes. First,
we note that Ex Ey∗ = (Ex∗ Ey )∗ , so our expression for S2 is the sum of a complex number and its
conjugate. Second, for a complex number z, we have the identity z + z ∗ = 2Re(zz ∗ ). That means:

S2 = 2Re[Ex Ey∗ ] = 2Re[Ex∗ Ey ] = 2Re[ax e−iφx ay e−iφy ] = 2ax ay Re[ei(φy −φx ) ] = 2ax ay cos(φ)
(5.45)

where φ = φy − φx , and recall that thanks to Euler we have eiθ = cos(θ) + i ∗ sin(θ).

We can repeat this trick to get S3 ; this time, we use a coordinate transformation to get to L and
R polarizations:

79 Q C a m p 2 0 2 3
C H A P T E R 5 . F R O M L IG H T T O O P T IC A L Q U B IT S

      
EL 1 1 i Ex 1 Ex + iEy
=√ =√ (5.46)
ER 2 1 −i Ey 2 Ex − iEy

and recalling Figure 5.23 on what S3 means:

1
∗
S3 = ER ER −EL EL∗ = [(Ex −iEy )(Ex∗ +iEy∗ )−(Ex +iEy )(Ex∗ −iEy∗ )] = i(Ex Ey∗ −Ex∗ Ey ) (5.47)
2

And to get the expression in terms of the electric field amplitudes, we note that the final expression
in brackets is the difference between 2 complex conjugates, and we have this handy identity for a
complex number z: z − z ∗ = −2Im(zz ∗ ). And we push on:

S3 = −2Im[Ex Ey∗ ] = 2Im[Ex∗ Ey ] = 2ax ay Im[ei(φy −φx ) ] = 2ax ay sin(φ) (5.48)

where φ = φy − φx , and Euler comes to the rescue again. And there we go! The Stokes vector in
terms of the electric field!

80 Q C a m p 2 0 2 3
C h a p te r 6

S u p e r c o n d u c t in g c ir c u it s

fo r Q u a n tu m C o m p u t in g

Lecture notes by Fernando Valadares

6 .1 F ro m S u p e rc o n d u c to rs to Q u a n tu m C o m p u te rs

8 April, 1911. Kamerlingh Onnes arrives at his laboratory at the Leiden University, in the Nether-
lands. There, his team had already set up the day’s experiment. They would test a lingering doubt
in the physics community: what happens to the resistance of metals at temperatures close to ab-
solute zero? Some believed the electronic motion inside the material would become “frictionless”
and thus reduce the resistance to zero. But others argued the electrons themselves would freeze
and block the flow of any current. The arena was open for a heated debate. And the expertise of
Onne’s team in cooling liquid helium to temperatures down to 1.1 K was a good opportunity to
settle this question.

The team transferred liquid helium to a cryostat that contained a resistor made of mercury. This
cryostat was a very elaborate glass container that took months to be crafted, and it was meant to
isolate the materials inside from the external temperatures. After transfer, the system was further
cooled by pumps and, by the end of the afternoon, when the temperatures reached values lower
than 4 K, Kamerlingh Onnes noted down the Dutch sentence in his notebook Kwik nagenoeg nul
- mercury (resistance) practically zero. He had observed the phenomenon of superconductivity for
the first time.

We will all live to see breakthroughs that will change our lifestyle and perspective on the world.
I was in shock the first time I heard about this sorcery called Wi-Fi, or the streaming services
that eventually substituted the adamant technology of video cassettes. And I bet the scientific

81
C H A P T E R 6 . S U P E R C O N D U C T IN G C IR C U IT S F O R Q U A N T U M C O M P U T IN G

community back in Kamerlingh Onnes’ time was impressed by the properties of his team’s discovery,
which would eventually play a part in another futuristic technology.

Figure 6.1: Heike Kamerlingh Onnes on the right side and Gerrit Flim in the Leiden laboratory,
around 1911. Image taken from [1].

In 1981, 67 years after Onnes’ experiment, Richard Feynman delivered a keynote speech proposing
the concept of a quantum computer. He argued that a machine able to mimic the quantum behavior
of nature would be much more efficient in simulating molecules and materials than regular (or
classical) computers ever could. Besides materials simulation, there are many possible applications
of quantum computing in the fields of optimization and cryptography, for example. So far, however,
there is still no quantum computer efficient enough to offer an advantage over classical computers.

The prospective uses of this abstract machine attracted the interest of the scientific community
and industry. And, in most recent years, prototypes based on many physical systems - trapped
ions, optics, and superconducting circuits - started to appear. As for the latter case, it turns
out that superconductivity is a quantum phenomena that can be observed in macroscopic scales,
and therefore we can build circuits that have quantum properties. The relatively large scale of
the circuits also make them easier to fabricate, making this platform one of the most popular at
the present time. The purpose of this lecture is then to talk about quantum hardware based on
superconducting systems and to give an introduction to the field in general. It is divided in two
parts. The first part explores some of the macroscopic properties of superconducting materials: the
persistent current flow, the magnetic field expulsion, and the intriguing phenomenon of quantum
lock. On the second part, we move on to superconducting circuits and show how they can be used
to build a qubit, the basic unit of quantum computers.

82 Q C a m p 2 0 2 3
C H A P T E R 6 . S U P E R C O N D U C T IN G C IR C U IT S F O R Q U A N T U M C O M P U T IN G

Figure 6.2: a) Movement of electrons on normal conductors. The electrons (orange circles) will
bump, or scatter, onto the ions (green circles), limiting their speed. (b) Cooper pair interacting
with the ion lattice on a superconductor while avoiding the scattering of individual electrons.

6 .2 S u p e r c o n d u c t iv it y

6 .2 .1 Z e ro R e s is t a n c e a n d P e r s is t e n t c u r r e n t s

Normal metals like copper and gold are very efficient in carrying electric currents and, in many
situations, their resistance can be considered to be zero. But this is just an approximation. In any
metal, flowing electrons constantly bump (or more precisely, scatter) onto the ions that compose
the material. Because of these obstacles, the electrons reach a maximum velocity that is related
to the resistivity of the metal.

But metals have a hidden feature, a very weak interaction that went by unnoticed for too long: the
fields cast out by ions can make two electrons attract each other. In principle, these two electrons
should be pushed apart due to their negative charges. However, when they are moving in opposite
directions, the lattice of ions effectively makes them pair up. When two electrons are connected in
that way (forming a Cooper pair), they avoid the ion scattering altogether. Without the scattering,
there is no maximum velocity and the resistivity becomes actually zero.

This microscopic mechanism was proposed by Bardeen, Cooper and Schrieffer in 1957 to explain
superconductivity. Their calculations also show that, even though this electron-pairing effect is
always present, high thermal energy easily breaks the Cooper pairs. That is why superconductivity
occurs at very cold temperatures.

But if there is nothing there to stop the electrons, can we make them flow forever without giving
them more energy? The answer is yes. By exciting a closed circuit made with a superconducting
wire, one can create a persistent current. The strength of this current can be probed with external
magnets that react to the magnetic field it creates. To demonstrate this remarkable effect at a

83 Q C a m p 2 0 2 3
C H A P T E R 6 . S U P E R C O N D U C T IN G C IR C U IT S F O R Q U A N T U M C O M P U T IN G

conference in 1932, Gerrit Flim (Figure 6.1) flew from the Leiden laboratory, the closest source of
liquid helium, to London with the cryostat on his lap. Inside the cryostat and surrounded by liquid
helium was a ring of superconducting lead in which Flim had excited a current of 200 amperes.
After arriving at the conference, the audience witnessed that the ring still carried the current that
had been started before the flight.

The zero resistance property would later be proven essential in building superconducting circuits.
Without it, the quantum information carried by the electrons is destroyed by the scattering with
the ions. The circuits of quantum computers should then be cooled down as much as possible to
be well within the superconducting regime.

6 .2 .2 T h e M e i s s n e r E ff e c t a n d Q u a n tu m L o c k in g

As was mentioned in the last section, the uninterrupted motion of Cooper pairs (also known as
supercurrent) preserves the information that they carry. This information is a phase. Phases are
properties of waves, not of particles. But electrons, depending on the conditions, can behave as
either. In a superconductor, therefore, the electrons forming Cooper pairs preserve their wave-like
characteristics, and this change in behavior enables a whole new property.

When the superconductor is placed inside a magnetic field, a supercurrent will start flowing on its
surface. And it flows with a direction and velocity that blocks the magnetic field from entering
the bulk of the material. This property is called perfect diamagnetism, or the Meissner effect,
because the material cancels the external field by producing the exact opposite field. This is a
direct consequence of the electron phase, and consequently doesn’t happen in normal conductors
(see Figure 6.3).

This phenomenon also finds a use in superconducting quantum computers. Stray magnetic fields
can reduce the quality of the circuits, but this problem is solved by placing them inside a super-
conducting enclosure that acts as a filter to this noise.

When taken to the limit, however, the perfect diamagnetism can break. Under a sufficiently strong
magnetic field, superconductors can react in two ways: they either lose their superconductivity
altogether, turning back to a normal material, or they allow the field to penetrate through very
narrow tubes of normal conductor known as vortices. The response depends if the material is a
type-I or type-II superconductor, respectively.

Figure 6.4 shows how these vortices look like. Supercurrent flows in a spiral movement around the
vortices, directing the flow of the field and blocking if from penetrating the rest of the material.
But there is more to it. The laws of quantum mechanics require that the magnetic flux through a
vortex should be equal to a constant, the flux quantum

h
Φ=B·A= = Φ0 . (6.1)
2e

84 Q C a m p 2 0 2 3
C H A P T E R 6 . S U P E R C O N D U C T IN G C IR C U IT S F O R Q U A N T U M C O M P U T IN G

Figure 6.3: Depiction of the Meissner effect. When placed in a constant magnetic field, a normal
conductor won’t necessarily react to it (depends on the material). On the other hand, supercon-
ductors will expel all of the field from its interior. Adapted from [2].

The magnetic flux Φ is the product between the field magnitude B and the vortex area A; the flux
quantum Φ0 is calculated from Plank’s constant h and the electron charge e.

This constraint leads to a very unexpected effect. Suppose the field is generated by an external
magnet. If one tries to move the magnet, the supercurrents will fight back the change in the field
strength, pulling the magnet back. This effect is called quantum locking (or flux pinning) and can
be used for frictionless transport. Figure 6.4 shows how the effect looks like in practice.

Figure 6.4: Depiction of the quantum locking effect. The diagram in the left shows how a quantum
of magnetic flux can go through a superconductor by creating thin channels of normal conductor
called vortices. The figure in the right is a demonstration of a magnet locked to a superconductor.

85 Q C a m p 2 0 2 3
C H A P T E R 6 . S U P E R C O N D U C T IN G C IR C U IT S F O R Q U A N T U M C O M P U T IN G

6 .3 S u p e r c o n d u c t in g C ir c u it s

6 .3 .1 B it s a n d Q u b it s

The first transistors were developed in mid-twentieth century, enabling the revolutionary digital
electronics industry. Transistors are gates in a circuit, either allowing or blocking the passage of
current in its conducting paths. They can hold a unit of binary information, a bit, that can be in
state 1 or 0 depending whether the current is flowing or not. Since mathematical operations can be
translated into binary operations, computers use hundreds of millions of transistors to manipulate
the flow of current and perform calculations much faster than a human ever could.

Digital quantum computers build on top of this classical system by using quantum bits (qubits).
Similar to the classical bit, the qubit is a system that has two independent states. These are
commonly labeled by the symbols |0i, also known as ground state, and by |1i, the excited state.
Their characteristic property is having well-defined energies E0 and E1 , with ∆E = E1 − E0 > 0.
The qubit frequency fq (in hertz) is obtained in the expression ∆E = hfq , where h is Planck’s
constant.

Before we start, one important comment: the quantum bit is a mathematical abstraction. In other
words, it is a useful layout that can describe a myriad of real systems, superconducting circuits
included. The overall logic of operation covered in this section finds parallel in many other quantum
computing schemes, such as those based on trapped ions and polarized light.

In the next section we will explore the hydrogen atom as a prototype qubit. Then, it will be shown
how a transmon, one type of superconducting device, can work similarly to an artificial atom.
The section will end with the demonstration of the experimental control and measurement of the
transmon state - the basic toolkit for operating quantum computers.

6 .3 .2 T h e A to m a s a Q u b it

One prototypical example of a qubit is an atom. Suppose the hydrogen atom, in which a single
electron may leap between atomic shells (Figure 6.5). By placing the atom in a sufficiently cold
environment, the electron is highly probable to be in the least energetic shell |0i. By shining the
atom with a laser with frequency that matches the transition |0i ↔ |1i, the electron would receive
energy and flip to the first excited state.

In the atom example, the laser plays the role of a drive with frequency fd =f|0i↔|1i . The transition
from |0i to |1i occurs once the laser has transferred an energy ∆E to the atom. It is then reasonable
to ask what happens if you suddenly halt the operation halfway through, providing an energy of
∆E/2, for example. Does the atom reach a half-excited state |0.5i? The answer is that no such

86 Q C a m p 2 0 2 3
C H A P T E R 6 . S U P E R C O N D U C T IN G C IR C U IT S F O R Q U A N T U M C O M P U T IN G

Figure 6.5: Schematic representation of atomic shells and their respective energy levels. The
uneven energy spectrum allows us to restrain the electron to the first two shells.

state exists, and thus the electron ends up in a quantum superposition state of both |0i and |1i.
The electron is effectively residing in two atomic shells at the same time.

Another important case to consider is when the laser continues to shine on the atom after the elec-
tron has leaped to |1i. Counter-intuitively, the electron does not absorb more energy to go to outer
shells (|2i, |3i, etc.). The energy levels of an atom are uneven, that is, f|0i↔|1i 6= f|1i↔|2i 6= f|1i↔|3i ...
Consequently, if the drive is tuned with the first transition, it is detuned with respect to the higher
transitions. The detuned transitions do not listen to the drive, just like a radio that is tuned into
the wrong station. Instead, the prolonged action of the laser makes the electron flip back to |0i.
This curious effect is called a stimulated emission, whereby the electron returns to the laser the
energy it had previously absorbed.

The Bloch sphere (Figure 6.6) is a helpful way of visualizing these dynamics. In a brief description,
the excited state is depicted at the base of a sphere, while the ground state is at its very top. Each
other point on the surface of the sphere is on a superposition of both states and can be uniquely
determined by two angles: the azimutal angle ϕ, and the polar angle θ. In the example of the
atom, the laser makes the polar angle change in time as θ = Ω∆t, where Ω depends on the strength
of the drive and ∆t is the laser exposure time. Therefore, as long as the laser is active, the electron
state rotates between |0i and |1i, passing through superposition states. This is known as Rabi
oscillation, an effect that we will observe experimentally by the end of this class.

Finally, suppose that you want to determine in which shell the electron is. This can be done by
measuring its energy, which would certainly give an energy values of E0 for |0i and E1 for |1i. But
the measurement of a superposition state halfway between |0i and |1i (θ = π/2) would not result

87 Q C a m p 2 0 2 3
C H A P T E R 6 . S U P E R C O N D U C T IN G C IR C U IT S F O R Q U A N T U M C O M P U T IN G

Figure 6.6: The Bloch sphere, at the left. Each point on its surface is a possible state of the qubit,
and can be uniquely determined by the azimuthal (ϕ) and polar (θ) angles. When the energy of
the qubit is measured, it collapses either to the excited state with probability sin2 θ2 or to the

ground state with probability cos2 θ2 .

in (E0 + E1 )/2, as you might have already suspected. Instead, it would have a 50% chance of
resulting in either E0 or E1 . Also, the sheer act of measuring the qubit would destroy its original
state, displacing it to |0i or |1i depending on the result of the measurement. This is another
inherently quantum effect, the quantum state collapse. To be precise, the probabilities of collapse
of a superposition state to either |0i or |1i are cos2 θ2 and sin2 θ2 , respectively.

It is perfectly possible to use atoms in such a way and this is explored in many fields such as
cavity quantum electrodynamics (QED). But since nature offers many different systems that can
be used as qubits, science has the option to pick whichever is most suitable for performing quantum
computing - that is, a system easy to operate and to customize.

6 .3 .3 B u ild in g th e A r t i fi c i a l A t o m

So far in the race for creating efficient and advantageous quantum computers, no physical system
has yet been declared a winner - each has its advantages and caveats. Nevertheless, implementations
based on superconductor circuits are nowadays very popular not only in academia but also in the
nascent industry.

88 Q C a m p 2 0 2 3
C H A P T E R 6 . S U P E R C O N D U C T IN G C IR C U IT S F O R Q U A N T U M C O M P U T IN G

These circuits use rather simple materials such as aluminum and sapphire, and their fabrication can
take advantage of the well-known processes used for producing classical computer chips. Besides,
the qubits can be designed to have transition frequencies on the order of gigahertz (109 Hz),
whereas these frequencies can get up to 1015 Hz in atomic qubits. Laser drives can consequently
be replaced by standard electronic signals for added simplicity. Another significant advantage is
that superconducting materials have infinite conductivity and their electrons can travel without
any energy dissipation, a nice property to have if you want the qubit you placed at |1i to maintain
that state for a reasonable time.

This does not mean quantum computing based on these systems is trivial. The superconductivity
phenomenon only happens when the circuits are cooled down to really low temperatures, which
can be in the order of a few kelvins above absolute zero for ordinary conductors such as aluminum.
For that reason, all devices must be stored inside a dilution refrigerator, a cryogenic device that
can reaches temperatures down to 10 mK.

Figure 6.7: At the left side, a superconducting microstrip resonator deposited over a dielectric,
represented in grey and green, respectively; at the right side, schematic representation of the evenly
spaced energy levels of a resonator.

The problem in question is to find such a circuit that can be operated as a two-level system.
Consider a resonator as a first attempt. There are many different architectures to build microwave
resonators, one of which is depicted in Fig. 6.7. Its physical behavior is the electromagnetic analog
of a guitar string: it carries a field with well-defined frequency of oscillation that can be tuned
depending on its shape, and the amplitude of the field depends on the amount of energy it is given.
The central difference is that, in the superconducting phase, the energy spectrum of this field is
quantized in distinct levels, and each level can be labeled as an individual quantum state. The
flaws of such device become apparent once one analyzes the values of excitation energies. The
transition |0i ↔ |1i has the same characteristic frequency as |1i ↔ |2i, which is also the same for
any two consecutive levels. Applying a drive with frequency fd = f0↔1 would then cause a leaking

89 Q C a m p 2 0 2 3
C H A P T E R 6 . S U P E R C O N D U C T IN G C IR C U IT S F O R Q U A N T U M C O M P U T IN G

to the higher levels, making it impossible to isolate two single states and use it as a qubit.

It is possible to correct this issue and reproduce the convenient uneven energy spectrum of atoms.
Using the resonator as a basis, the transmon accomplishes this task by introducing the Josephson
Junction, which is a nonlinear element that disturbs the circuit. Each energy level of the spectrum
is changed in a different way, separating the transition frequencies. Figure 6.8 shows one possible
transmon architecture, in which a resonator is split in two halves connected by the junction. This
circuit allows |0i and |1i to be addressed individually with the use of electronic drives, mimicking
the hydrogen atom and laser example.

Figure 6.8: At the left side, a superconducting microstrip resonator deposited over a dielectric,
represented in grey and green, respectively; at the right side, schematic representation of the evenly
spaced energy levels of a resonator.

We have explored two crucial elements of the field of circuit quantum electrodynamics, the resonator
and the transmon, and concluded that the latter can be used as a qubit. In truth, both find many
different uses and designs in superconducting quantum computing, and the list grows by day. In
the next section, these elements will be used to experimentally demonstrate the Rabi oscillations
in a transmon, leading to drive operations capable of flipping its state between |0i and |1i.

6 .3 .4 E x p e r im e n t : C o n t r o llin g th e T ra n s m o n S ta te

The objective of this experiment is to demonstrate Rabi oscillations between the two lowest trans-
mon states by continuously changing the superposition angle θ.

As we have seen, the action of a drive makes the qubit state evolve as θ = Ω∆t, so this angle can be
controlled by changing either the exposure time ∆t or the drive strength Ω. Since time in quantum
computing is a rather scarce resource (the quantum states decay very quickly), we will keep the
duration of all drives to a small value while varying Ω. This experiment is known as Power Rabi.

90 Q C a m p 2 0 2 3
C H A P T E R 6 . S U P E R C O N D U C T IN G C IR C U IT S F O R Q U A N T U M C O M P U T IN G

First, we will briefly describe the laboratory settings required for conducting this experiment.
Then, we define the experimental procedure and the expected result. At the end, the collected
data will be shown and discussed.

E q u ip m e n t

The qubit circuit (QC) consists of a transmon sitting on an insulating sapphire (α-Al2 O3 ) plate
attached to the interior of a superconducting resonant cavity (Fig. 6.9). The resonant cavity is
also a type of resonator, but here it will simply aid in the readout of the transmon state. The
superconducting parts are made out of aluminum (Al), and the Josephson Junction consists of
amorphous aluminum oxide (AlOx ). The characteristic frequencies of the cavity and the transmon
are 8.7069 GHz and 4.110 GHz, respectively.

Figure 6.9: Qubit circuit used in the experiment. The transmon is placed on a sapphire substrate
inside a resonant cavity.

To make the aluminum superconducting, the qubit circuit is placed inside a Bluefors dilution
refrigerator (DR) and cooled down to around 10 mK (see Fig. 6.10). Two SMA cables connect
the system to the outside of the DR, and these are attached to an external control circuit (CC).
A computer is connected to the CC and is able to program the electromagnetic signals sent to the
fridge and read the signals that come back.

T h e E x p e r im e n t

The experiment consists in the following steps:

1. The transmon and cavity are prepared in the ground state;

91 Q C a m p 2 0 2 3
C H A P T E R 6 . S U P E R C O N D U C T IN G C IR C U IT S F O R Q U A N T U M C O M P U T IN G

2. A 4.110 GHz pulse is sent to the QC and drives the qubit for a fixed time ∆t =900 ns and
variable strength αΩ, where α is a scaling factor ranging from -1.6 to 1.6;

3. A 8.7069 GHz pulse is sent to the QC, interacts with the cavity and then goes back to the
CC. The received signal power indicates the transmon state;

4. The value of α is updated and steps 1 to 3 are repeated;

5. Steps 1 to 4 are repeated Nreps = 4000 times, and the received signal power is averaged;

6. The average signal power is plotted against α to indicate the qubit state evolution as a
function of θ.

Step 1 is simply performed by waiting for 200 µs, because the energy stored in both cavity and the
transmon eventually dissipates and the QC returns to the ground state. Step 2 makes the transmon
oscillate in the Bloch sphere, and polar angle of the final state is θ = αΩ∆t. The actual value of
Ω is difficult to predict beforehand, and it is not relevant for the experiment. Step 3 consists on
the qubit state measurement, for the received measurement pulse will have a power level of P|0i
or P|1i depending on the transmon state. If the transmon is in a superposition state, the average
power level will depend on the quantum state collapse probabilities: P=P|0i cos2 ( θ2 )+P|1i sin2 ( θ2 ).
Defining ∆ = P|0i − P|1i /2 and using the identity 2 sin2 ( θ2 ) = 1 − cos θ, we can write

P = P|0i + ∆ (cos (αΩ∆t) − 1) . (6.2)

The P versus α graph is thus expected to have the shape of a cosine.

92 Q C a m p 2 0 2 3
C H A P T E R 6 . S U P E R C O N D U C T IN G C IR C U IT S F O R Q U A N T U M C O M P U T IN G

Figure 6.10: The dilution refrigerator used in our experiment (top), and the temperature sensors
display showing the circuit temperature of approx. 9.2 mK in the bottom-right panel (bottom).

93 Q C a m p 2 0 2 3
C H A P T E R 6 . S U P E R C O N D U C T IN G C IR C U IT S F O R Q U A N T U M C O M P U T IN G

R e s u lt s

The experiment was performed and the P versus α graph is shown in Figure 6.11. The sinusoidal
shape is evident and agrees with the prediction. Note that, for α = 0, the qubit is necessarily in
the ground state since the action of the drive is null. Therefore, we can conclude from the graph
that P|0i ≈ 1e-6 and P|1i ≈ 0.4e-6.

Figure 6.11: Power Rabi results. The orange line represents experimental results, while the blue
line corresponds to a fitted cosine function. The vertical blue bars show the standard deviation
(in other words, amplitude) of the measurement noise. Data points corresponding to ground and
excited state are highlighted.

The Rabi results can be used to design qubit operations: picking a drive with α = 0.932 performs
a qubit flip operation, that takes the |0i state to |1i and vice-versa; by instead picking a drive with
half the strength, α =0.466, the qubit ends in a quantum superposition of |0i and |1i.

R e f e r e n c e s f o r t h is C h a p t e r

[1] Van Delft, D., & Kes, P. (2010). The discovery of superconductivity. Physics Today, 63(9),
38-43. https://doi.org/10.1063/1.3490499

[2] Annett, J. F. (2004). Superconductivity, superfluids and condensates (Vol. 5). Oxford
University Press.

94 Q C a m p 2 0 2 3
Wednesday

95
C h a p te r 7

T h e S c h r ö d in g e r E q u a t io n

Lecture notes by Lim Kian Hwee

Edited by Zaw Lin Htoo

In the previous chapters, we have learnt how to describe quantum systems, what happens when we
measure them, and a small handful of operations we can perform. We have also had some practice
in describing physical apparatuses within the formalism, and encountered some odd phenomena
that can be observed only with quantum objects.

However, everything we have discussed thus far has been exceedingly static—we prepare a state,
and it stays that way, unchanged, until we perform a measurement or transformation on it. Yet,
we know that is not usually the case. An apple under the influence of earth’s gravity falls to the
ground over time, an asteroid in space far from every planet moves in a straight line due to its
initial momentum. Physical objects are dynamic: they change over time according to the external
forces acting upon them. What, then, are the dynamics of a quantum system?

This chapter brings together everything we have learnt in the past five chapters, culminating in
the final puzzle piece that completes the full description of a quantum system.

Ta k e a w a y s

By the end of this chapter, the reader should:

• Retread the historical steps taken during the development of quantum mechanics.

• Understand what it means for the Schrödinger equation to describe the dynamics of a quan-
tum system.

• Know how to use the Schrödinger equation to find the state |ψ(t)i of a quantum system at
a later time t given the initial state |ψ(0)i.

96
C H A P T E R 7 . T H E S C H R Ö D IN G E R E Q U A T IO N

• Have an appreciation of the Schrödinger equation for continuous degrees of freedom (for the
interested reader).

7 .1 H is t o r ic a l D e v e lo p m e n t o f Q u a n t u m M e c h a n ic s

In Chapter 1, we learnt that two major problems—the ultraviolet catastrophe and the photoelectric
effect—pointed to the inadequacies of classical physics, which led to the development of quantum
mechanics. It is worthwhile to remind ourselves about its historical development before we proceed
with the topic of this chapter proper.

7 .1.1 C la s s ic a l P a r t ic le s a n d W a v e T h e o r ie s o f L ig h t

As we will see when we delve into the mathematical details later, much about our modern under-
standing of matter began from our attempts at uncovering the nature of light. At the heart of this
understanding is the long-standing debate fought by the biggest names in the history of physics:
is light is a particle, or a wave?

Newton’s corpuscles. In his 1704 treatise Opticks, Isaac Newton held his position in the former
camp. He described light to be made up of small particles, called “corpuscles”, that obey his
eponymous laws of motion. These particles of light are assigned different connate (innate) proper-
ties to account for their different colours. Newton’s corpuscular theory of light could successfully
explain reflection—particles bouncing off a surface—and refraction—particles accelerated laterally
upon entering a denser medium.

Normal Inc
cor iden
pu t
scle
s
cl d
In rpu

Rarer medium
us te
co

es
ci sc

rp ec
de le

Denser medium
co efl
nt s

Re pus
co
R

fra cle
r
cte s
d

Reflecting surface Refracting medium

Figure 7.1: Reflection and refraction according to Newton’s corpuscular theory of light.

Huygens’ and Fresnel’s light waves. Meanwhile, Newton’s contemporaries Robert Hooke
and Christiaan Huygens, and later Augustin-Jean Fresnel, held the latter view. By taking light
to be a wave that traveled at different speeds in different media, they showed mathematically
that refraction could be easily explained as the medium-dependent propagation of light waves.
The resulting Huygens–Fresnel principle was extremely successful at reproducing the behavior of

97 Q C a m p 2 0 2 3
C H A P T E R 7 . T H E S C H R Ö D IN G E R E Q U A T IO N

light. Subsequently, after Thomas Young’s demonstration of the wave interference of light with his
double-slit experiment in 1801, which could not be explained by Newton’s corpuscles, the classical
debate was settled in Hooke’s, Huygens’, and Fresnel’s favour.

Light source

Double slit
Screen

(b) A schematic of the double slit experiment, vi-

sualised by the wavefronts of light. Constructive
(a) Refraction of light under Huygen’s principle.
and destructive interference lead to the fringe pat-
Figure adapted from [1].
tern on the screen. Figure adapted from [2].

Figure 7.2: Huygen’s principle states that every point on a wavefront is itself the source of spherical
wavelets, and the secondary wavelets emanating from different points mutually interfere. The sum
of these spherical wavelets forms the wavefront. Applying Huygen’s principle of wave propagation
to light explains the phenomena of refraction and interference.

Maxwell’s electromagnetic waves. In 1861, James Clerk Maxwell united the laws of electricity
and magnetism into a single set of equations. Then, he noticed a few years later that a solution to
his equations predicted the existence of electromagnetic waves—one that propagates at the speed
of light—and deduced that light was a type of electromagnetic wave. This theory led to predictions
that agreed with many observed phenomena of light, and was so successful that the notion of light
being a particle completely vanished: at least, until the end of the 19th century.

z
Magnetic field
y component

Wavelength

Electric field component

x

Figure 7.3: Schematic of light as an electromagnetic wave. Figure adapted from [3].

98 Q C a m p 2 0 2 3
C H A P T E R 7 . T H E S C H R Ö D IN G E R E Q U A T IO N

7 .1.2 In a d e q u a c ie s o f C la s s ic a l P h y s ic s

The photoelectric effect. The photoelectric effect is the phenomenon where electrons are emit-
ted when a metal is illuminated with light. A peculiar observation was noted when performing
experiments of the photoelectric effect: regardless of its intensity, light below a threshold frequency
would not cause any electrons to be emitted. This contradicts a purely wave description of light,
which predicted that increasing the intensity of light should also increase the energy of the light
absorbed by the electrons on the metal. Hence, if light were just a wave, increasing its intensity
should lead to electron emission, regardless of its frequency.

Einstein’s solution. To explain the photoelectric effect, Albert Einstein took Max Planck’s
concept of the quanta of light seriously. Planck’s solution to the black-body radiation problem
implied that each quanta of light—which you would now know as a photon—carries an energy
proportional to its angular frequency, E = h̄ω. Here, h̄ = h/2π is the reduced Planck constant.
Einstein proposed a literal reading of this implication—light exists as individual photons—which
provided an explanation for the photoelectric effect: each photon needs to have a frequency large
enough to have enough energy to “knock out” electrons from the metal. Increasing the intensity of
light increases the number of photons hitting the metal every second, but the actual energy that
each photon carries is proportional to its angular frequency ω.

de Broglie’s matter waves. After it was shown that light sometimes behaved like particles
(like in the photoelectric effect) and sometimes like waves (like in Young’s double slit experiment),
Louis de Broglie conjectured that the same must be true of matter. Outlined in a series of short
notes in 1923, he reasoned that matter—electrons and other objects with well-established particle
properties—should also exhibit features that are characteristic of waves. This “matter wave” has
an associated “de Broglie” wavelength λ = h/p, where p is the momentum of the object and h is
the Planck constant.
D
ou

Sc
bl

re
e

en
sl

Electrons
it

Electron
beam gun

Figure 7.4: Double slit experiment with electrons. Notice how an interference pattern is formed,
indicating the wave nature of electrons. Figure adapted from [2].

Electron interference. In 1927, the Davisson-Gerner experiment showed that electrons fired at a

99 Q C a m p 2 0 2 3
C H A P T E R 7 . T H E S C H R Ö D IN G E R E Q U A T IO N

Nickel surface exhibited Bragg diffraction peaks, which provided evidence about the hypothesised
wave nature of electrons. A more visual demonstration can be seen from the double slit electron
diffraction experiments: the 1961 Jönsson experiment with electron beams and the Merli-Missiroli-
Pozzi experiment with single electrons. These electron diffraction results cannot be explained by
classical physics, which treated electrons purely as particles.

7 .1.3 L a t e r D e v e lo p m e n t s o f Q u a n t u m T h e o ry

After Einstein’s solution of the photoelectric effect and de Broglie’s hypothesis, the next big de-
velopments were Schrödinger’s wave mechanics, and Heisenberg’s matrix mechanics. At the time,
these were two seemingly different approaches to explain the line spectrum of atoms. However, it
was later shown by Dirac and a few other contemporaries that these two approaches were actually
equivalent, and could be combined in a single formalism. This combined formalism is what we call
quantum mechanics today: one that has been very successful in explaining the microscopic world,
and one that we will dip our toes into in this chapter.

7 .2 F u lly D e s c r ib in g a P h y s ic a l S y s t e m

For a complete description of an object in a physical system, we need to know the answer to the
following questions:

• How are things (for the object)? How do we describe the state of an object? More importantly,
what are the complete set of quantities we require to describe all the possible states that the
object can have?

• What can we measure? How do we describe physical quantities that we can measure? Given
that the object is in a particular state, what would these quantities be observed to be?

• What is going on outside? How can we describe the external forces that are acting upon the
object?

• How do things change? Given the external forces acting on the object, and its initial state,
how does the state of that object evolve in time?

100 Q C a m p 2 0 2 3
C H A P T E R 7 . T H E S C H R Ö D IN G E R E Q U A T IO N

Time t′ Time t

p(t′) how the p(t)

what we measure

object is how things change

what's going
on outside x(t)
x(t′)
F(t′) F(t)

Figure 7.5: The four aspects of a physical system one needs to characterise for a complete descrip-
tion, shown here for a classical system. We will need to figure out the quantum analogues of these
aspects in order to fully describe a quantum system.

7 .2 .1 H o w T h in g s A r e : T h e S ta te o f a n O b je c t

Classical mechanics. In most of classical physics, we often model everything—a car along a
straight road, planets orbiting around each other, etc.—as a point particle: an object with a mass,
occupying some position in space, and moving with some speed. This simple description goes a
long way: a large solid body can be described as a collection of point particles with fixed relative
distances (rigid body physics), while a liquid or a gas can be described as a collection of point
particles with certain inter-particle forces (fluid mechanics).

For a system with a single particle with d degrees of (spatial) freedom, it suffices us to know its
position x ∈ Rd and momentum p = m dx
dt ∈ R to completely determine its state: “how things
d

are” for the object. Every possible pair of vectors (x, p) ∈ R2d describe every possible state of a
point particle, and we often speak about the state of a point particle belonging to a phase space
or configuration space R2d .

Quantum mechanics. In the preceding chapters, we have learnt that a quantum object is
described by the mathematical object |ψi ∈ Cd . Unlike the classical case, d here refers to the
number of distinct outcomes we can observe when a measurement is made.1

1 To be more precise, the number of distinct outcomes we can observe when a complete set of measurements
are made. What makes a set of measurements complete depends on the context. For example, when discussing
entanglement and nonlocality, although the pairs of qubits could have taken any position in space—which would
result in an infinite number of outcomes—we assumed that each qubit can only be in one of two locations. So, the
complete set of measurements are the two-dimensional qubit measurements (0/1), and the two-dimensional location
measurements (“here”/“there”). This makes four possible distinct outcomes, and hence a four-dimensional Hilbert
space C4 .

101 Q C a m p 2 0 2 3
C H A P T E R 7 . T H E S C H R Ö D IN G E R E Q U A T IO N

It bears repeating that for a quantum system, |ψi completely determines its state: “how things
are” for the object. Every possible state |ψi describes every possible state of that quantum object,
and |ψi belongs to the Hilbert space Cd .

7 .2 .2 W h a t W e M e a s u r e : P h y s ic a l O b s e r v a b le s

Classical mechanics. Physical quantities we can observe—like position and momentum—are

easily described in classical physics. If an object is in a certain state, there will be a single real
number associated to the observable “position along the x-axis”, and a statement like “x = 3 m”
makes sense. The results of “what we measure” of a classical object are simply real numbers whose
values depend on the state of the object.

Quantum mechanics. We run into problems if we attempt such value assignments to the quan-
tum formalism. After all, we know that for a given measurement with outcomes q ∈ {q1 , q2 , . . . , qn },
and associated orthonormal states {|q1 i , |q2 i , . . . , |qn i},

• the only possible outcomes are q ∈ {q1 , q2 , . . . , qn },

• if the object is in the state |ψi before the measurement, the probability of observing the
outcome q = qk is Pr(q = qk |ψ) = |hqk |ψi| ,
2

• after the observation that q = qk , the object collapses into the state |qk i, regardless of what
the state of the object was before.

So, as q does not have a definite value, it does not make sense to assign some real number to
q. However, it does make sense to speak about the statistics of q. As a first step, consider the
expectation value of q given that the system is in the state |ψi:
n
qk Pr(q = qk |ψ)
X
hqiψ =
k=1
Xn
= qk hψ|qk ihqk |ψi
k=1
n
!
X
= hψ| qk |qk ihqk | |ψi
k=1
| {z }
Q

= hψ|Q|ψi .

As you might have encountered in a previous chapter, sandwiching Q with the state |ψi returns
the expectation value of q. In fact, it can be shown that every higher-order statistics of q can also
be obtained from Q: q 2 ψ
= hψ|Q2 |ψi, q 3 ψ
= hψ|Q3 |ψi, . . . , hq m iψ = hψ|Qm |ψi, . . . .

As knowing Q provides us with all the statistics of q, Q is the quantum analogue of the physical
observable. Hence, “what we measure” of a quantum object is described by a Hermitian operator
that we call a quantum observable.

102 Q C a m p 2 0 2 3
C H A P T E R 7 . T H E S C H R Ö D IN G E R E Q U A T IO N

Exercise 7.1. Given Q = |qk ihqk |, show that Qm = |qk ihqk |. Conclude that
Pn Pn m
k=1 qk k=1 qk
hq iψ = hψ|Q |ψi, as mentioned above.
m m
Hint: remember that orthonormality implies that
hqj |qk i = 0 if j 6= k, and 1 if j = k.

Exercise 7.2. Consider the states |ψi and |ψ 0 i = eiφ |ψi. Show that hψ|Q|ψi = hψ 0 |Q|ψ 0 i. (This
means that states are unique only up to a global phase eiφ : |ψ 0 i and eiφ |ψi describe the same state,
as every physically measurable quantity cannot distinguish the two.)

Exercise 7.3. Note that when we write Q = |qk ihqk |, we are writing it in the diagonal
Pn
k=1 qk
form: each term in the sum is the eigenvalue qk multiplied by |qk ihqk |. Be careful: it is only in
a diagonal form if every “ket-bra” |qk ihqk | has the same state for the “ket” |qk i and the “bra”
hqk | = |qk i !
†

Take, for example, Q = |0ih1| + |1ih0|. It is not in the diagonal form because h0| 6= |1i for the first
†

term and h1| 6= |0i for the second term. By solving for the eigenvalues q1 , q2 and the corresponding
†

eigenstates |q1 i , |q2 i, write out Q in its diagonal form.

7 .2 .3 W h a t is G o in g o n O u t s id e : E x t e r n a l F o r c e s a s P o t e n t ia ls

Classical mechanics. When it comes to describing external influences acting upon an object,
it is often easier work with scalar potentials (assuming our system is conservative) instead of the
force-based vectorial description of Newtonian mechanics. For example, the gravitational potential
V (x) = mgz, or the potential of a sinusoidal driving force V (x, t) = F0 x cos(ωt). Then, to fully
describe “what is going on outside”, we need only specify the potential V (x, t).

Quantum mechanics. We turn to Niels Bohr’s correspondence principle, which states that
when objects in quantum mechanics get “large enough” and “energetic enough”, we should recover
classical mechanics. This means that although we can measure only the statistics hQi , Q2 , . . .
of a physical observable in a quantum system, as the systems gets very large and very energetic,
Q should start acting like a classical observable q.

In other words, for every classical observable (say, x), there is some corresponding quantum ob-
servable (say, X), which describes the same physical quantity. Hence, we can similarly describe
“what is going on outside” by specifying the same potential V (X, t), except that the potential is
now a quantum operator. Note that X here refers to the quantum observable that corresponds to
the position vector x, and not the Pauli operator.

103 Q C a m p 2 0 2 3
C H A P T E R 7 . T H E S C H R Ö D IN G E R E Q U A T IO N

7 .2 .4 H o w T h in g s C h a n g e : T im e E v o lu t io n

Classical mechanics. In classical mechanics, if you know the current state (x, p) of an object,
you can work out exactly what its state will be at a later point in time:
time evolution more time evolution
(x0 , p0 ) := (x(0), p(0)) −−−−−−−−−→ (x(t0 ), p(t0 )) −−−−−−−−−−−−→ (x(t), p(t)). (7.1)
| {z } | {z }
state at time t = 0 state at time t 0

The first equation that describes the time evolution of a

p
classical object is Newton’s second law,

dp(t) (x(0), p(0)) using

= −∇V (x, t), (7.2) Newton's
dt
equations
which you might recognise in its more fashionable form with x
0
dp(t)
dt = ma(t) and F(x, t) = −∇V (x, t).
The second equation is the almost trivial
(x(t), p(t))
dx(t) p(t)
= . (7.3)
dt m
Together, Equations (7.2) and (7.3) should be read as the Figure 7.6: The phase space (p
rules that govern the time evolution of an object in classi- against x) trajectory of an object un-
cal mechanics: the arrow in equation (7.1), or “how things der free fall. Equations (7.2) and
change” over time. They describe, mathematically, how the (7.3) govern the time evolution of
state (x(t), p(t)) should be updated. p(t) and x(t).

Exercise 7.4. Consider a rocket with the mass m = 100 kg. The thrust provided by the engine
results in a potential V (x) = −10 N · z. If the initial state of the rocket is (x(0), p(0)) with
x(0) = (2 m)j + (3 m)k and p(0) = −(200 kg m s−1 )k, what is the state of the rocket (x(t), p(t)) at
time t = 10 s?

Quantum mechanics. What about the motion of a quantum object? However the system
changes, the updated state must be another state in the same Hilbert space.
time evolution more time evolution
|ψ0 i := |ψ(t = 0)i −−−−−−−−−→ |ψ(t0 )i −−−−−−−−−−−−→ |ψ(t)i . (7.4)
| {z } | {z }
state at time t = 0 state at time t0

What we are missing here is how exactly the transformation |ψ(t0 )i → |ψ(t)i occurs.

7 .3 G u e s s in g th e Q u a n tu m T im e E v o lu t io n

Let us answer a simpler question: what type of transformation can it be? Since we know that the
transformation must bring a valid state to another valid state, it must satisfy

(7.5)
!
hψ(t0 )|ψ(t0 )i = hψ(t)|ψ(t)i

104 Q C a m p 2 0 2 3
C H A P T E R 7 . T H E S C H R Ö D IN G E R E Q U A T IO N

for any state |ψ(t)i and any time t. The meaning is clear: the probability of finding “state at
time t” at “state at time t” should be the same as the probability of finding “state at time t0 ” at
“state at time t0 ”. When we have a complete description of a quantum object and the dynamics
of the system, we expect the object to be, well, somewhere in the system. Now, this seems like a
painfully obvious statement that amounts to nothing more than an exercise in tautology, but this
requirement leads us to the demand that the transformation of a state over time must be one that
preserves the inner product of two states: a unitary transformation!

hψ(t0 )|ψ(t0 )i = hψ(t)|ψ(t)i

1

using
⟩=
|ψ

U(t,t′)
=⇒ hψ(t0 )|I|ψ(t0 )i = hψ(t0 )|U† (t, t0 )U(t, t0 )|ψ(t0 )i
|ψ(t)⟩
⟨ψ

(7.6)
= hψ(t)|ψ(t)i
|ψ( ′
t )⟩
=⇒ |ψ(t)i = U(t, t0 ) |ψ(t0 )i ,

where U(t, t0 ) is the unitary operator that brings the state

of a quantum object at time t0 to its state at time t.

Clearly, we need U(t, t) = I, since the identity operator is

the unitary operator that brings the state to itself. Then,
we consider a time interval 0 ≤ ∆t  1 so small that we (t)⟩
(t)|ψ
can linearly approximate U(t + ∆t, t) as t)⟩ ΔtW
+Δ
|ψ(t
)⟩
|ψ(t
U(t + ∆t, t) = I + ∆tW(t) + |{z}
... ,
terms containing
higher powers of ∆t
(7.7)
=⇒ |ψ(t + ∆t)i = U(t + ∆t, t) |ψ(t)i
≈ |ψ(t)i + ∆tW(t) |ψ(t)i ,

for some operator W(t) to be found. This approximation is justified because U(t + ∆t, t) → I as
∆t → 0, and the other terms would be negligible when 0 . ∆t. The inverse transformation gives

U† (t + ∆t, t) = I + ∆tW† (t) + |{z}

... ,
terms containing
higher powers of ∆t

|ψ(t)i = U† (t + ∆t, t)U(t + ∆t, t) |ψ(t)i (7.8)

≈ |ψ(t)i + ∆tW(t) |ψ(t)i + ∆tW† (t) |ψ(t)i

≈ |ψ(t + ∆t)i + ∆tW† (t) |ψ(t)i .

In order for both sides of the equation to agree, we need ∆tW† (t) |ψ(t)i = −∆tW(t) |ψ(t)i for
all |ψ(t)i. Figure 7.7 illustrates this requirement geometrically. In other words, we must have
W† (t) = −W(t), or W(t) = ih̄ H(t)
1
for some Hermitian operator H† (t) = H(t). The constant h̄ is
for dimension purposes, which will be clear later.

105 Q C a m p 2 0 2 3
C H A P T E R 7 . T H E S C H R Ö D IN G E R E Q U A T IO N

⟩ (t)⟩ ⟩ † (t)|ψ
(t)⟩
Δt) (t)|ψ t+Δ
t)
|ψ (t+ ΔtW |ψ( ΔtW

)⟩ ) ⟩
|ψ(t |ψ(t

Figure 7.7: The difference between the final and initial state over an infinitesimal
time interval is given by the vector ∆tW(t) |ψ(t)i, while the difference between the
initial and final state is ∆tW† (t) |ψ(t)i. They have to be negatives of each other for
the time evolution to be unitary.
Hence, we have
∆t
|ψ(t + ∆t)i = U(t + ∆t, t) |ψ(t)i = |ψ(t)i + H |ψ(t)i + |{z}
... ,
ih̄
terms containing
higher powers of ∆t
(7.9)
|ψ(t + ∆t)i − |ψ(t)i
=⇒ ih̄ = H |ψ(t)i + |{z}
... .
∆t
terms containing
higher powers of ∆t

By taking ∆t → 0, we arrive at the differential form of the time evolution operator

d d
ih̄ U(t) = H(t)U(t) or ih̄ |ψ(t)i = H(t) |ψ(t)i . (7.10)
dt dt
These are two common forms of the Schrödinger equation, which tells us that the evolution of a
state is dictated by the Hermitian operator H(t).

7 .3 .1 T h e H a m ilt o n ia n O p e ra to r

We know that the operator H(t) should be Hermitian, but which Hermitian operator should it be?
To figure this out, let us consider a system that consists of a single photon travelling along a fixed
path. We use
(7.11)
 T
|En i := |En (0)i =
b 0, · · · , 0, 1, 0, · · · , 0

to describe a photon with energy En at time t = 0. For simplicity, we will assume that the energies
are different: En 6= Em if n 6= m. By accounting for all possible energies {E1 , E2 , . . . , En , . . . }, we
know that {|E1 i , |E2 i , . . . , |En i , . . . } must form an orthonormal basis. What we mean by this is

106 Q C a m p 2 0 2 3
C H A P T E R 7 . T H E S C H R Ö D IN G E R E Q U A T IO N

that the outcome of measuring the energy of the system must be one of the values of En :

Pr(En |ψ)
X
1=
n
X
= hψ|En ihEn |ψi
(7.12)
n
!
X
= hψ| |En ihEn | |ψi .
n

Since this must be true for every state |ψi, we conclude that |En ihEn | = I. This is sometimes
P
n
known as the completeness relation, and is true for any chosen orthonormal basis.

Now, what do we know about the time evolution of light? We know that it must have a sinusoidal
time dependence with frequency ω. From the normalisation condition hEn (t)|En (t)i = 1, this
oscillation must be of the form

(7.13)
T
b 0, · · · , 0, e−iωn t , 0, · · · .

|En (t)i =

In addition, thanks to Planck and Einstein, we know that the angular frequency of the photon is
related to its energy by En (t) = En = h̄ωn . So,

d d
ih̄ U(t, 0) |En i = ih̄ |En (t)i = En (t) |En (t)i (7.14)
dt dt
Finally, multiplying by hEn |ψ(0)i = hEn (t)|ψ(t)i on both sides, and taking a sum over all values
of En , we have

d
U(t, 0) |En ihEn |ψ(0)i =
X X
ih̄ En (t) |En (t)ihEn (t)|ψ(t)i
n
dt n
! !
d
ih̄ U(t, 0) (7.15)
X X
|En ihEn | |ψ(0)i = En (t) |En (t)ihEn (t)| |ψ(t)i
dt n n
| {z }
I
!
d d
ih̄ U(t, 0) |ψ(0)i = ih̄ |ψ(t)i =
X
En (t) |En (t)ihEn (t)| |ψ(t)i .
dt dt n

Comparing with equation (7.10), we make the identification

H(t) := (7.16)
X
En (t) |En (t)ihEn (t)| .
n

The Hermitian operator that generates time evolution is called the Hamiltonian, which is the
physical observable that describes the energy of the system.

In the preceding discussion, we have only shown that the Hamiltonian is the correct operator that
appears on the right-hand side of the Schrödinger equation for the case of a single photon. Taking
inspiration from de Broglie’s hypothesis, which conjectures that particles would also exhibit wave
features, we can make the jump that the Hamiltonian would be the correct operator that appears
in the Schrödinger equation for every system. In every quantum system that has been studied

107 Q C a m p 2 0 2 3
C H A P T E R 7 . T H E S C H R Ö D IN G E R E Q U A T IO N

since the beginnings of quantum mechanics, the Hamiltonian has turned out to be the choice that
gives correct predictions to experimental observations.

Exercise 7.5. Consider a qubit in the state |ψi = √1 (|0i

5
+ 2 |1i) and the Hamiltonian operator
H = E0 (X + Z).

(a) Express the eigenstates of H in the computational basis {|0i , |1i}.

(b) If we were to measure the energy of the system, what are the possible values that we can
obtain? What are the probabilities of obtaining those values?

(c) Denoting the eigenstates of H as {|E1 i , |E2 i}, rewrite H in the form of equation (7.16).

7 .3 .2 S o lv in g th e S c h r ö d in g e r E q u a t io n

In a closed system where energy is conserved, H(t) = H does not depend on time. Then,
d
ih̄ U(t, 0) = HU(t, 0). (7.17)
dt
This is a differential equation with the known solution U(t, 0) = e−iHt/h̄ , where we should under-
stand the exponentiation of an operator as the Maclaurin series
∞  k
1 −it
Hk . (7.18)
X
−iHt/h̄
e =
k! h̄
k=0

Note that if we write the Hamiltonian in its diagonal form, that is, in terms of its eigenstates,

H= (7.19)
X
En |En ihEn | ,
n
th
then, from Exercise 7.1, the k power of the Hamiltonian is

Hk = (7.20)
X
Enk |En ihEn | .
n

As such, the time evolution operator is

∞  k X
−iHt/h̄
X 1 −it
e = Enk |En ihEn | .
k! h̄ n
k=0
∞  k
X X 1 −it
= Enk |En ihEn |
n
k! h̄ (7.21)
k=0
| {z }
just the scalar e−iEn t/h̄
X
= e−iEn t/h̄ |En ihEn | .
n

The time evolution of a general state |ψ(0)i is therefore

X
e−iHt/h̄ |ψ(0)i = e−iEn t/h̄ |En ihEn |ψ(0)i
(7.22)
n
X 
= hEn |ψ(0)i e−iEn t/h̄ |En i .
n

108 Q C a m p 2 0 2 3
C H A P T E R 7 . T H E S C H R Ö D IN G E R E Q U A T IO N

Exercise 7.6. Using the power series in equation (7.18), show that U(t, 0) = e−iHt/h̄ satisfies the
Schrödinger equation.

Exercise 7.7. With the Hamiltonian H from Exercise 7.5, use equation (7.21) to write U(t, 0) =
e−iHt/h̄ as a matrix using the standard representation. How long should we wait—what value
should t be—so that the time evolution U(t, 0) becomes the Hadamard gate =
b √12 11 −1 up to a
 1 

complex phase?

7 .3 .3 E x a m p le : E le c t r o n in a m a g n e t i c fi e l d

An electron has an intrinsic spin angular momentum, so-called because it invokes the image of a
classical charged sphere rotating about its centre of mass. The observable for the spin angular
momentum of an electron along the y axis is Sy = 2 Y,
h̄
where Y is the Pauli operator. Similarly,
Sx = h̄2 X and Sz = h̄2 Z, where X and Z are also Pauli operators.

From classical electromagnetism, we also know that the y-component of the magnetic moment of
a particle with mass me , charge −e, and angular momentum Sy , is µy = − 2m
e
e
Sy . Using the
correspondence principle, the quantum observable of the magnetic moment of an electron is

e gµB
µy = −g Sy = − Y.
2m 2
| {ze}
≡µB /h̄

Where µB ≈ 9.27 × 10−24 J T−1 is the Bohr magneton and g ≈ 2 is the g-factor of the electron.
The g-factor is a correction factor not present in the classical case, but is required to account for
relativistic effects.

Finally, the energy of an object with magnetic moment µy in a magnetic field with magnetic field
strength By (and Bx = Bz = 0) is E = −µy By . So, the Hamiltonian of an electron in such a
magnetic field is
gµB By
H= Y. (7.23)
2
Don’t worry if you didn’t quite get what was going on above, you can take this paragraph as
the starting point. Given that the Hamiltonian of an electron in a magnetic field is given in
equation (7.23), how does the state |ψ(0)i = |0i evolve in time?

Illustrative solution. We leave the proper solution to the reader (Exercise 7.8). Instead, we shall
take a more “illustrative” approach to gain some intuition for solving the Schrödinger equation.

To start, let us consider a general (ignoring complex phases) state |θ(0)i ≡ cos θ |0i + sin θ |1i,
which in the standard vector representation is

cos θ
 
|θ(0)i = . (7.24)
sin θ
b

109 Q C a m p 2 0 2 3
C H A P T E R 7 . T H E S C H R Ö D IN G E R E Q U A T IO N

With the same representation, we can write equation (7.23) as a matrix:

 
gµB By 0 −i
H= . (7.25)
i 0
b
2

Let us consider how the state |θ(0)i will evolve over a tiny step in time ∆t. From equation (7.9),

∆t
|θ(∆t)i ≈ |θ(0)i + H |θ(0)i
 ih̄
cos θ gµB By ∆t 0 −1 cos θ
   
=
b
sin θ
+
2h̄ 1 0 sin θ (7.26)

cos θ gµB By ∆t − sin θ

   
= + .
sin θ 2h̄ cos θ
 − sin θ 
Geometrically, is a vector that is always perpendicular to sin θ , regardless of the value
 cos θ 
cos θ
of θ. We recognise this as a tiny anticlockwise rotation of angle ∆θ = . So,
gµB By ∆t
2h̄

|θ(∆t)i = U(∆t, 0) |θ(0)i


gµB By
 
gµB By
 (7.27)
(probably) = cos θ + ∆t |0i + sin θ + ∆t |1i .
2h̄ 2h̄

A finite time interval t can be broken down into many infinitesimal time steps ∆t to give
t/∆t times

|θ(t)i = U(∆t, 0) · · · U(∆t, 0) |θ(0)i

z }| {
    (7.28)
gµB By gµB By
(probably) = cos θ + t |0i + sin θ + t |1i .
2h̄ 2h̄

Therefore, we recognise the time evolution as a uniform rotation with angular frequency ω = 2h̄ .
gµB By

Hence, given |ψ(0)i = |0i, the state at some later time is

|ψ(t)i = U(∆t, 0) |0i

(7.29)
(probably) = cos(ωt) |0i + sin(ωt) |1i .

Exercise 7.8. In this exercise, we will solve the Schrödinger equation in the “proper” way, and
double-check the intuitive approach we took above.

(a) Rewrite equation (7.23) in its diagonal form (i.e. in the form of equation (7.19)).

(b) Then, using equation (7.22), solve for |ψ(t)i with the state |ψ(0)i = |0i. Verify that equa-
tion (7.29) is correct.

(c) What is the probability of measuring |0i at time t?

Exercise 7.9. Repeat Exercise 7.8 with the Hamiltonian H = gµB Bz

2 Z. What do you notice
about the probability of measuring |0i over time?

110 Q C a m p 2 0 2 3
C H A P T E R 7 . T H E S C H R Ö D IN G E R E Q U A T IO N

7 .4 S u m m a ry

Historical development of Quantum Mechanics

Quantum mechanics was developed to address the inadequacies of classical mechanics,
which failed to explain certain experimental phenomena such as the photoelectric effect
and electron diffraction. An important concept is how quantum objects have both wave-
like and particle-like properties, and many principles are extensions from the behaviour
of the photon, which are individual packets of light.

Fully Describing a Physical System

What we require for a full description of a physical system in the classical and quantum
cases are

Classical Quantum
State of an object (x, p) |ψi
Physical observables Real scalars x, p Hermitian operators X, P
Potentials V (x, t) V (X, t)
p(t)
Time evolution dp(t)
dt = −∇V (x, t), dx(t)
dt = m
d
ih̄ dt |ψ(t)i = H(t) |ψ(t)i

The Schrödinger equation and the Hamiltonian operator H

The Schrödinger equation
d
ih̄ |ψ(t)i = H(t) |ψ(t)i
dt
is the rule that governs the dynamics (i.e the time evolution) of a quantum system,
similar to how Newton’s laws govern the dynamics of a classical system. It involves the
Hamiltonian H(t), the quantum observable that describes the energy of the system.

The unitary time evolution operator

Since quantum states must always be normalised, the dynamics of the quantum system
must preserve the inner product. Hence, the time evolution of a system is described by
a unitary operator
|ψ(t)i = U(t, 0) |ψ(0)i .

For time independent H(t) = H, the time evolution operator is given by

U(t, 0) = e−iHt/h̄ ,

where we understand the exponentiation of an operator H in terms of the Maclaurin

series expansion of the exponential. If H is written in terms of its eigenvalues—so,

111 Q C a m p 2 0 2 3
C H A P T E R 7 . T H E S C H R Ö D IN G E R E Q U A T IO N

H=
P
n En |En ihEn |—then

U(t, 0) =
X
e−iEn t/h̄ |En ihEn | .
n

112 Q C a m p 2 0 2 3
C H A P T E R 7 . T H E S C H R Ö D IN G E R E Q U A T IO N

7 .5 C o n t in u o u s C a s e o f th e S c h r ö d in g e r E q u a t io n

This topic is a little advanced, hence its relegation to the appendix of the current chapter, and
meant only for the interested reader. That said, any discussion about the Schrödinger equation
is incomplete without touching on the continuous case, even if the discussion were minimal and
hand-wavy like the one in this section. As such, feel free to refer to more complete textbooks on
this topic.

The continuous case of the Schrödinger equation is required when a state |ψi of a quantum system
requires continuous—uncountably infinite—degrees of freedom, like a particle in one-, two-, or
three-dimensional space.

7 .5 .1 O n e - d im e n s io n a l D is c r e t e S p a c e

We first consider a quantum particle that occupies a one-dimensional lattice with a spacing ∆x:

… (k−3)Δx (k−2)Δx (k−1)Δx kΔx (k+1)Δx (k+2)Δx (k+3)Δx …

x

Δx

That is, if we measure the position of the object, the only possible outcomes are x ∈ {k∆x}∞
k=−∞ ,
where k is an integer. Hence, the post-measurement states |ki ∈ {. . . , |−2i , |−1i , |0i , |1i , |2i , . . . }
are orthonormal, and the position observable X (position observable, not Pauli operator) is
∞
X= (7.30)
X
k∆x |kihk| .
k=−∞

We can also use {|ki}∞

k=−∞ as an orthonormal basis to write a general state |ψi of the system

∞
(7.31)
X
|ψi = ψk |ki ,
k=−∞

with the coefficient ψk = hk|ψi.

Just like before, ψk is a complex number such that |ψk | is the probability of finding the particle
2

at x = k∆x. The only change is that, somehow, the Hilbert space is infinite dimensional. So, the
position representation of |ψi ∈ C∞ is, somehow, an infinite dimensional vector
 . 
.
 . 
ψk−1 
(7.32)
 
|ψi =
b ψk  .

ψk+1 
..
 
.

113 Q C a m p 2 0 2 3
C H A P T E R 7 . T H E S C H R Ö D IN G E R E Q U A T IO N

While there might be a myriad of practical problems—say, the fact that we only have a finite
amount of ink to write down the vector—there is nothing preventing us from defining an infinite
dimensional vector, so long as we keep the same rules of matrix multiplication and other vector
operations. For example, the inner product between two states is defined in the same way, except
with an infinite sum: ∞
(7.33)
X
hφ|ψi = φ∗k ψk .
k=−∞

An important rule we have to keep for such an infinite-dimensional state to make physical sense is
to keep them normalised:
∞
(7.34)
X 2
hψ|ψi = |ψk | = 1.
k=−∞

Exercise 7.10. A quantum object is in the state |ψi. Using the Born rule, what is the probability
of finding the position of that object in the interval a ≤ x ≤ b, where a = ka ∆x, b = kb ∆x, and
ka < k b ?

7 .5 .2 O n e - d im e n s io n a l C o n t in u o u s S p a c e

What about about the case where the one-dimensional space is continuous? We can simply take
∆x → 0: a line is, after all, an infinite number of points that are infinitely close to each other.

x
Δx → 0

Before we actually do that, it is helpful to rewrite the following

∞
|ki hk|
X= (position observable), (7.35)
X
∆x k∆x √ √
k=−∞
∆x ∆x
∞
ψk |ki
(general state), (7.36)
X
|ψi = ∆x √ √
k=−∞
∆x ∆x
∞
φ∗ ψk
(inner product), (7.37)
X
hφ|ψi = ∆x √ k √
k=−∞
∆x ∆x
∞
ψ∗ ψk
(normalisation condition). (7.38)
X
hψ|ψi = ∆x √ k √ =1
k=−∞
∆x ∆x

As we take ∆x → 0, the discrete sum approaches the integral dx. At the same
P∞ R∞
k=−∞ ∆x → −∞

114 Q C a m p 2 0 2 3
C H A P T E R 7 . T H E S C H R Ö D IN G E R E Q U A T IO N

√
time, since x has replaced k as the index, we need to redefine2 k∆x → x, ψk / ∆x → ψ(x), and
√
|ki / ∆x → |xi. With this, we have
Z ∞
X= dx x |xihx| (position observable), (7.39)
−∞
Z ∞
|ψi = dx ψ(x) |xi (general state), (7.40)
−∞
Z ∞
hφ|ψi = dx φ∗ (x)ψ(x) (inner product), (7.41)
−∞
Z ∞
dx |ψ(x)| = 1 (normalisation condition). (7.42)
2
hψ|ψi =
−∞

Now, instead of an infinitely long vector with entries {ψk }∞

k=−∞ to describe our state in the position
representation, we have a function ψ(x). The requirement that the state is normalisable translates
to the integral of |ψ(x)| from −∞ to ∞ being one. For obvious reasons, a function like ψ(x) such
2

that −∞ dx |ψ(x)| < ∞ is called square integrable.

R∞ 2

Exercise 7.11. A quantum object is in the state |ψi = dx ψ(x) |xi. Using the Born rule,
R∞
−∞
what is the probability of finding the position of that object in the interval a ≤ x ≤ b, written as
an integral in terms of the function ψ(x)?

Exercise 7.12. Show that the function

q
 2 cos πx , x ∈ − L2 , L2

 
L L
ψ(x) =
0, otherwise


is square integrable and normalised.

7 .5 .3 T h e M o m e n tu m O p e ra to r

To utilise the correspondence principle in the continuous case, apart from the position observable
X, we also require a momentum observable P. Again, we appeal to the behaviour of light. Consider
a single photon with definite momentum p,
Z ∞
|φp (t)i = dx φp (x, t) |xi . (7.43)
−∞

Because it has a definite momentum, every time we measure its momentum, we will always measure
it to be p. That is, |φp (t)i is an eigenstate of P:
Z ∞
P |φp (t)i = dx {p φp (x, t)} |xi . (7.44)
−∞

2 This is a sloppy—though hopefully well-motivated—approach for taking the limit. There are proper ways to
do the same, but such technical details are beyond the scope of these notes.

115 Q C a m p 2 0 2 3
C H A P T E R 7 . T H E S C H R Ö D IN G E R E Q U A T IO N

Now, how does φp (x, t) depend on p? Firstly, we know that travelling light waves have the position
and time dependence in a form like cos 2πx
λ − ωt and sin λ − ωt . In addition, from the time
2πx

dependence of a photon, it must be that φp (x, t) ∝ ei(2πx/λ−ωt) . Now, from the de Broglie relation
λ = hp , we have φp (x, t) ∝ ei(xp/h̄−ωt) . Finally, to eliminate the dependence on p in equation (7.44),
we note that ∂
∂x φp (x, t) = ip/h̄. Hence,
Z ∞  
h̄ ∂
P |φp (t)i = dx φp (x, t) |xi . (7.45)
−∞ i ∂x

We use equation (7.45) as the defining equation for the momentum operator in the position repre-
sentation: that is, we guess that the momentum operator acts on every object in the same way it
acts on a photon. So, Z ∞  
h̄ ∂
P |ψ(t)i = dx ψ(x, t) |xi . (7.46)
−∞ i ∂x
This turns out to be the correct choice for the momentum operator.

Exercise 7.13. For a general state |ψi, work out X(P |ψi) and P(X |ψi). Hence, conclude that
[X, P] = XP − PX = ih̄I. This is the known as the canonical commutation relation, which can serve
as another possible starting point to derive the position representation of the momentum operator.

7 .5 .4 T h e S c h r ö d in g e r ` ` W a v e '' E q u a t io n

Now that the position and momentum operators are defined for the continuous case, we can define
the Hamiltonian operator. Since the energy in a classical system is E(t) = p2 /2m + V (x, t),
invoking the correspondence principle,

P2
H(t) = + V (X, t). (7.47)
2m
Then, the Schrödinger equation is

P
 2 
d
|ψ(t)i = + V (X, t) |ψ(t)i
ih̄
dt 2m
Z ∞ Z ∞ (7.48)
h̄2 ∂ 2
   
∂
dx ih̄ ψ(x, t) |xi = dx − ψ(x, t) + V (x, t)ψ(x, t) |xi .
−∞ ∂t −∞ 2m ∂x2

As we expanded |ψ(t)i in the position representation on both sides, equation (7.48) can be thought
of as an equation that relates the coefficients in the braces on the left and the coefficients in the
braces on the right. Isolating just those terms,

∂ h̄2 ∂ 2
ih̄ ψ(x, t) = − ψ(x, t) + V (x, t)ψ(x, t). (7.49)
∂t 2m ∂x2
This is another popular format of the Schrödinger equation, which is reminiscent of wave equations
in classical physics. For this reason, equation (7.49) is sometimes called the “Schrödinger wave
equation”, and ψ(x, t) the “wave function”.

116 Q C a m p 2 0 2 3
C H A P T E R 7 . T H E S C H R Ö D IN G E R E Q U A T IO N

Solving the Schrödinger equation for the continuous case for time-independent potentials V (X, t) =
V (X) is the same as the discrete case: one finds the eigenstates of the Hamiltonian H |En i =
En |En i, then uses equation (7.22) to find |ψ(t)i.

Exercise 7.14. Show that the state |ψi = dx ψ(x) |xi with the function ψ(x) from Exer-
R∞
−∞
cise 7.12 is an eigenstate of the Hamiltonian

P2
H=
2m
within the “box” x ∈ − L2 , L2 . What is the energy of that state?
 

117 Q C a m p 2 0 2 3
C H A P T E R 7 . T H E S C H R Ö D IN G E R E Q U A T IO N

E x e r c is e S o lu t io n s

Solution 7.1. We use a proof by induction. Assume Qm = |qk ihqk | is true. Then,
Pn m
k=1 qk

Qm+1 = Qm Q
n
! n 
X X
= qkm |qk ihqk |  qj |qj ihqj |
k=1 j=1
n
X n
X 
= qkm |qk i qj hqk |qj i hqj |
k=1 j=1
| {z }
0 if j 6= k, 1 if j = k
| {z }
only j = k term survives
X
= qkm |qk i (qk hqk |)
k=1
X
= qkm+1 |qk ihqk | ,
k=1

which proves the inductive step. Since the base case Q1 = |qk ihqk | is true, Qm =
Pn 1
k=1 qk

k=1 qk |qk ihqk | is true for all m ≥ 1.

Pn m

Solution 7.2. A direct application of |ψ 0 i = eiφ |ψi and hψ 0 | = |ψ 0 i = e−iφ hψ| gives us
†

hψ 0 |Q|ψ 0 i = e−iφ hψ| Q eiφ |ψi

= e−iφ eiφ hψ|Q|ψi

= hψ|Q|ψi .

Solution 7.3. We write Q as a matrix using the standard representation |0i =

b [ 10 ], |1i =
b [ 01 ]:

Q = |0i (h1|) + |1i (h0|)

† †

   †    †
1 0 0 1
= +
0 1 1 0
   
1   0  
= 0 1 + 1 0
0 1
 
0 1
= .
1 0

The eigenvalues qk must each satisfy the characteristic equation det[Q − qk I] = 0:

det[Q − qk I] = 0
−qk 1
= qk2 − 1 = 0
1 −qk

118 Q C a m p 2 0 2 3
C H A P T E R 7 . T H E S C H R Ö D IN G E R E Q U A T IO N

=⇒ qk = 1 or −1.
 
ak
Let us call q1 = 1 and q2 = −1. We wish to find the eigenstates |qk i = . Remembering from
bk
b
the previous exercise that global phases represent the same state, we can take ak > 0 to be a real
number. Then, from the normalisation condition hqk |qk i = 1,
2 2
1 = |ak | + |bk |
q
=⇒ ak = 1 − |bk | .
2

Then, we consider the cases q1 and q2 separately. For q1 = 1,

Q |q1 i = q1 |q1 i
  "q 2
# "q
2
#
0 1 1 − |b1 | = 1 − |b1 |
1 0 b1 b1
q
=⇒
2
1 − |b1 | = b1
1
∴ b1 = √ .
2
Meanwhile, for q2 = −1,

Q |q2 i = q2 |q2 i
  "q 2
# "q
2
#
0 1 1 − |b2 | = − 1 − |b2 |
1 0 b2 b2
q
=⇒
2
1 − |b2 | = −b2
1
∴ b2 = − √ .
2
Hence, the eigenstates are
   
1 1 1 1
b √
|q1 i = = |+i b √
|q2 i = = |−i .
2 1 2 −1
Now, we can write Q in the diagonal form

Q = q1 |q1 ihq1 | + q2 |q2 ihq2 |

= |+ih+| − |−ih−| .

Solution 7.4. First, we use equation (7.2) to solve for p(t). We have
dp(t)
= −∇V (x) = 10 N
dt
Z p(t) Z t
dp = dt0 (10 N)k
p(0) 0 (7.50)
∴ p(t) = 10 N · t + p(0)
= 10 N · t − 200 kg m s−1 k.

119 Q C a m p 2 0 2 3
C H A P T E R 7 . T H E S C H R Ö D IN G E R E Q U A T IO N

Then, using equation (7.3), we have

dx(t) 1 dp
=
dt  dt
m
10 N · t − 200 kg m s−1

= k
100 kg
= 0.1 m s−2 · t − 2 m s−1 k


Z x(t) Z t (7.51)
dx = dt0 0.1 m s−2 · t − 2 m s−1 k


x(0) 0
t2
 
∴ x(t) = 0.1 m s−2 · − 2 m s−1 · t k + x(0)
2
= (2 m)j + 0.05 m s−2 · t2 − 2 m s−1 · t + 3 m k.

Therefore, at t = 10 s, we have (x(10 s), p(10 s)) with

x(10 s) = (2 m)j + (483 m)k,

(7.52)
p(10 s) = (8 m s−1 )k

Solution 7.5.

(a) First, we find the matrix representation of our Hamiltonian H in the standard representation
b [ 10 ] and |1i =
|0i = b [ 01 ]. From the previous chapters, we know that
   
0 1 1 0
X= , Z=
1 0 0 −1
b b

which gives  
1 1
H=
b E0 .
1 −1

(b) The only possible outcomes when we measure the energy of the system are the eigenvalues
En of H. We find the eigenvalues En by solving the characteristic equation

det[H − En I] = 0
E0 − En E0
= En2 − 2E02 = 0
E0 −E0 − En
√ √
=⇒ En = 2E0 or − 2E0 .
√ √
Therefore, the possible values of energy we would measure are E1 = 2E0 and E2 = − 2E0 .

The outcome probability of measuring En is given by the Born rule Pr(En |ψ) = |hEn |ψi| .
2

To find the eigenstates |En i, we repeat the steps from Exercise 7.3 and write
  "q 2
#
an 1 − |b n |
|En i = =
bn
b
bn

120 Q C a m p 2 0 2 3
C H A P T E R 7 . T H E S C H R Ö D IN G E R E Q U A T IO N

by taking an > 0 real, and imposing the normalisation condition hEn |En i. Then, we consider
√
the first case E1 = 2E0

H |E1 i = E1 |E1 i
"q # "p #
  2 √ 2
1 1 1 − |b1 | 1 − |b 1 |
E0 = 2E0
1 −1 b1 b1
q √
=⇒
2
1 − |b1 | − b1 = 2b1
1
=⇒ b1 = p √ .
4+2 2
√
For the case E2 = − 2E0 ,

H |E2 i = E2 |E2 i
 "q # "p
2
#
1 − |b2 | = −√2E
2

1 1 1 − |b2 |
E0 0
1 −1 b2 b2
q √
=⇒
2
1 − |b1 | − b1 = − 2b1
1
=⇒ b1 = − p √
4−2 2
Therefore, we have
√  √ 
1 2+1 1 2−1
|E1 i = √ |E2 i = √ ,
−1
b p b p
4+2 2 1 4−2 2
and the outcome probabilities are
2
√ !†   
1 2+1 1 1
Pr(E1 |ψ) = |hE1 |ψi| =
2
p √ √
1 5 2
b
4+2 2
√
2
2+3
= √
≈ 0.571
5 4+2 2
2
√ !†   
1 2−1 1 1
Pr(E2 |ψ) = |hE2 |ψi| =
2
√ √
−1
p
5 2
b
4−2 2
√
2
2−3
= √
≈ 0.429.
5 4−2 2

(c) The diagonal form of the Hamiltonian is simply

√
H = E1 |E1 ihE1 | + E2 |E2 ihE2 | = 2E0 (|E1 ihE1 | − |E2 ihE2 |)

Solution 7.6. We simply perform the derivative on the power series:

∞  k
d 1 −it
U(t, 0) = Hk
X
dt k! h̄
k=0

121 Q C a m p 2 0 2 3
C H A P T E R 7 . T H E S C H R Ö D IN G E R E Q U A T IO N

∞  k−1  
1 −it −i
Hk
X
= k
k! h̄ h̄
k=0
∞  k−1
1 X 1 −it
= H Hk
ih̄ (k − 1)! h̄
k=1
1
= HU(t, 0)
ih̄
1 d
∴ ih̄ U(t, 0) = HU(t, 0).
ih̄ dt

Solution 7.7. From our solution to Exercise 7.5, the diagonal form of the Hamiltonian is
√
H= 2E0 (|E1 ihE1 | − |E2 ihE2 |).

Then, using equation (7.21), the time evolution operator is

√ √
U(t, 0) = e−i 2E0 t/h̄
|E1 ihE1 | + ei 2E0 t/h̄ |E2 ihE2 |
√ √
e−i 2E0 t/h̄  e−i 2E0 t/h̄
√  √ 
2 + 1 √ 2 − 1 √ 
= √ 2+1 1 + √ 2 − 1 −1
1 −1
b
4+2 2 4−2 2
" √
2 √
√
√
#
1 √
2 + 1 2 − 1 2 + 1 2 − 1
= √
√
√
√
√
e−i 2E0 t/h̄
4+2 2 4−2 2 2+1 2−1 2−1
" √
2 √
√
√
# √ !
2 − 1 2 + 1 − 2 − 1 2 + 1
+ √
√
√
ei 2E0 t/h̄
− 2−1 2+1 2+1
!
√
√  √  √
1 2+1 √ 1 −i 2E0 t/h̄ 2 − 1 √ −1 i 2E0 t/h̄
= √ e + e
2 2 1 2−1 −1 2+1
√ √  √  √  
1 2 cos 2E0 t
h̄ − i sin 2E0 t
−i sin 2E0 t
= √  √  h̄ √ √  h̄  √  .
2 −i sin 2E h̄
0t
2 cos 2E0 t
h̄ + i sin 2E h̄
0t

√
If we choose the time 2E0 t
h̄ = π/2, then
     
πh̄ 1 0−i −i 1 1 1
U t= √ ,0 = √ = −i × √ ,
2 2E0 2 −i 0+i 2 1 −1

which we recognise as the Hadamard gate, up to the complex phase e−iπ/2 . This is how specific
unitary gates like the Hadamard gate, or the Pauli gates, are performed in the lab: some instrument
(usually an electromagnetic pulse) changes the energy of the system so that it corresponds to the
relevant Hamiltonian, the system is allowed to evolve for a fixed amount of time, after which the
instrument is deactivated. In this case, we need to ensure that the instrument was activated for a
time t = √πh̄ .
2 2E0

122 Q C a m p 2 0 2 3
C H A P T E R 7 . T H E S C H R Ö D IN G E R E Q U A T IO N

Solution 7.8.

(a) Using the same steps we performed in Exercise 7.3, we can work out that
   
1 1 1 1
Y = |+yih+y| − |−yih−y| , where |+yi =b √ , b √
|−yi = .
2 i 2 −i
Then, equation (7.23) can be rewritten as
   
gµB By gµB By
H= |+yih+y| + − |−yih−y| .
2 2

We can identify the energy eigenstates as |E± i = |±yi with the eigenvalues E± = ± .
gµB By
2

(b) Using equation (7.22) with E± = ± B2 y and |ψ(0)i = |0i,

gµ B

   
|ψ(t)i = hE+ |ψ(0)i e−iE+ t/h̄ |E+ i + hE− |ψ(0)i e−iE− t/h̄ |E− i
gµB By gµB By
   
= hy+ |0i e−i 2h̄ t |y+ i + hy− |0i ei 2h̄ t |y− i

The coefficients include the inner product

 
1   1 1
h±y|0i = √ 1 ∓i =√ .
2 0 2

Placing this back in to our expression, and defining ω := 2h̄ ,

gµB By

e−iωt eiωt
|ψ(t)i = √ |y+ i + √ |y− i
2 2
cos(ωt) − i sin(ωt) 1 1 cos(ωt) + i sin(ωt) 1
   
1
= √ ·√ + √ ·√
2 i 2 −i
b
2 2
   
1 0
= cos(ωt) + sin(ωt)
0 1
b cos(ωt) |0i + sin(ωt) |1i .
=

Indeed, equation (7.29) is correct.

(c) Using the Born rule,

Pr(0|ψ(t)) = |h0| (cos(ωt) |0i + sin(ωt) |1i)| = cos2 (ωt).

2

Solution 7.9. The diagonal form of H = gµB Bz

2 Z is

H = h̄ω |0ih0| − h̄ω |1ih1| .

Then, given the initial state |ψ(0)i = |0i, the state of the system at time t is

|ψ(t)i = h0|0i e−iωt |0i + h1|0i eiωt |1i = e−iωt |0i .

The probability of measuring |0i at a later time is

Pr(0|ψ(t)) = h0| e−iωt |0i

2
= 1.

123 Q C a m p 2 0 2 3
C H A P T E R 7 . T H E S C H R Ö D IN G E R E Q U A T IO N

Hence, we will always find the system to be in the |0i state. This is simply a statement about the
conservation of energy: because energy is conserved in an isolated system, once you prepare the
system in a certain state of the energy, it remains in that state. For this reason, energy eigenstates
are sometimes called “stationary states”.

Solution 7.10. Given |ψi = ψk |ki, the probability that the object is in the interval
P∞
k=−∞
ka ∆x = a ≤ x ≤ b = kb ∆x is the sum of the probabilities of the object being in the position x
that are in this range:
kb kb
Pr(a ≤ x ≤ b|ψ) = Pr(x|ψ) =
X X 2
X 2
|hk|ψi| = |ψk | .
a≤x≤b k=ka k=ka

Solution 7.11. We rewrite our answer from Exercise 7.10 in the form of equations (7.35)—(7.38):
kb
ψ ∗ ψk
Pr(a ≤ x ≤ b|ψ) =
X
∆x √ k √ .
k=ka
∆x ∆x

Recognising how we went from equation (7.38) to equation (7.42), we take dx

Pkb R xb
k=ka ∆x → xa
√
and ψk / ∆x → ψ(x), which results in
Z xb
Pr(a ≤ x ≤ b|ψ) = dx |ψ(x)| .
2

xa

Solution 7.12. The normalisation condition requires that hψ|ψi = 1. Substituting the function
from the exercise,
Z ∞
dx |ψ(x)|
2
hψ|ψi =
−∞
Z −L
2 2
dx cos πx



= L
L −L
2

−L πx πx 2
ei L + e−i L
Z 
2 2
= dx
L −L
2
2
Z −L
1 2
dx ei2 L + e−i2 L + 2
πx πx

=
2L −L
2
 πx πx  L2
1 ei2 L e−i2 L
= π + π + 2x
2L i2 L −i2 L −L 2

= 1.

Hence, ψ(x) is normalised, and of course, square integrable.

124 Q C a m p 2 0 2 3
C H A P T E R 7 . T H E S C H R Ö D IN G E R E Q U A T IO N

Solution 7.13. (Please note that there was a typo in the printed copy which made this question
unsolvable.) Let us start with X(P |ψi).
Z ∞   
h̄ ∂
X(P |ψi) = X dx ψ(x) |xi
−∞ i ∂x
Z ∞  
h̄ ∂
= dx ψ(x) X |xi
−∞ i ∂x | {z }
|xi is an eigenstate of X with eigenvalue x
Z ∞ 

h̄ ∂
= dx
ψ(x) x |xi
−∞ i ∂x
Z ∞  
h̄ ∂
= dx x ψ(x) |xi .
−∞ i ∂x

Be careful about what the differential operator acts upon: by definition of the momentum operator,
the partial derivative acts only on ψ(x, t) in this case. So, when we drag the x into the bracket,
we place it on the left-hand side to avoid any ambiguity.

As for P(X |ψi), we shall take a little more care. Define |ψ 0 i := X |ψi. Then,
Z ∞
|ψ 0 i = dx ψ 0 (x) |xi
−∞
Z ∞
:= X |ψi = dx {xψ(x)} |xi .
−∞

We make the identification ψ 0 (x) = xψ(x). Then,

P(X |ψi) = P |ψ 0 i
Z ∞  
h̄ ∂ 0
= dx ψ (x) |xi
−∞ i ∂x
Z ∞  
h̄ ∂
= dx [xψ(x)] |xi
−∞ i ∂x
Z ∞  
h̄ h̄ ∂
= dx ψ(x) + x ψ(x) |xi
−∞ i i ∂x
Z ∞ Z ∞  
h̄ h̄ ∂
= dx ψ(x) |xi + dx x ψ(x) |xi .
i −∞ −∞ i ∂x
| {z } | {z }
|ψi X(P|ψi)

Rearranging, we have

h̄
X(P |ψi) − P(X |ψi) = − |ψi
i
[X, P] |ψi = ih̄ |ψi .

Since the above steps are valid for any arbitrary state |ψi, we conclude that [X, P] = ih̄I.

Solution 7.14. The action of the Hamiltonian upon the state |ψi is
∞
h̄2 ∂ 2
Z  
H |ψi = − ψ(x) |xi
−∞ 2m ∂x2

125 Q C a m p 2 0 2 3
C H A P T E R 7 . T H E S C H R Ö D IN G E R E Q U A T IO N

L
h̄2 ∂ 2
Z  q 
2

L cos L
2 πx


= − |xi
−L 2
2m ∂x2
Z L2  2  2 q 
h̄ π
L cos L
2 πx


= |xi
−L 2
2m L
Z L2 q
h̄2 π 2

L cos L
2 πx


= |xi
2mL2 − L2
| {z }
|ψi
2 2
h̄ π
= |ψi .
2mL2

Hence, |ψi is an eigenstate of H with energy E = 2mL2 .

h̄2 π 2

R e f e r e n c e s f o r t h is C h a p t e r

[1] Huygens–Fresnel principle - Wikipedia. https://en.wikipedia.org/wiki/Huygens%E2%80%

93Fresnel_principle (accessed May 24, 2021).

[2] Double-slit experiment - Wikipedia. https://en.wikipedia.org/wiki/Double-slit_experiment

(accessed May 24, 2021).

[3] Light Is An Electromagnetic Wave - Light Electromagnetic Wave Transparent PNG - Free
Download on NicePNG. https://www.nicepng.com/ourpic/u2w7y3r5i1i1w7w7_light-is-an-e
lectromagnetic-wave-light-electromagnetic-wave/ (accessed May 24, 2021).

126 Q C a m p 2 0 2 3
C h a p te r 8

E n t a n g le m e n t a n d N o n lo c a lit y

This chapter explores what happens if we have a quantum state consisting of not just one, but two
qubits. The first section introduces these two-qubit states and the phenomenon of entanglement.
Teleportation and no-cloning are covered as well. Then, the second section explores the counter-
intuitive behaviour of entanglement further by introducing the concept of Bell nonlocality.

8 .1 E n t a n g le m e n t a n d T e le p o r t a t io n

Lecture notes by Gan Beng Yee

8 .1.1 In t r o d u c t io n

By now, you should have a fair understanding on the quantum mechanical properties of a qubit
and how it differs from a classical object. Will there be any new quantum phenomenon if we
consider more than one qubit? In this section, we shall look at the physics of multi-qubits quantum
system and understand how a more peculiar quantum effect: quantum entanglement emerges from
the mathematical structure of Kronecker product and the principle of superposition. This exotic
phenomenon will serves as a resource that enables the transmission of quantum information between
two places.

Ta k e a w a y s

By the end of this section, the reader should:

• Know how to construct a general two-qubit quantum state and perform transformations on
quantum states using two-qubit quantum operations.

127
C H A P T E R 8 . E N TA N G L E M E N T A N D N O N L O C A L IT Y

• Able to differentiate entangled quantum states from separable quantum states and comment
on the properties of entangled quantum states.

• Understand the non-cloning property of non-orthogonal quantum states.

• Gain intuitions on the working principle of quantum teleportation protocol and explains its
significant.

8 .1.2 T w o - q u b it Q u a n t u m S y s te m s

Before going deeply into quantum entanglement and teleportation, we have to first introduce
the mathematical structure of multi-qubits quantum systems. For simplicity, we only consider
two-qubits systems in this section1 . We will connect the linear algebra’s concept, i.e: Kronecker
product, you have learned in the previous lecture to the physics of two-qubits quantum systems.

8 .1.3 M a t h e m a t ic a l S t r u c t u r e o f T w o - q u b it Q u a n t u m S ta te s

D ir a c r e p r e s e n t a t io n o f t w o - q u b it q u a n t u m s ta te s

Imagine that each lab in CQT NTU and CQT NUS holding one qubit with the respective com-
putation basis {|0iNTU , |1iNTU } and {|0iNUS , |1iNUS }2 . To describe their joint quantum state, we
have to introduce a special mathematical structure called Kronecker product, i.e: ⊗. If the qubit
in NTU is in quantum state |0iNTU and qubit in NUS is in quantum state |0iNUS , then the joint
quantum state of these two qubits is

|ψitot = |0iNTU ⊗ |0iNUS .

In fact, there are four different combinations that we could consider, which are

|0iNTU ⊗ |0iNUS , |0iNTU ⊗ |1iNUS , |1iNTU ⊗ |0iNUS , |1iNTU ⊗ |1iNUS

To simplify the notation, we drop the label “NTU” and “NUS” and we use the sequence of qubits
as the label, i.e: the first (second) qubit always denotes the qubit in CQT NTU’s (NUS) lab

|iiNTU ⊗ |jiNUS := |ii ⊗ |ji := |iji ,

where i, j ∈ {0, 1}. We will be using these notations interchangeably. With the simplified notation,
we then have
|00i , |01i , |10i , |11i (8.1)

1 The generalization to multi-qubits systems is straightforward where you just need to apply Kronecker product
multiple times to combine quantum states of multiple qubits into one joint quantum state.
2 The qubits could be in the same lab/place and the label of NTU and NUS can be changed to something else.

128 Q C a m p 2 0 2 3
C H A P T E R 8 . E N TA N G L E M E N T A N D N O N L O C A L IT Y

as the computation basis for two-qubit quantum systems. By invoking the principle of superposi-
tion, we could construct arbitrary two-qubit quantum states

|ψitot = α |00i + β |01i + γ |10i + σ |11i . (8.2)

where α, β, γ, σ ∈ C and |α|2 + |β|2 + |γ|2 + |σ|2 = 1, i.e: the quantum states are normalised.

Let’s imagine now the qubits in CQT NTU’s and CQT NUS’s labs are in their respective general
quantum state

|ψiNTU := a |0iNTU + b |1iNTU and |ψiNUS := c |0iNUS + d |1iNUS

with a, b, c, d ∈ C, |a|2 + |b|2 = 1 and |c|2 + |d|2 = 1. Their joint quantum state |ψitot is

|ψitot := |ψiNTU ⊗ |ψiNUS .

Again, to simplify of the notation, we drop the label “NTU” and “NUS” and the first (second)
qubit always denotes the qubit in CQT NTU’s (NUS) lab. The action of the Kronecker product
is identical to the standard multiplication operation, i.e: ×, where we cross multiple each of the
elements between two blocks of variables

|ψitot = (a |0i + b |1i) ⊗ (c |0i + d |1i)

= ac |0i ⊗ |0i + ad |0i ⊗ |1i + bd |1i ⊗ |0i + bd |1i ⊗ |1i
= ac |00i + ad |01i + bc |10i + bd |11i . (8.3)

The expression in equation (8.2) and equation (8.3) are the same when we have

α = ac, β = ad, γ = bc, σ = bd.

However, there is one special class of quantum state, i.e: entangled quantum states, in which the
above relation can never be found. We will study this special class of quantum state in section 8.1.5.
The joint quantum state |ψitot is a quantum state that lives in a complex space of four dimensions,
i.e: |ψitot ∈ C2 ⊗ C2 . The Kronecker product combines two 2-dimensional complex vectors into
one 4-dimensional complex vector, allowing us to describe the quantum states of the two-qubit
system with a simple but elegant structure.

V e c t o r r e p r e s e n t a t io n o f t w o - q u b it q u a n t u m s ta te s

The action of the Kronecker product can be seen clearly when we consider the vector representation
of two-qubit quantum states. If we are given two complex vectors
   
a c
and ,
b d

129 Q C a m p 2 0 2 3
C H A P T E R 8 . E N TA N G L E M E N T A N D N O N L O C A L IT Y

then the Kronecker product between these two vectors is

    
c ac
a · d  ad
   
a c
⊗ :=    =   .
b d  c   bc 
b·
d bd
Recall the computational basis (|0i , |1i) for a single qubit
   
1 0
|0i = , |1i = .
0 1
b b

The vector representation of the computation basis in equation (8.1) is therefore

       
1 0 0 0
0 1 0 0
|00i =
b0 , |01i =
 b 0 , |10i =
 b1 , |11i =
 b 0 ,
 (8.4)
0 0 0 1
With all of the bases introduced, we can now find the vector representation of two-qubit quantum
states |ψitot in equation (8.3)
 
    ac
a c ad
|ψitot = (a |0i + b |1i) ⊗ (c |0i + d |1i) = ⊗ = 
b d  bc 
b
bd

Some Examples

Let’s compute the vector representation of |00i state

    
1 1
1 · 0  0
   
1 1
|00i = ⊗ =   =   .
0 0 1  0
b 
0·
0 0

Exercise 8.1. Vector representations of bipartite quantum states. Verify the vector calculations
in equation (8.4). Then, show that
 
ac
ad
|ψi := ac |00i + ad |01i + bc |10i + bd |11i =
b bc  .


bd

Exercise 8.2. Vector representations of bipartite quantum states. Recall the quantum states
   
1 1 1 1
b √
|+i = , |−i = b √ .
2 1 2 −1
Carry out the vector calculations for |+i ⊗ |+i and |−i ⊗ |−i.

130 Q C a m p 2 0 2 3
C H A P T E R 8 . E N TA N G L E M E N T A N D N O N L O C A L IT Y

T h e B o rn r u le f o r t w o - q u b it q u a n t u m s ta te s

Similar to the single qubit quantum systems, the Born rule gives us the statistics of measurement
results of two-qubit states. Before applying the Born rule, we have to learn how to obtain the
inner product between two-qubit states

|φitot = |φiNTU |φiNUS and |ψitot = |ψiNTU |ψiNUS .

The inner product between these two states is

hφ|ψitot := (hφ|NTU hφ|NUS ) · (|ψiNTU |ψiNUS ) = hφ|ψiNTU · hφ|ψiNUS .

where hφ|NTU hφ|NUS is the Hermitian conjugate of |φiNTU |φiNUS

† †
[|φiNTU |φiNUS ]† := |φiNTU |φiNUS = hφ|NTU hφ|NUS

The rule of thumb is that the inner product of the Dirac representation of the quantum states
is acting in a slot by slot basis. In other words, each bra of the quantum states will only meet
with the respective ket of the quantum states, i.e: hφ|NTU (hφ|NUS ) only meet with |ψiNTU (|ψiNUS ).

Now, we will apply the Born rule to see how probable a measurement on the quantum state |ψitot
would produce the outcome |φitot . Let’s take the vector representation of these two quantum states
to be
   
ψ1 φ1
ψ2  φ2 
|ψitot =
bψ3  ,
 |φitot =
bφ3  .


ψ4 φ4

with |ψ1 |2 + |ψ2 |2 + |ψ3 |2 + |ψ4 |2 = 1 and |φ1 |2 + |φ2 |2 + |φ3 |2 + |φ4 |2 = 1. Then, the probability
of getting outcome |φitot given the quantum state |ψitot is
 2
ψ1
2   ψ2 
P (φ|ψ) := |hφ|ψi| =
b φ1 φ2 φ3 φ4 · 
ψ3  .


ψ4

Some Examples
Let’s do some examples here. Consider the calculations for P (00|Φ+ ) where
 
1
1 1 1 0
Φ+ := √ |00i + √ |11i =
b √  .
2 2 2 0

1

131 Q C a m p 2 0 2 3
C H A P T E R 8 . E N TA N G L E M E N T A N D N O N L O C A L IT Y

The probability of getting |00i from |Φ+ i is

2
1 1
P (00|Φ+ ) = h00| · ( √ |00i + √ |11i)
2 2
2
1 1
= √ h00|00i + √ h00|11i
2 2
2
1 1
= √ (h0|0i · h0|0i) + √ (h0|1i · h0|1i)
2 2
2
1 1
= √ (1) + √ (0)
2 2
1
=
2
Alternatively, we could find P (00|Φ+ ) in the vector representation
  2
1
2
 1 0 1 1
P (00|Φ+ ) =

1 0 0 0 ·√   = √ =
2 0 2 2
1

By applying the Born rule to |ψitot in equation 8.2, we get the probability of obtaining different
computation basis (|00i , |01i , |10i , |11i)
2 2
P (00|ψ) = |h00|ψi| = |α|2 , P (01|ψ) = |h01|ψi| = |β|2 ,
(8.5)
2 2
P (10|ψ) = |h10|ψi| = |γ|2 , P (11|ψ) = |h11|ψi| = |σ|2 .

The coefficients α, β, γ and σ of the quantum state in equation (8.2) have similar meaning as
the one in the single qubit state, in which they are the probability amplitudes of the respective
computation basis and the statistics of measurement outcome for the respective basis is given by
its norm-squared. The principle of normalization has to be respected, i.e:

|α|2 + |β|2 + |γ|2 + |σ|2 = 1.

in order for the |ψi to be a valid quantum state.

What we have been discussing so far is only the case with measurement performed on both qubits.
What if now, we are only interested on the measurement statistic of one qubit? Will the Kronecker
product structure gives us the correct single qubit measurement statistics? The short answer is
yes. Recall the quantum state |ψitot in equation (8.3)

|ψitot = (a |0i + b |1i) ⊗ (c |0i + d |1i) = ac |00i + ad |01i + bc |10i + bd |11i .

If we perform measurement on the first qubit, the possible output state is |00i (|10i) and |01i
(|11i) if we obtained |0i (|1i) from the first qubit. Therefore, the probability of getting |0i for the
first qubit is the sum of the probabilities of getting |00i and |01i

P (00|ψ) + P (01|ψ) = |ac|2 + |ad|2

132 Q C a m p 2 0 2 3
C H A P T E R 8 . E N TA N G L E M E N T A N D N O N L O C A L IT Y

= |a|2 · (|c|2 + |d|2 )

= |a|2

with |c|2 + |d|2 = 1 by definition while the probability of obtaining |1i from the first qubit is

P (10|ψ) + P (11|ψ) = |b|2 .

Indeed, these are the correct single qubit measurement statistics of the first qubit. Similarly, the
probability of getting |0i and |1i from the second qubit are

P (00|ψ) + P (10|ψ) = |c|2 and P (01|ψ) + P (11|ψ) = |d|2 ,

respectively. This agreed with our common sense as it make sense to have the measurement results
of different qubits to be independent to each other. However, not all quantum states behave in
the way. There is one class of quantum state with a strange property: the measurement outcome
of one qubit depends on the measurement outcome of the other qubit. Keep this in mind and we
will address this class of quantum state in section 8.1.5.

Exercise 8.3. Born rule. Verify that the measurement probability of quantum state |ψitot of
equation (8.2) in computation basis is the same as equation (8.5), in both Dirac representation
and vector representation.

Exercise 8.4. Born rule. What is the probability of getting |01i, |10i and |11i given the quantum
state |Φ+ i, i.e: P (01|Φ+ ), P (10|Φ+ ) and P (11|Φ+ )? You only need to find the answers using vector
representation.

8 .1.4 Q u a n tu m O p e r a t io n s f o r T w o - q u b it Q u a n t u m S y s te m s

In section 4.5, we have learned that quantum operations O can perform some transformations on
single qubit quantum systems and the operations are represented by a 2 × 2 matrix since the qubit
space is 2 dimensional, i.e: C2 . This gives us a hint of what should the dimensionality of the
matrix that represent the two-qubit quantum operations be: it needs to be a 4 × 4 matrix as the
two-qubit space is four dimensional, i.e: C2 ⊗ C2 .

Now, let’s imagine that both CQT NTU’s and CQT NUS’s lab can apply their respective quantum
operations ONTU and ONUS on their qubits |ψiNTU and |ψiNUS . What is the two-qubit quantum
operation that acts on the joint quantum state |ψitot := |ψiNTU ⊗|ψiNUS ? We need a mathematical
operation that combines ONTU and ONUS into a quantum operation that acts on the larger space.
Fortunately, we know which operation that does exactly what we want, i.e: the Kronecker product
⊗. We can combine ONTU and ONUS using the Kronecker product, constructing a two-qubit
quantum operation

Otot := ONTU ⊗ ONUS

133 Q C a m p 2 0 2 3
C H A P T E R 8 . E N TA N G L E M E N T A N D N O N L O C A L IT Y

that acts on the joint quantum state |ψitot .

M a t r ix r e p r e s e n t a t io n o f t w o - q u b it q u a n t u m o p e r a t io n s

Matrix representation of two-qubit quantum operations allows us to see directly the operative effect
of the Kronecker product. Suppose that the matrix representation of quantum operations ONTU
and ONUS are    
a b e f
ONTU = and ONUS =
c d g h
b b

Then, the Kronecker product of these two quantum operations is

Otot := ONTU ⊗ ONUS

   
a b e f
= ⊗
c d g h
b
      
e f e f ae af be bf
a · g h
b·
g h  ag ah bg bh 
(8.6)
 
=  = .
 e f e f   ce cf de df 
c· d·
g h g h cg ch dg dh

The Kronecker product does what we want it to do: combines two 2 × 2 matrices into one 4 × 4
matrix. Next, the hermitian conjugate O†tot of the operator Otot is defined as

a∗ e∗ a∗ g ∗ c∗ e ∗ c∗ g ∗
 
a∗ f ∗ a∗ h∗ c∗ f ∗ c∗ h∗ 
O†tot := O†NTU ⊗ O†NUS =
b b∗ e ∗
. (8.7)
b∗ g ∗ d∗ e∗ d∗ g ∗ 
b∗ f ∗ b∗ h∗ d∗ f ∗ d∗ h∗

Exercise 8.5. Matrix representation of bipartite quantum operations. Recall that

       
1 0 0 1 0 −i 1 0
I = , X = , Y = , Z =
0 1 1 0 i 0 0 −1
b b b b

Find any two out of 16 possible combinations of OA ⊗ OB where OA , OB = {I, X, Y, Z}.

h i
Exercise 8.6. Matrix representation of bipartite quantum operations. Verify I⊗Z |01i = − |01i,
h i h i
X ⊗ Z |01i = − |11i, and X ⊗ Y |01i = −i |10i in matrix representation.

Exercise 8.7. Hermitian conjugate of two-qubit quantum operations. Verify the hermitian con-
jugate of quantum operation in equation (8.6) as equation (8.7)

Exercise 8.8. Hermitian conjugate of two-qubit quantum operations. Find the hermitian conju-
gate of I ⊗ Z, X ⊗ Z, and X ⊗ Y.

134 Q C a m p 2 0 2 3
C H A P T E R 8 . E N TA N G L E M E N T A N D N O N L O C A L IT Y

U n it a r y o p e ra to rs

Similar to the single qubit system, the unitary operator will also preserve the norm of the two-
qubit quantum states. If we have a unitary operator UNTU with UNTU U†NTU = I and UNUS with
UNUS U†NUS = I, then the joint unitary operator of the two-qubit quantum systems would be

Utot := UNTU ⊗ UNUS .

The unitarity of the joint unitary operation follow immediately from the unitarity of its parts

Utot U†tot = (UNTU ⊗ UNUS )(U†NTU ⊗ U†NUS )

= UNTU U†NTU ⊗ UNUS U†NUS
= I ⊗ I = I4

where Id denote the d-dimension identity operator3 .

Exercise 8.9. Unitarity of two-qubit quantum operations. Verify that I ⊗ Z, X ⊗ Z and X ⊗ Y

are unitary operators, i.e:

(U†A ⊗ U†B ) · (UA ⊗ UB ) = (UA ⊗ UB ) · (U†A ⊗ U†B ) = I4 ,

where UA = {I, X} and UB = {Z, Y} in matrix representation.

O p e ra to rs o n t w o - q u b it q u a n t u m s ta te s

The action of two-qubit quantum operations on the joint quantum states is defined as

(ONTU ⊗ ONUS ) |ψiNTU ⊗ |ψiNUS := ONTU |ψiNTU ⊗ ONUS |ψiNUS . (8.8)

The rule of thumb is that the quantum operation is only acting on the respective quantum state,
i.e: ONTU (ONUS ) only acting on |ψiNTU (|ψiNUS ). This is important as it allows us to describe the
situation where the quantum operation is only acting on one of the qubits while leaving the other
untouched. For example, if we want to perform transformation on |ψiNUS but keeping |ψiNTU as
it is, then the joint quantum operator would be Otot = INTU ⊗ ONUS

Otot |ψitot = (INTU ⊗ ONUS ) · (|ψiNTU ⊗ |ψiNUS )

= INTU |ψiNTU ⊗ ONUS |ψiNUS
= |ψiNTU ⊗ ONUS |ψiNUS

3 The I with no subscript is always denotes as 2-dimensional identity operator unless otherwise stated.

135 Q C a m p 2 0 2 3
C H A P T E R 8 . E N TA N G L E M E N T A N D N O N L O C A L IT Y

Some Examples

Let’s consider some example: Apply I ⊗ Z and X ⊗ Z on |01i

h i
I ⊗ Z |01i = I |0i ⊗ Z |1i = |0i ⊗ (− |1i) = − |01i
h i
X ⊗ Z |01i = X |0i ⊗ Z |1i = |1i ⊗ (− |1i) = − |11i
h i
X ⊗ Y |01i = X |0i ⊗ Y |1i = |1i ⊗ (−i) |0i = −i |10i

Exercise 8.10. Find the operation that perform transformation on |ψiNTU but keeping |ψiNUS
untouched.

8 .1.5 Q u a n tu m E n t a n g le m e n t

In section 8.1.3, we have learned that arbitrary two-qubit quantum states can be constructed via
a linear superposition of the computation basis for two qubits

|00i , |01i , |10i , |11i ,

as long as the quantum states are normalized. Thus, we could define the following state
1 1
Φ+ := √ |00i + √ |11i . (8.9)
2 2
Now imagine that Alice and Bob, each holding one qubits of |Φ+ i and stay at the same place. If
Alice perform a measurement on her qubit in the computation basis (|0i, |1i) and Bob does noth-
ing, there will be a 50% chance of getting either |00i or |11i. If Bob perform measurement on his
qubit in (|0i, |1i) basis afterward, he will find his outcome always aligns with Alice’s outcome, i.e:
his outcome will be |0i (|1i) if Alice’s outcome is |0i (|1i). This observation is strange and perculiar
as Alice’s measurement has forces Bob’s qubit to choose the state that is aligned with Alice’s state.
This effect will still exist even if they are separated by a vast distances, i.e: Alice’s qubit is placed
on Earth while Bob’s qubit is placed on the Alpha Centauri that is 4.367 light years4 away from
Earth! Albert Einstein found this absurb and called it a “spooky action at a distance” as it seems
like one qubit “instantly knows” the outcome of the other qubit, appearing to be transmitting
information faster than the speed of light, hence, violating his theory of relativity. However, this
does not actually contradict physics since no information is actually being transmitted!

4 One light year is the distance travel by light in a year which is about 9 trillion km.

136 Q C a m p 2 0 2 3
C H A P T E R 8 . E N TA N G L E M E N T A N D N O N L O C A L IT Y

Entangled Quantum State:

If a quantum state |ψi cannot be written as |ψ1 i ⊗ |ψ2 i for any choice of |ψ1 i and |ψ2 i, then
the qubits are said to be entangled and |ψi is an entangled quantum state. Otherwise, it is
a separable quantum state.

Not all the quantum states has this perculiar property. Consider the following quantum state

1 1
|ψi = √ |00i + √ |01i .
2 2
If Alice measure her qubit (the first qubit) in computation basis, she will get the outcome |0i with
certain while Bob will have equal chance of getting |0i and |1i by measuring his qubit in the same
basis. In this case, the measurement on one qubit has no effect on the other qubit because the
joint quantum state |ψi can be separated into two single qubit states

1 1 1 1
|ψi = √ |00i + √ |01i = |0i ⊗ ( √ |0i + √ |1i).
2 2 2 2
This kind of quantum states are called separable quantum state.

Careful readers will notice that the quantum state |Φ+ i can also be expressed in the
complementary basis {|+i , |−i}

1 1
Φ+ = √ |++i + √ |−−i . (8.10)
2 2
which has a similar structure as equation 8.9. Given these two expressions, it seems like
it is possible for Alice to have a faster than speed of light communication with Bob by
measuring her qubits either in (|0i , |1i) basis or (|+i , |−i) basis.

Let’s imagine that Alice and Bob shared a copy of quantum state |Φ+ i and have a pre-
agreed information encoding scheme where Alice will measure her qubit in computation
(complementary) basis if she wants to communicate 0 (1) with Bob. The working principle
of the scheme is that if Alice measure in (|0i , |1i) basis, Bob will get |0i or |1i with
certainty given that he also measure in this basis. Similarly, if Alice measure in (|+i , |−i)
basis, Bob will get |+i or |−i with certainty given that he also measure in (|+i , |−i) basis.

This scheme will fail if there is no classical communication between Alice and Bob. This
is because Bob will not know when Alice will performs her measurement as well as not
knowing Alice’s choice of measurement basis. Bob can only make an informed decision of
which basis to measure if and only if Alice communicate her measurement outcome with
Bob through classical communication channel. As the information transmission speed of
the classical communication channel is limited by the speed of light, therefore, faster than

137 Q C a m p 2 0 2 3
C H A P T E R 8 . E N TA N G L E M E N T A N D N O N L O C A L IT Y

speed of light communication is not possible using this schemea .

a In fact, all other schemes can be shown to be not violating the theory of relativity in the same way.

On the other hand, quantum state |Φ+ i cannot be written as the Kronecker product of two single
qubit states |ψ1 i and |ψ2 i, i.e: |Φ+ i 6= |ψ1 i ⊗ |ψ2 i, with vector representation
   
a c
|ψ1 i = and |ψ2 i =
b d
b b

with |a|2 + |b|2 = 1 and |c|2 + |d|2 = 1. This can be proof by contradiction. Let’s assume that |Φ+ i
can be written as the Kronecker product of |ψ1 i and |ψ2 i, ie.: |Φ+ i = |ψ1 i ⊗ |ψ2 i
   
1 ac
1  0  ad
√   =  .
2 0  bc 
1 bd

By comparing the second entry of both vectors, we can see that ad = 0 and this implies that a = 0
or d = 0. Hence, we must have ac = 0 or bd = 0, but this is not true as ac = bd = √1 .
2
This
leads to a contraction, therefore, we can conclude that the qubits in quantum state |Φ i cannot +

be described independently of the state of the others, i.e: |Φ+ i =

6 |ψ1 i ⊗ |ψ2 i. The measurement
statistics of both qubits will be meaningful only if they are compared against each other, i.e:
their measurement outcomes in the same basis are always aligned. Otherwise, their measurement
statistics will appear to be random where all outcomes appear with equal probability. Thus, all
qubits in the system should be treated as one entity as the whole quantum state is more than the
sum of its parts. This kind of quantum states are called entangled quantum states and they contain
the exotic quantum machanical resource: quantum entanglement.

Quantum entanglement is one of the important quantum mechanical properties that has huge im-
plication on quantum information and technologies. It appears everywhere, ranging from quantum
cryptography to quantum computation and it might be one of the resources that could poten-
tially help us to achieve quantum advantages in quantum computation5 . The inseparability of
the two-qubit state into a Kronecker product of two single qubit states induces many interesting
quantum phenomenons that has no classical counterpart. In the next section, we shall see one of
the quantum phenomena powered by quantum entanglement: quantum teleportation.

Exercise 8.11. Verify that

1 1 1 1
Φ+ := √ |00i + √ |11i = √ |++i + √ |−−i .
2 2 2 2

5 Quantum advantage is the ultimate goal of quantum computation. The aim is to demonstrate a programmable
quantum computer that can solve a problem that no classical computer can solve in any feasible amount of time.

138 Q C a m p 2 0 2 3
C H A P T E R 8 . E N TA N G L E M E N T A N D N O N L O C A L IT Y

Exercise 8.12. Separable quantum states. Prove the following two-qubit states are NOT entan-

gled:        
1 1 1 0
0 1 1 1 1 1 0
|ψ1 i =
b0 ,
 b √ 
|ψ2 i = , |ψ3 i =
b  , b √ 
|ψ4 i = 
2 0
 2 1 2 1 
0 0 1 −1

Exercise 8.13. Entangled quantum states. Find the vector representations of

1 1 1 1
Φ+ := √ |00i + √ |11i Φ− := √ |00i − √ |11i
2 2 2 2
1 1 1 1
Ψ+ := √ |01i + √ |10i Ψ− := √ |01i − √ |10i .
2 2 2 2
and show that they are entangled states. These 4 states are called as Bell’s states.

139 Q C a m p 2 0 2 3
C H A P T E R 8 . E N TA N G L E M E N T A N D N O N L O C A L IT Y

8 .1.6 Q u a n tu m T e le p o r t a t io n

In the classical world, we can copy and transmit information to other people. For example, you
could just press “control + c” follows by “control + v” in your keyboard, to copy the text from
one place to another and click the “send” button to send the text. Or, you could photocopy a
document and pass it to your friends without any issue. However, this is not the case in the
quantum world. Quantum mechanics imposed tight restrictions on the way we copy and transmit
quantum information. In this section, we shall see how the restriction arises from the unitarity of
quantum evolution and the way to bypass it.

U n k n o w n q u a n tu m s ta te s c a n n o t b e c lo n e d

Figure 8.1: A classical cloning of the classical bits can be done using XEROX. In contrast, there
exist no quantum XEROX to clone an unknown quantum state.

In this digital era, it is a norm to copy and transmit information between places. We has taken for
granted that a XEROX6 machine can scan, copy, and print documents without any restriction. Is
there a quantum version of XEROX machine that could copy quantum information like its classical
counterpart? Unfortunately, quantum information do not behave in the same way as the classical
information. The theory of quantum mechanics has imposed tight restriction on which types of
quantum information we can copy.

6 It is just the old way of calling photocopier.

140 Q C a m p 2 0 2 3
C H A P T E R 8 . E N TA N G L E M E N T A N D N O N L O C A L IT Y

No-Cloning Theorem:

It is impossible to create an independent and identical copy of an arbitrary unknown quan-

tum state.

Suppose we are given a magical quantum device, i.e: the quantum XEROX machine to copy
unknown quantum states |ψi by replicating it on a “blank” quantum state |0i7 . This is a quantum
operation, hence, has a corresponding matrix representation U. By assumption, this matrix U
satisfies
U(|ψi ⊗ |0i) = |ψi ⊗ |ψi

for any quantum state |ψi. Now, we will show that this quantum device has to respect the
constraint imposed by quantum mechanics, i.e: it cannot copy arbitrary quantum states. To see
this, we input two different quantum states |ψi and |φi into this device to get

U(|ψi ⊗ |0i) = |ψi ⊗ |ψi (8.11a)

U(|φi ⊗ |0i) = |φi ⊗ |φi . (8.11b)

Let’s multiply equation 8.11b by the hermitian conjugate of equation 8.11a

[hψ| ⊗ h0|]U† U[|φi ⊗ |0i] = [hψ| ⊗ h0|]U† |φi ⊗ |φi . (8.12)

Since any unitary matrices preserve inner products, the left hand side (LHS) of equation 8.12 will
becomes

[hψ| ⊗ h0|]U† U[|φi ⊗ |0i] = (hψ| ⊗ h0|) U† U (|φi ⊗ |0i)

= (hψ| ⊗ h0|)(|φi ⊗ |0i)

= hψ|φi · h0|0i
= hψ|φi

while for the right hand side (RHS) of equation 8.12, we have that

[hψ| ⊗ h0|]U† |φi ⊗ |φi = (hψ| ⊗ hψ|)(|φi ⊗ |φi)

= hψ|φi · hψ|φi
2
= hψ|φi .

Hence, equation 8.12 can be re-expressed as

2
hψ|φi = hψ|φi .

7 Just like you use photocopier to print documents on blank papers.

141 Q C a m p 2 0 2 3
C H A P T E R 8 . E N TA N G L E M E N T A N D N O N L O C A L IT Y

This equation will holds only when hφ|ψi = 0 or 1, i.e: |φi has to be a quantum state that is
orthogonal to |ψi or identical to |ψi8 , which is not true for every choice of two states! In other
words, quantum mechanics allows us to create a quantum device that copy |ψi and those quantum
states that are orthogonal to it, but not to clone a general quantum state. This is pointless as it is
not worth the effort to just create a machine that can copy a limited number of quantum states.
For a qubit system, there are infinite number of orthogonal bases but each of them only consists
of two quantum states.

A direct consequence of non-cloning theorem is that we cannot copy arbitrary quantum states
and send them to other party, which is inconvenience if we want to share quantum information9 .
You may wonder if there is a way to bypass the constrain imposed by the no-cloning theorem.
Fortunately, this can be done if we have an entangled quantum state, the exotic quantum state
that we have introduced in Section 8.1.5.

T r a n s m it t in g q u a n tu m in f o r m a t io n b e tw e e n q u a n tu m s y s te m s

A glimpse on quantum teleportation We will first introduce the protocol without any deriva-
tions in order for you to get an intuition on how quantum teleportation works and bring back
mathematics afterward to get a complete picture of the protocol. I should warn those readers who
wish to skip the next section. Without looking at the derivations, we will not know what is hap-
pening behind the scene and quantum teleportation would appears to be a black magic. Although
the mathematics seems tedious and daunting, it will be rewarding as you will see that quantum
teleportation is just the consequence of the principle of superposition.

Suppose Alice wants to send the qubit with quantum state |ψiA0 = a |0i+b |1i to Bob while sharing
a Bell state |Φ+ iAB with Bob. The joint quantum state of these three qubits system is

|ψitot = |ψiA0 ⊗ Φ+ AB
. (8.13)

The subscripts A, A0 and B denote who keeps the qubit: A, A0 for Alice and B for Bob. It can
be shown that quantum state |ψitot could be rewritten as

1h +    
|ψitot = Φ A0 A
a |0iB + b |1iB + Φ− A0 A
a |0iB − b |1iB
2    i
+ Ψ+ A0 A
b |0iB + a |1iB + Ψ+ A0 A
b |0iB − a |1iB

8 To be precise, |φi has to be identical to |ψi up to a global phase but it is not important in our discussion here.
9 Imagine a world that does not allows you to copy and paste any information in your computer or photocopy a
document using a photocopier at you wish.
10 Source: Gisin, N. (2017). Quantum-teleportation experiments turn 20.

142 Q C a m p 2 0 2 3
C H A P T E R 8 . E N TA N G L E M E N T A N D N O N L O C A L IT Y

Figure 8.2: Quantum teleportation protocol10 . Alice (sender) and Bob (receiver) shared a pair
of entangled photons. To send an unknown quantum state (red photon) to Bob, Alice has to
perform a Bell’s measurement (Bell-state analyser) on her qubits (red and grey photons) and send
the measurement outcome (classical signal) to Bob. Bob will apply a unitary transformation on
his qubit (blue photon) conditioning on Alice’s outcome. The transformed quantum state will be
identical to the unknown quantum state.

where |Φ+ i, |Φ− i, |Ψ+ i and |Ψ− i are called Bell’s states and they are defined as
1 1 1 1
Φ+ := √ |00i + √ |11i Φ− := √ |00i − √ |11i
2 2 2 2
1 1 1 1
Ψ+ := √ |01i + √ |10i Ψ− := √ |01i − √ |10i . (8.14)
2 2 2 2
All of these Bell’s states are entangled quantum state (see exercise.8.13) and they form a basis
set that spans the four dimensional qubit space. Alice will then measures qubits on her side, i.e:
the first two qubits labelled by A and A0 in the Bell’s basis (|Ψ+ i, |Ψ− i, |Φ+ i, |Φ− i)11 and four
different outcomes will occur with equal probability, i.e: 4.
1
If Alice’s qubits collapse to

• |Φ+ iA0 A , then Bob’s qubit is in state |φiB = a |0iB + b |1iB

• |Φ− iA0 A , then Bob’s qubit is in state |φiB,1 = a |0iB − b |1iB

• |Ψ+ iA0 A , then Bob’s qubit is in state |φiB,2 = b |0iB + a |1iB

• |Ψ− iA0 A , then Bob’s qubit is in state |φiB,3 = b |0iB − a |1iB .

You could see that after Alice’s measurement, Bob’s qubit is identical to |ψiA0 in the first scenario
while for the other three scenarios, Bob can performs some transformations to transform |φiB,n

11 see question.8.17 to learn how to performs Bell’s measurement.

143 Q C a m p 2 0 2 3
C H A P T E R 8 . E N TA N G L E M E N T A N D N O N L O C A L IT Y

to |φiB if he knows Alice’s measurement outcome. However, at this moment, Bob do not know
whether Alice has measured her qubits or not, so his qubit is still appear to be random from his
perspective. Thus, Alice has to informs Bob her measurement outcome, i.e: which of the four
Bell’s state her qubit has collapsed to so that Bob can make an informed decision. Note that it
will only takes 2 (classical) bits to send this information to Bob. If Alice says

• |Φ+ iA0 A , then Bob knows his qubit is in state |ψiB = a |0iB + b |1iB . He does nothing on
his qubit since his qubit is already in the correct state

I(a |0iB + b |1iB ) = a |0iB + b |1iB .

• |Φ− iA0 A , then Bob knows his qubit is in state |φiB,1 = a |0iB −b |1iB . He applies Z operation
on his qubit
Z(a |0iB − b |1iB ) = a |0iB + b |1iB .

• |Ψ+ iA0 A , then Bob knows his qubit is in state |φiB,2 = b |0iB +a |1iB . He applies X operation
on his qubit
X(b |0iB + a |1iB ) = a |0iB + b |1iB .

• |Ψ− iA0 A , then Bob knows his qubit is in state |φiB,3 = b |0iB −a |1iB . He applies Z operation
follows by X operation on his qubit

XZ(b |0iB − a |1iB ) = X(b |0iB + a |1iB ) = a |0iB + b |1iB .

Notice now that Bob’s quantum state is the same as the quantum state |ψiA0 for all four scenarios.
Voila, Alice has sent an unknown quantum state to Bob! Interestingly, Alice and Bob still have
no knowledge on the values a and b after the protocol! Alice only performs Bell’s measurement
on her qubits and send the respective measurement outcome to Bob while Bob performs a unitary
operation conditioning on Alice’s measurement outcome. All of these operations do not reveal any
information about a and b.

Did Alice and Bob violate the no-cloning theorem? No, they did not because the original copy of
|ψiA0 is completely destroyed right after Bell’s measurement. After this step, the joint quantum
state of Alice’s qubits becomes one of the Bell’s states and the measurement statistics of qubit
A0 is now random12 , hence, it contains no information about its original state |ψiA0 . In other
words, there is only one copy of |ψi present in the protocol and the quantum state is “transport”
between two places, hence the name quantum teleportation. Actually, the naming of quantum
teleportation might be misleading as no quantum system is physically transported from one place
to the other in the protocol. It is the quantum information stored in the form of quantum state

12 Given that Alice does not compare it against the measurement outcome of the other qubit.

144 Q C a m p 2 0 2 3
C H A P T E R 8 . E N TA N G L E M E N T A N D N O N L O C A L IT Y

being transmitted from one quantum system to another quantum system. Quantum teleportation
can be summarized neatly as

2 Classical Bits + 1 Shared Entanglement = 1 Qubit

where we can use two classical bits (Bell state’s measurement outcome Alice sent to Bob) and one
share entanglement (|Φ+ i shared by Alice and Bob) to “transport” one qubit (|ψi).

Exercise 8.14. Verify that Bell’s states in equation 8.14 are orthogonal to each other. This
means that they could forms a set of basis that spans the four dimensional qubit space. You may
want to attempt exercise.8.13 first before trying this exercise.

Missing pieces of quantum teleportation The “black magic” of quantum teleportation works
if and only if the following relation holds
1h +    
|ψiA0 ⊗ Φ+ AB
= Φ A0 A
a |0iB + b |1iB + Φ− A0 A
a |0iB − b |1iB
2    i
+ Ψ+ A0 A
b |0iB + a |1iB + Ψ+ A0 A
b |0iB − a |1iB . (8.15)

We shall see that this is actually the principle of superposition at work. To simplify the notations,
we will drop the index A, A0 and B but it should be clear that the first two qubits belong to Alice
while Bob holds the third qubit. First, we expand the LHS of equation 8.15 and group the first
two qubits into one entity
1
|ψi ⊗ Φ+ = (a |0i + b |1i) ⊗ √ (|00i + |11i)
2
1
= √ (a |00i ⊗ |0i + a |01i ⊗ |1i + b |10i ⊗ |0i + b |11i ⊗ |1i). (8.16)
2
It can be shown that the computational basis can be written as the superposition of the Bell’s
basis13
1 1 1 1
|00i = √ Φ+ + √ Φ− , |11i = √ Φ+ − √ Φ− (8.17a)
2 2 2 2
1 1 1 1
|01i = √ Ψ+ + √ Ψ− , |10i = √ Ψ+ − √ Ψ− (8.17b)
2 2 2 2
Substitute them into equation 8.16 yield
1 h  1 1   1 1 
|ψi ⊗ Φ+ = √ a √ Φ+ + √ Φ− ⊗ |0i + a √ Ψ+ + √ Ψ− ⊗ |1i
2 2 2 2 2
 1 1   1 1 
+ b √ Ψ+ − √ Ψ− ⊗ |0i + b √ Φ+ − √ Φ−

⊗ |1i) .
2 2 2 2

13 This should not be new to you as you have learned in section 4.4.1 that (|0i , |1i) and (|+i , |−i) basis can be
expressed as the superposition of each other.

145 Q C a m p 2 0 2 3
C H A P T E R 8 . E N TA N G L E M E N T A N D N O N L O C A L IT Y

We bring the factor √1

2
out of the square bracket and expand each of the terms, giving us

1 h   
|ψi ⊗ Φ+ = a Φ+ |0i + a Φ− |0i + a Ψ+ |1i + a Ψ− |1i
2    i
+ b Ψ+ |0i − a Ψ− |0i + b Φ+ |1i − b Φ− |1i . (8.18)

By grouping the terms for each of the Bell’s basis, we finally reach the following expression

1h +    
|ψi ⊗ Φ+ = Φ a |0i + b |1i + Φ− a |0i − b |1i
2    i
+ Ψ+ b |0i + a |1i + Ψ+ b |0i − a |1i (8.19)

which is the RHS of equation 8.15.

You might wonder in which steps have we apply the principle of superposition. The answer is in
each and every steps, from the start to the end. This is not surprising as the theory of quantum
mechanics is build upon the principle of superposition14 . However, I would want to comment
on two important and crucial aspects of the principle that make quantum teleportation possible.
First, the principle of superposition treats all bases in equal footing15 , allowing us to re-express
the quantum state of Alice’s qubits in term of Bell’s basis. The “black magic” will not be possible
if nature prefer the computation basis over the Bell’s basis. Next, the principle of superposition
ensure that the resultant quantum state is the same irrespective of the order of the summands16 .
This allows us to group the terms in equation 8.18 according to the Bell’s state, enabling Alice to
get 4 distinct outcomes with associated Bob’s quantum states. With the above derivations, I hope
you will appreciate the principle of superpositon and realise its importance in quantum mechanics.
It is one of the essential elements that make quantum mechanics differ from the classical theory
and it induces many interesting quantum phenomenons that could not be realized in a world that
is purely classical.

Exercise 8.15. Computational and Bell’s basis. Verify equation 8.17a and equation 8.17a: Show
that the computational basis can be written as the superposition of the Bell’s basis.

14 Of cause there are other elements contain in the theory of quantum mechanics but they are not our focus.
15 This is important in every aspect of quantum mechanics and first demonstrated in the Stern Gerlach experiment.
16 It respects the order of Kronecker product.

146 Q C a m p 2 0 2 3
C H A P T E R 8 . E N TA N G L E M E N T A N D N O N L O C A L IT Y

8 .1.7 E x e r c is e s

Exercise 8.16. (Advanced) Transformation between Bell’s states. Find the quantum operations
that transform |Φ+ i into |Φ− i, |Ψ+ i and |Ψ− i. Hint: You only need to find the quantum operations
OA and keep the quantum operation OB as an identity operation, i.e: OB = I.

Exercise 8.17. (Advanced). Preparation of Bell’s states. One can prepare Bell’s states |Φ+ i , |Φ− i , |Ψ+ i
and |Ψ− i by applying a Hadamard gate H⊗I and followed by a CNOT gate on |00i , |01i , |10i , |11i.
The CNOT is a conditional quantum operation where it will flips the second qubit if and only if
the first qubit is |1i. The CNOT gate and the Hadamard gate are defined as
   
1 0 0 0 1 0 1 0
0 1 0 0 1 0 1 0 1
CNOT =
b , H⊗I=
b √  .
0 0 0 1 2 1
 0 −1 0
0 0 1 0 0 1 0 −1

Verify
h i h i
CNOT H ⊗ I |00i = Φ+ , CNOT H ⊗ I |10i = Φ−
h i h i
CNOT H ⊗ I |01i = Ψ+ , CNOT H ⊗ I |11i = Ψ− .

One can apply the quantum operations in reverse order on the Bell’s states to recover the com-
putational basis. This could be used to perform Bell’s measurements, i.e: measurements that
distinguish Bell’s states.

Exercise 8.18. (Advanced). Superdense coding. You have seen in section 8.1.6 how an entan-
gled quantum state can be used to send a quantum state from one place to another. Let’s consider
another scenario in this exercise. Suppose that Alice and Bob are sharing a Bell state

1 1
Φ+ AB
= √ |00iAB + √ |11iAB .
2 2
This time, Alice is not allowed to send any classical information directly to Bob but she is allowed
to send her qubit to Bob through a noiseless quantum channel while Bob has a Bell-state analyser
that can distinguish the Bell’s states. Propose a protocol to help Alice send 2 bits of classical
information to Bob in this restricted setting. hint: Find a way to encode 2 bits of classical
information in Bell’s states.

147 Q C a m p 2 0 2 3
C H A P T E R 8 . E N TA N G L E M E N T A N D N O N L O C A L IT Y

8 .1.8 E x e r c is e S o lu t io n s

Solution 8.1.
    
1 1
1 · 0  0
   
1 1
1. |00i = ⊗ =   =  
0 0 1  0
b 
0·
0 0
    
0 0
    1·
1 0 1 = 1
2.
   
|01i = ⊗ =
0 1 0 0
b   
0·
1 0
    
1 0
0 · 0  0
   
0 1
3. |10i = ⊗ =   =  
1 0 1  1
b 
1·
0 0
    
0 0
0 · 1  0
   
0 0
4. |11i = ⊗ =   =  
1 1 0  0
b 
1·
1 1
Substitute the vector representation of each of the basis into the quantum state |ψi yield
         
1 0 0 0 ac
0 1 0 0 ad
|ψi =
b ac 
0 + ad 0 + bc 1 + bd 0 =  bc 
        

0 0 0 1 bd

Solution 8.2.

    
1 1
    1 ·
1 1   = 12 1
1
1. |+i ⊗ |+i =
   
b √12 ⊗ √12 = 12 
1 1  1  1
1·
1 1
    
1 1
    1·
1 1 −1 1 −1
2. |−i ⊗ |−i =
   
b √12 ⊗ √12 = 12    = 2 −1
−1 −1  1 
−1 ·
−1 1

148 Q C a m p 2 0 2 3
C H A P T E R 8 . E N TA N G L E M E N T A N D N O N L O C A L IT Y

Solution 8.3.

1.   2
1
2  0
= |α|2 .

|hψ|00i| = α β γ σ · 
0
0

2 2
|hψ|00i| = |(α∗ h00| + β ∗ h01| + γ ∗ h10| + σ ∗ h11|) · |00i|
2
= |(α∗ h00|00i + β ∗ h01|00i + γ ∗ h10|00i + σ ∗ h11|00i|
2
= |(α∗ h0|0i · h0|0i + β ∗ h0|0i · h1|0i + γ ∗ h1|0i · h0|0i + σ ∗ h1|0i · h1|0i|
2
= |(α∗ (1) + β ∗ (0) + γ ∗ (0) + σ ∗ (0)| = |α|2

2.   2
0
2  1
= |β|2 .

|hψ|01i| = α β γ σ ·
0


2 2
|hψ|01i| = |(α∗ h00| + β ∗ h01| + γ ∗ h10| + σ ∗ h11|) · |01i|
2
= |(α∗ h00|01i + β ∗ h01|01i + γ ∗ h10|01i + σ ∗ h11|01i| = |β|2

3.   2
0
2  0
= |γ|2 .

|hψ|10i| = α β γ σ ·
1


2 2
|hψ|10i| = |(α∗ h00| + β ∗ h01| + γ ∗ h10| + σ ∗ h11|) · |10i|
2
= |(α∗ h00|10i + β ∗ h01|10i + γ ∗ h10|10i + σ ∗ h11|10i| = |γ|2

4.   2
0
2  0
= |σ|2

|hψ|11i| = α β γ σ · 
0
1

2 2
|hψ|11i| = |(α∗ h00| + β ∗ h01| + γ ∗ h10| + σ ∗ h11|) · |11i|
2
= |(α∗ h00|11i + β ∗ h01|11i + γ ∗ h10|11i + σ ∗ h11|11i| = |σ|2

149 Q C a m p 2 0 2 3
C H A P T E R 8 . E N TA N G L E M E N T A N D N O N L O C A L IT Y

Solution 8.4.

1.   2
0
1   1
P (Φ+ |01) = √ 1 0 0 1 · 
0 =0
2
0

2.   2
0
1   0
P (Φ+ |10) = √ 1 0 0 1 · 
1 =0
2
0

3.   2
0
2
1   0 1 1
P (Φ+ |11) = √ 1 0 0 1 · 
0 = √ =
2 2 2
1

Solution 8.5.
     
1 0 0 0 0 1 0 0 0 −i 0 0
0 1 0 0 1 0 0 0 i 0 0 0
I⊗I=
b  , I⊗X=
b  , I⊗Y= b  
0 0 1 0 0 0 0 1 0 0 0 −i
0 0 0 1 0 0 1 0 0 0 i 0
     
1 0 0 0 0 0 1 0 0 0 0 1
0 −1 0 0  0 0 0 1 0 0 1 0
I⊗Z=
b 
0
, X ⊗ I =
b  , X ⊗ X = b  
0 1 0  1 0 0 0 0 1 0 0
0 0 0 −1 0 1 0 0 1 0 0 0
     
0 0 0 −i 0 0 1 0 0 0 −i 0
0 0 i 0 0 0 0 −1 0 0 0 −i
X⊗Y= , X ⊗ Z = , Y⊗I=

b  b  b  
0 −i 0 0 1 0 0 0  i 0 0 0
i 0 0 0 0 −1 0 0 0 i 0 0
     
0 0 0 −i 0 0 0 −1 0 0 −i 0
0 0 −i 0  0 0 1 0  0 0 0 i
Y⊗X= , Y⊗Y= , Y ⊗ Z =

b 
0 b  b  
i 0 0  0 1 0 0  i 0 0 0
i 0 0 0 −1 0 0 0 0 −i 0 0
     
1 0 0 0 0 1 0 0 0 −i 0 0
0 1 0 0  1 0 0 0  i 0 0 0
Z⊗I= , Z⊗X= , Z ⊗ Y =

b    
0 0 −1 0  0 0 0 −1 0 0 0 i
b  b 
0 0 0 −1 0 0 −1 0 0 0 −i 0
 
1 0 0 0
0 −1 0 0
Z⊗Z=  
0 0 −1 0
b 
0 0 0 1

150 Q C a m p 2 0 2 3
C H A P T E R 8 . E N TA N G L E M E N T A N D N O N L O C A L IT Y

Solution 8.6.
    
1 0 0 0 0 0
h i 0 −1 0 0 1 1
1. I ⊗ Z |01i =
b0
  = −  =
0 b − |01i
0 1 0  0
0 0 0 −1 0 0
    
0 0 1 0 0 0
h i 0 0 0 −1 1 0
2. X ⊗ Z |01i =b1
  = −  =
0 b − |11i
0 0 0  0
0 −1 0 0 0 1
    
0 0 0 −i 0 0
h i 0 0 i 0 1 0
3. X ⊗ Y |01i =b    = −i   =
1 b −i |10i
0 −i 0 0  0
i 0 0 0 0 0

Solution 8.7.

O†tot := O†NTU ⊗ O†NUS

 ∗
c∗
  ∗
g∗

a e
=
b ∗ ⊗ ∗
b d∗ f h∗
 ∗
g∗ e∗ g∗
  ∗ ∗
a∗ g ∗ c∗ e∗ c∗ g ∗
   
∗ e a e
a · ∗ c∗ ·

 f∗ h∗ ∗
f ∗ h∗  a∗ f ∗ a∗ h∗ c∗ f ∗ c∗ h∗ 
= = .
 e g∗ e g ∗   b∗ e∗ b∗ g ∗ d∗ e∗ d∗ g ∗ 
b∗ · ∗ d∗ · ∗
f h∗ f h∗ b∗ f ∗ b∗ h∗ d∗ f ∗ d∗ h∗

Solution 8.8.

1. (I ⊗ Z)† = I† ⊗ Z† = I ⊗ Z
 †  
1 0 0 0 1 0 0 0
0 −1 0 0 0 −1 0 0
(I ⊗ Z) =
†
b
0
 = 
0 1 0 0 0 1 0
0 0 0 −1 0 0 0 −1

2. (X ⊗ Z)† = X† ⊗ Z† = X ⊗ Z
 †  
0 0 1 0 0 0 1 0
0 0 0 −1 0 0 0 −1
(X ⊗ Z)† =

b  = 
1 0 0 0 1 0 0 0
0 −1 0 0 0 −1 0 0

3. (X ⊗ Y)† = X† ⊗ Y† = X ⊗ Y
 †  
0 0 0 −i 0 0 0 −i
0 0 i 0 0 0 i 0
(X ⊗ Y)† =
  
b = 
0 −i 0 0 0 −i 0 0
i 0 0 0 i 0 0 0

151 Q C a m p 2 0 2 3
C H A P T E R 8 . E N TA N G L E M E N T A N D N O N L O C A L IT Y

Solution 8.9.

1.
   
1 0 0 0 1 0 0 0
0 −1 0 0 0 −1 0 0
(I ⊗ Z) · (I ⊗ Z) =
†
b
0
· 
0 1 0  0 0 1 0
0 0 0 −1 0 0 0 −1
 
1 0 0 0
0 1 0 0
=
0 b I4
=
0 1 0
0 0 0 1

2.
   
0 0 1 0 0 0 1 0
0 0 0 −1 0 0 0 −1
(X ⊗ Z) · (X ⊗ Z) =
†
b
1
· 
0 0 0  1 0 0 0
0 −1 0 0 0 −1 0 0
 
1 0 0 0
0 1 0 0
=
0
=b I4
0 1 0
0 0 0 1

3.
   
0 0 0 −i 0 0 0 −i
0 0 i 0 0 0 i 0
(X ⊗ Y) · (X ⊗ Y)† =

b · 
0 −i 0 0  0 −i 0 0
i 0 0 0 i 0 0 0
 
1 0 0 0
0 1 0 0
=0
= b I4
0 1 0
0 0 0 1

Solution 8.10. The joint quantum operation: Otot = ONTU ⊗ INUS

Otot |ψitot = (ONTU ⊗ INUS ) · (|ψiNTU ⊗ |ψiNUS )

= ONTU |ψiNTU ⊗ INUS |ψiNUS
= ONTU |ψiNTU ⊗ |ψiNUS

Solution 8.11. First, we identify that the computation basis (|0i , |1i) can be written in a equal

superposition of the complementary basis (|+i , |−i)

1 1
|0i = √ (|+i + |−i), |1i = √ (|+i − |−i)
2 2
Substitute it into |Φ+ i
1 1
Φ+ := √ |00i + √ |11i
2 2

152 Q C a m p 2 0 2 3
C H A P T E R 8 . E N TA N G L E M E N T A N D N O N L O C A L IT Y

1 h1 1 i
=√ (|+i + |−i)(|+i + |−i) + (|+i − |−i)(|+i − |−i)
2 2 2
1 h i
= √ (|++i + |+−i + |−+i + |−−i) + (|++i − |+−i − |−+i + |−−i)
2 2
1 1
= √ |++i + √ |−−i .
2 2
Solution 8.12.

 
1    
0
1. |ψ1 i =
b = 1 ⊗ 1 = b |0i ⊗ |0i
0 0 0
0
 
1    
1 1 1 1
2.
  1
|ψ2 i =
b 2 =
√ ⊗ 2√ b |0i ⊗ |+i
=
0 0 1
0
 
1    
1 1 1 1
3.
 
b 2   = √12
|ψ3 i = ⊗ √12 b |+i ⊗ |+i
=
1 1 1
1
 
0    
0 0 1
4. 1
 1
|ψ4 i =
b √2   =
  ⊗ √2 =b |1i ⊗ |−i
1 1 −1
−1

Solution 8.13.
     
1 0 1
0 0 0
1. |Φ i := √2 |00i + √2 |11i =
+ 1 1 1 1 1
0 + 2 0 = 2 0
b √2   √   √  

0 1 1
ad = 0 implies that a = 0 or d = 0, hence, we need ac = 0 or bd = 0. However, we have
ac = bd = √1
2
for |Φ+ i. So |Φ+ i cannot be written as a product state, i.e: |Φ+ i 6= |ψ1 i ⊗ |ψ2 i
for any |ψ1 i and |ψ2 i, therefore, it is an entangled state.

     
1 0 1
0 0 0
2. |Φ i := √2 |00i − √2 |11i =
− 1 1 1 1 1
0 − 2 0 = 2  0 
b √2   √   √  

0 1 −1
ad = 0 implies that a = 0 or d = 0, hence, we need ac = 0 or bd = 0. However, we
have ac = √1
2
and bd = − √12 for |Φ− i. So |Φ− i cannot be written as a product state, i.e:
|Φ− i 6= |ψ1 i ⊗ |ψ2 i for any |ψ1 i and |ψ2 i, therefore, it is an entangled state.

     
0 0 0
1 0 1
3. |Ψ+ i := √12 |01i + √12 |10i =
  
b √12 
0 +
 √1   =
2 1
√1  
2 1

0 0 0
ac = 0 implies that a = 0 or c = 0, hence, we need ad = 0 or bc = 0. However, we have

153 Q C a m p 2 0 2 3
C H A P T E R 8 . E N TA N G L E M E N T A N D N O N L O C A L IT Y

ad = bc = √1
2
for |Ψ+ i. So |Ψ+ i cannot be written as a product state, i.e: |Ψ+ i 6= |ψ1 i ⊗ |ψ2 i
for any |ψ1 i and |ψ2 i, therefore, it is an entangled state.

     
0 0 0
1 0 1
4. |Ψ i := √2 |01i − √2 |10i =
− 1 1 1 1 1
0 − 2 1 = 2 −1
b √2   √   √  

0 0 0
ac = 0 implies that a = 0 or c = 0, hence, we need ad = 0 or bc = 0. However, we
have ad = √1
2
and bc = − √12 for |Ψ− i. So |Ψ− i cannot be written as a product state, i.e:
|Ψ− i 6= |ψ1 i ⊗ |ψ2 i for any |ψ1 i and |ψ2 i, therefore, it is an entangled state.

Solution 8.14.
       
1 1 0 0
1 0 1 0 1 1 1 1
Φ+ Φ− = Ψ+ = Ψ−
  
b √ 
= , b √  , b √  , b √ 
= 
2 0
 2  0  2 1
 2 −1
1 −1 0 0
 
1
 1 0
1. hΦ+ |Φ− i = √1 1

2
1 0 0 1 · √2 
 0  = 2 [1 − 1] = 0.


−1
 
0
 1 1
2. hΦ |Ψ i = √2 1 0 0 1 · √2 
1
+ +

1 = 0.


0
 
0
1
3. hΦ+ |Ψ− i = √12 1 0 0 1 · √12 
  
−1 = 0.


0
 
0
1
4. hΦ− |Ψ+ i = √12 1 0 0 −1 · √12 
  
1 = 0.


0
 
0
1
5. hΦ− |Ψ− i = √2 1 0 0 −1 · √2 
1
  1

−1 = 0.


0
 
0
1
6. hΨ+ |Ψ− i == √12 0 1 1 0 · √12 

 1
 
−1 2 [1 − 1] = 0.
0

Solution 8.15. Recall the Bell’s state:

1 1 1 1
Φ+ := √ |00i + √ |11i Φ− := √ |00i − √ |11i
2 2 2 2

154 Q C a m p 2 0 2 3
C H A P T E R 8 . E N TA N G L E M E N T A N D N O N L O C A L IT Y

1 1 1 1
Ψ+ := √ |01i + √ |10i Ψ− := √ |01i − √ |10i .
2 2 2 2
1. For |00i:
1 1 1  1 1  1  1 1 
√ Φ+ + √ Φ− = √ √ |00i + √ |11i + √ √ |00i − √ |11i
2 2 2 2 2 2 2 2
1 1 1 1
= |00i + |11i + |00i − |11i = |00i
2 2 2 2
2. For |11i:
1 1 1  1 1  1  1 1 
√ Φ+ − √ Φ− = √ √ |00i + √ |11i − √ √ |00i − √ |11i
2 2 2 2 2 2 2 2
1 1 1 1
= |00i + |11i − |00i + |11i = |11i
2 2 2 2
3. For |01i:
1 1 1  1 1  1  1 1 
√ Ψ+ + √ Ψ− = √ √ |01i + √ |10i + √ √ |01i − √ |10i
2 2 2 2 2 2 2 2
1 1 1 1
= |01i + |10i + |01i − |10i = |01i
2 2 2 2
4. For |10i:
1 1 1  1 1  1  1 1 
√ Ψ+ − √ Ψ− = √ √ |01i + √ |10i − √ √ |01i − √ |10i
2 2 2 2 2 2 2 2
1 1 1 1
= |01i + |10i − |01i + |10i = |10i
2 2 2 2

Solution 8.16.

1.      
0 0 1 0 1 0
0 0 0 1 1 0 1 1
(X ⊗ I) Φ+ =
b1
· √  = √  = b Ψ+
0 0 0 2 0 2 1
0 1 0 0 1 0
2.      
1 0 0 0 1 1
0 1 0 0 1 0 1 0
(Z ⊗ I) Φ+ =
b · √  = √  = b Φ−
0 0 −1 0 2 0 2 0 
0 0 0 −1 1 −1
3.      
0 0 1 0 1 0
0 0 0 1 1 0 1 1
(ZX ⊗ I) Φ+ =
−1
· √  = √  = b Ψ−
0 0 0 2 0 2 −1
0 −1 0 0 1 0

Solution 8.17.
 
1 0 1 0
h i 1 0 1 0 1
CNOT H ⊗ I = √  
2 0
 1 0 −1
1 0 −1 0

155 Q C a m p 2 0 2 3
C H A P T E R 8 . E N TA N G L E M E N T A N D N O N L O C A L IT Y

     
1 0 1 0 1 1
h i 0 1 0 1 0 0
1. CNOT H ⊗ I |00i = 1
√  · = 1
√  = b |Φ+ i
2 0 1 0 −1 0 2 0
b
1 0 −1 0 0 1
     
1 0 1 0 0 0
h i 0 1 0 1 1 1
2. CNOT H ⊗ I |01i =
 
b √12  · = √1   =b |Ψ+ i
0 1 0 −1 0 2 1

1 0 −1 0 0 0
     
1 0 1 0 0 1
h i 0 1 0 1 0 0
3. CNOT H ⊗ I |10i =
 
b √12  · = √1  =b |Φ− i
0 1 0 −1 1 2  0 

1 0 −1 0 0 −1
     
1 0 1 0 0 0
h i 0 1 0 1 0 1
4. CNOT H ⊗ I |11i =
 
b √12  · = √1  b |Ψ− i
=
0 1 0 −1 0 2 −1

1 0 −1 0 1 0

Solution 8.18. Alice and Bob begin by each having one qubit of the entangled state
1 1
Φ+ AB
= √ |00iAB + √ |11iAB .
2 2
Alice then performs a quantum operation on her qubit, depending on which two bits she want to
send. If she wants to send

• 00: She does nothing to her qubit

  1 1  1 1
IA ⊗ IB √ |00iAB + √ |11iAB = √ |00iAB + √ |11iAB
2 2 2 2
Alice and Bob are still sharing state |Φ+ iAB .

• 01: She applies the operation X her qubit

  1 1  1 1
XA ⊗ IB √ |00iAB + √ |11iAB = √ |10iAB + √ |01iAB
2 2 2 2
Alice and Bob are now sharing state |Ψ+ iAB .

• 10: She applies the operation Z her qubit

  1 1  1 1
ZA ⊗ IB √ |00iAB + √ |11iAB = √ |00iAB − √ |11iAB
2 2 2 2
Alice and Bob are now sharing state |Φ− iAB .

• 11: She applies the operation Z follows by operation X her qubit

  1 1    1 1 
XZA ⊗ IB √ |00iAB + √ |11iAB = XA ⊗ IB √ |00iAB − √ |11iAB
2 2 2 2
1 1
= √ |10iAB − √ |01iAB
2 2
Alice and Bob are now sharing state |Ψ− iAB .

156 Q C a m p 2 0 2 3
C H A P T E R 8 . E N TA N G L E M E N T A N D N O N L O C A L IT Y

In this case, Alice has encoded the 2 bits of classical information into the Bell’ state

• 00 encoded into |Φ+ i.

• 01 encoded into |Ψ+ i.

• 10 encoded into |Φ− i.

• 11 encoded into |Ψ− i.

Alice now sends her qubit to Bob through the noiseless quantum channel. Since Bob has the Bell’s
state analyser, he is able to distinguish the Bell’s states. Hence, he can decode the information
Alice send to him by performing the Bell’s measurement on both of the qubits. If the measurement
outcome is

• |Φ+ iAB then Bob knows the two bits are 00.

• |Ψ+ iAB then Bob knows the two bits are 01.

• |Φ− iAB then Bob knows the two bits are 10.

• |Ψ− iAB then Bob knows the two bits are 11.

We see that Bob always learns the two bits Alice wants to send! This encoding method is known
as superdense coding and it could be summarized neatly as

1 Qubit + 1 Shared Entanglement = 2 Classical Bits

where we can use one qubit (Alice’s qubit) and one share entanglement (|Φ+ i shared by Alice and
Bob) to send 2 bits of classical information.

157 Q C a m p 2 0 2 3
C H A P T E R 8 . E N TA N G L E M E N T A N D N O N L O C A L IT Y

8 .2 B e ll N o n lo c a lit y

Lecture notes by Goh Koon Tong and Cai Yu

Edited by Frits Verhagen

By now, you would have noticed that the outcome of a quantum measurement is not deterministic
in general. We can determine the probability of a measurement outcome but not the outcome
itself! This was something that many physicists in the early 20th century, who were too used to
classical physics, could not accept. In classical physics (and even in special and general relativity),
given the physical laws and initial conditions, one can predict the motion of any particle at any
time precisely.

This led to a popular belief at that time: that quantum physics is an “incomplete theory” and
the random nature of quantum measurements is due to our lack of access to certain information
about the system, so-called “hidden variables”. Their hope was to have a “complete theory” such
that quantum measurements will be deterministic when we also take these “hidden variables” into
account. In 1935, Albert Einstein, Boris Podolsky and Nathan Rosen (EPR) proposed the criteria
for a “complete physical theory”.

8 .2 .1 L o c a l R e a lis m

EPR proposed that any “complete physical theory” has to fulfil two criteria:

1. Locality: Instantaneous cause-and-effect between spatially separated events is not allowed.

Information takes time to propagate through space and its speed of propagation is at most
the speed of light (special relativity).

2. Realism: All measurable quantities of a system must have a pre-existing value, independent
of the observer. The act of measurement can be described as “opening a box and seeing
what is inside”. By opening the box, one does not change the content in the box but obtains
information about the content, which was already there before the measurement took place.

“Local realism” is a collective term for the two above-mentioned ideas: locality and realism. EPR
argued that since quantum theory violates local realism, it has to be an incomplete theory. One
can argue that it is very natural to assume that nature follows the principles of local realism.
Surprisingly, as we shall see, it has been shown in experiments that nature does not follow local
realism. This phenomenon is called “(Bell) nonlocality”.

Note that nonlocality means that the assumption of local realism has been falsified. Hence, it means
that the assumption(s) of realism or locality or both has/have been falsified. No “preferential
treatment” should be given to locality over realism just because the term was coined nonlocality.

158 Q C a m p 2 0 2 3
C H A P T E R 8 . E N TA N G L E M E N T A N D N O N L O C A L IT Y

8 .2 .2 B e ll E x p e r im e n t

In 1964, John Bell proposed certain experiments for which quantum mechanics predicts outcomes
that are not possible if one assumes local realism. Therefore, performing these experiments and
analyzing the measurement outcomes can be used to test if local realism is the correct description
of nature. Such a test is known as a Bell test or a Bell experiment.

In 1969, John Clauser, Michael Horne, Abner Shimony and Richard Holt (CHSH) reformulated
Bell’s work to a form that can be more easily verified in an experiment. This experiment is known
as a CHSH experiment and is a specific type of Bell experiment.

8 .2 .3 C H S H e x p e r im e n t

The setup of the CHSH experiment is as follows: There are two parties (Alice and Bob) who are
separated from each other over a large distance. During the experiment, Alice and Bob are not
allowed to communicate with each other.

Alice and Bob are given a measuring device each, which has two measurement settings labeled as
“0” and “1”. The experiment consists of a number of rounds, and in each round Alice and Bob
both receive a particle that is produced by a common source located between them and make a
measurement on their particle. Every measurement will result in one of the two possible outcomes
labeled as “+1” and “-1”. See Figure 8.3.

Now, for a single round, we denote the measurement settings of Alice’s and Bob’s measuring
devices as x and y respectively. Similarly, we denote the measurement outcomes of Alice’s and
Bob’s measuring devices as a and b respectively. Alice and Bob run a large number of rounds using
different choices for the measurement settings x, y and record their outcomes a, b. After having
completed all the rounds, they use their data to calculate the 16 (there are 4 variables a, b, x, y
that can each take 2 values) probabilities:

Pr(+1, +1|0, 0), Pr(+1, +1|0, 1), Pr(+1, −1|1, 1), etc. (8.20)

Figure 8.3: Setup of CHSH experiment

159 Q C a m p 2 0 2 3
C H A P T E R 8 . E N TA N G L E M E N T A N D N O N L O C A L IT Y

The meaning of such a probability Pr(a, b|x, y) for some values of a, b, x, y is as follows: the prob-
ability that Alice gets a measurement outcome a and Bob b, given that Alice uses the setting x
and Bob y. The objective of the experiment is to determine the value of S, where S is given by:

S = E00 + E01 + E10 − E11 , (8.21)

where the correlation Exy is given by

Exy =Pr(+1, +1|x, y) + Pr(−1, −1|x, y) − Pr(+1, −1|x, y) − Pr(−1, +1|x, y).

One can check that Exy is also equal to the average value of the product ab of the measurement
outcomes when using settings x, y.

8 .2 .4 B e ll in e q u a lit y

In such a setup, it is useful to think of local realism as the existence of some “instructions” from
the source which is carried by the particles to their respective measuring devices. The existence
of these “instructions” reflect the fact that by the principle of realism all measurable quantities
have a pre-existing value. When the particles arrive at Alice and Bob, each measuring device will
decide on which measurement outcome that it will output to its user based solely on the user’s
input setting and the “instruction” from the particle.

In other words, the measurement outcomes of a device ONLY depend on the measurement settings
fed into the device and the “instruction” carried by the particle from the source. Translating this
statement into equations, we have:

Pr(a, b|x, y, λ) = Pr(a|x, y, λ) · Pr(b|x, y, λ)

= Pr(a|x, λ) · Pr(b|y, λ)

where λ denotes the “instruction”. (Note that Pr(a, b|x, y, λ) here means the probability that we
get outcomes a, b given that we use settings x, y and instruction λ.) The first line above represents
the independence between outcomes a and b and the second line represents that the measurement
outcome of a device (i.e. a) does not depend on the setting of the other device (i.e. y). That this
is true is guaranteed by the principle of locality: if we place Alice’s and Bob’s devices far enough
from each other and do both measurements at almost the same moment, the measurements cannot
influence each other. Using the equation above, we can show that the correlations read:

Pr(λ)āx,λ b̄y,λ
X
Exy =
λ

where āx,λ = aPr(a|x, λ) (respectively b̄y,λ ) is the average of a (b) given its corresponding
P
a
settings x (y) and the instruction λ, and Pr(λ) is the probability that the source emits particles
with instruction λ. Thus, the expression of S will be:

Pr(λ) ā0,λ b̄0,λ + ā0,λ b̄1,λ + ā1,λ b̄0,λ − ā1,λ b̄1,λ ≡ Pr(λ)Sλ ,
X
X
S=
λ λ

160 Q C a m p 2 0 2 3
C H A P T E R 8 . E N TA N G L E M E N T A N D N O N L O C A L IT Y

where we have defined Sλ ≡ ā0,λ b̄0,λ + ā0,λ b̄1,λ + ā1,λ b̄0,λ − ā1,λ b̄1,λ .

Notice that at this point, the measurement outcomes given any λ are still probabilistic. However,
we can shift this randomness to the probability distribution of λ. Instead of probabilistic strategies,
we consider deterministic strategies, where given any instruction λ, we have āx,λ , b̄y,λ = +1 or − 1.

To find the maximum (or minimum) value that Sλ can have, one can try out the 16 possibilities
(there are four variables āx,λ b̄y,λ in Sλ and each can take the values +1 or − 1). You will find
that for each possibility −2 ≤ Sλ ≤ 2. By summing over with Pr(λ) (that is like computing the
average), one cannot exceed these values. So we have:

−2 ≤ S ≤ 2.

This result gives an upper bound to a measurable quantity S assuming that local realism holds.

This is known as the CHSH inequality which is the simplest type of Bell inequality. A Bell inequality
is a (usually linear) constraint imposed on a measurable quantity (correlations or probabilities) by
local realism. We conclude that if local realism describes nature, then any Bell inequality must
hold in a Bell experiment.

8 .2 .5 B e ll V io la t io n w it h Q u a n tu m M e c h a n ic s

According to quantum mechanics, there exist some states and measurements which violates the
CHSH inequality. Consider the case where Alice and Bob share the quantum state:
1 1
Φ+ = √ |00i + √ |11i ,
2 2
and measurements:

input outcome +1 outcome −1

A0 |0i |1i
A1 |+i |−i
π 5π
B0 8 8
−π 3π
B1 8 8

where |φi = cos(φ) |0i + sin(φ) |1i. In this notation, |0i , |1i , |+i and |−i correspond to φ =
0, π/2, π/4 and − π/4 respectively.

We would like to compute the probabilities Pr(a, b|x, y). Let us work out the probability of ob-
serving |θi |φi given the state |Φ+ i:

Pr = | hθ| φ Φ+ |2
1
= |(cos θ cos φ h00| + cos θ sin φ h01|
2
+ sin θ cos φ h10| + sin θ sin φ h11|)(|00i + |11i)|
2

1
= (cos θ cos φ + sin θ sin φ)2
2

161 Q C a m p 2 0 2 3
C H A P T E R 8 . E N TA N G L E M E N T A N D N O N L O C A L IT Y

1
= cos2 (θ − φ).
2

Let us calculate E00 as an example and the rest are left as exercise.

Example:

E00 = Pr(00|00) + Pr(11|00) − Pr(01|00) − Pr(10|00)

 
1 π π 5π 5π π π
= cos2 (0 − ) + cos2 ( − ) − cos2 (0 − ) − cos2 ( − )
2 8 2 8 8 2 8
 
1 1 1 1 1 1 1 1 1
= (1 + √ ) + (1 + √ ) − (1 − √ ) − (1 − √ )
2 2 2 2 2 2 2 2 2
1
=√
2

Exercise: Similarly as above, calculate E01 , E10 , E11 .

If computed correctly, we should have

1 1 1 1 √
S = E00 + E01 + E10 − E11 = √ + √ + √ − (− √ ) = 2 2.
2 2 2 2
√
In fact, it can be shown that 2 2 is the maximum value of S allowed by quantum mechanics.
√
S ≤ 2 2 is also known as the Tsirelson’s bound. If we were to perform the CHSH experiment with
√
the measurements and state considered here, we would obtain a value S = 2 2, which violates the
CHSH inequality, which means quantum theory does not obey the principle of local realism.

8 .2 .6 C o n c lu s io n

The violation of Bell inequalities in the laboratories throughout of the world concludes that the
behavior of nature is not constrained by local realism. The first experimental Bell violation was
shown by Aspect et al. in 1982. However, this experiment contained some loopholes that were
hard to overcome. In 2015 loophole-free Bell violation was first demonstrated. Check out the three
papers in 2015, authored by Hensen et al., Giustina, Versteegh et al. and Shalm et al. respectively.
Hence, nature is better described by quantum theory than local realism. While nothing could be
said about whether or not quantum theory is a “complete theory”, at least it has been made known
that EPR’s proposal is wrong. The philosophical implications arising from the nonlocal nature of
the universe are far-reaching. However, Bell violation is nothing more than a direct consequence
of quantum mechanics, so one can argue that nothing is too “spooky” about nonlocality.

162 Q C a m p 2 0 2 3
C h a p te r 9

C ry p to g ra p h y

Edited by Enrique Cervero

Before starting the lecture, please take a few minutes to think about the questions:

• What is cryptography?

• What is the importance of cryptography?

• What are the main types of cryptography?

No matter how much you know about the questions, hopefully, after class is finished, you will get
some ideas for these questions :) Now, let’s get started!!

9 .1 W h a t is C r y p t o g r a p h y a n d W h y is it Im p o r t a n t ?

We approach the first questions in two directions. Firstly, we talk about the origin of the word
“cryptography”. Secondly, the purpose and a general scenario with the commonly-used terminolo-
gies for cryptography are introduced.

History of the word “Cryptography”

Cryptography is the study of secret (crypto-) writing (-graphy). The word cryptography comes
from the ancient Greek word kryptos, meaning “hidden” or “secret”, and the word graphein, mean-
ing “writing”. Therefore, it should make some sense that one important aspect of cryptography is
the process of writing secret messages that can only be understood by those that know the secrets.

Cryptography, in one form or another, has been around for thousands of years. The first instance of
its use is thought to date back to the Ancient Egyptians, around 2000 BCE. Of course, cryptography
is also an important aspect of our modern, data-driven, world. Throughout this vast expanse of
time cryptography has taken many forms and its definition has somewhat evolved.

Purpose and Scenario

163
C H A P T E R 9 . C R Y P T O G R A P H Y

The purpose of cryptography is to protect data transmitted in the likely presence of an adversary.
There are five pillars of cryptology:

• Privacy/Confidentiality - keep communication private. The information can only be

understood by those it is intended for

• Integrity - detect unauthorized alteration to communication. The information may not be

altered without the change being detectable

• Authenticity - confirm identity of sender. The identity of the sender/receiver of the infor-
mation can be identified

• Non-Repudiation - prove that communication was received. The sender/receiver may not
deny or change any messages/information they created/transmitted, at a later time

After talking about the goal of cryptography, next, let’s introduce a general scenario in cryptogra-
phy. A typical cryptography scenario includes two distanced parties, commonly referred to as Alice
and Bob, who want to communicate and transmit some secret information via a probably insecure
channel. A malicious eavesdropper, referred to as Eve, intends to get the hidden information and
remain undiscovered.

To illustrate, Alice firstly encrypts the “plaintext” using a predetermined key, where plaintext is
the unencrypted data that Alice wants to send to Bob. Then, she sends the resulting “ciphertext”
to Bob over the public channel, where ciphertext is encrypted text transformed from plaintext
using an encryption algorithm. Upon receiving the ciphertext, Eve cannot determine what the
plaintext was. However, since Bob knows the encryption key, he can decrypt the ciphertext and
get the plaintext. The encryption and decryption are based upon the type of cryptography scheme
being employed and some form of keys. If we describe the process in formula, this process can be
written as: C = E*k(P) and P = D*k(C), where P = plaintext, C = ciphertext, E = the encryption
method, D = the decryption method, and k = the key.

Notes on terminologies
Some important terminologies to keep in mind moving forward, include

• Eavesdropper - a malicious adversaries whose aim is to circumvent a cryptographic scheme

to impersonate a user, alter data that they do not have permission to change, or gain access
to information that they should not be able to access

• Plaintext - Some message, image, or data that is human comprehend-able

• Ciphertext - Some information that has been modified, so that it is not human comprehend-
able

• Encryption - The process of taking plaintext, and turning it into ciphertext

• Decryption - The process of taking ciphertext, and returning it to the original plaintext

• Key - Some secret that allows for the encryption and/or decryption of information

164 Q C a m p 2 0 2 3
C H A P T E R 9 . C R Y P T O G R A P H Y

What is the importance of cryptography? One may ask why should we care about cryptog-
raphy? Cryptography is an important part of our everyday lives, even if we don’t notice it all of
the time. Let’s show some examples below to know how cryptography is used in our daily life!

• The last time you logged into social media or used your passcode or fingerprint to unlock
your phone; you were authenticating your own identity.

• If you have ever used WhatsApp, Telegram, or iMessage; then you may have seen the phrase
“End-to-End Encrypted”, which implements both data integrity and confidentiality so
that your messages can’t be read or changed in transit to the recipient.

• If you have “read receipts” turned on in any of your messaging apps, then this is a simple
form of non-repudiation, where you can’t later deny that you saw a message...

Imagine a world where anyone could pick up your phone, unlock it, and access all your social media
to say/do anything as if they were you; or, one where anyone could read and edit any messages
you sent before they were received; or, perhaps worst of all, a world where you would never know
if you were being left on “read”. Hopefully, you can see how crazy things could get if it weren’t for
cryptography.

Of course, these examples are on an individual or user level. However, cryptography has a large
impact on nearly every business and country around the world. It is not hard to imagine how a
business may use cryptography to protect customers’ information, secure transactions, and protect
any research and development from falling into the hands of their competitors. Analogously,
governments may want to prevent certain entities from having access to information that may be
a threat to their national security and interests.

In most areas of research and engineering, we are trying to solve problems that arise naturally
from the physical world. For example, a civil engineer who is building a bridge may run into
some engineering challenges; however, once they are solved, there is a bridge that is most likely
going to last. With cryptography, however, we are not only solving problems that arise from the
natural world. Cryptographers and adversaries are continuously trying to outsmart one another.
As cryptographers create more secure cryptographic schemes, adversaries are continuing to look
for vulnerabilities in these schemes. So, in this way, cryptography is a continuously evolving field
of research and engineering.

I hope now you can have more feelings about how cryptography plays such an important role in
our modern world and our everyday lives :)

What are the main types of cryptography? Throughout this lecture, we will be looking at
two types of cryptography: classical cryptography and quantum cryptograpy.

As the name suggests, classical cryptography encompasses the technologies we use for communi-
cation and sharing secrets using current computer architectures. Classical cryptography comes in
two flavours:

165 Q C a m p 2 0 2 3
C H A P T E R 9 . C R Y P T O G R A P H Y

• Secret-Key Cryptography

• Public-Key Cryptography

Secret-Key cryptography, or symmetric cryptography, requires both the sender and the receiver to
previously agree on a secret key which they may use for communication —note that the agreement
of the secret key is a problem in and of itself. Public-Key cryptography, or asymmetric cryptog-
raphy, allows a receiver with a secret key to decrypt ciphers from arbitrary senders, who in turn
encrypt their messages with the receiver’s public key.

On the other hand, the term ‘quantum cryptography’ may refer to two different fields:

• Quantum Key Distribution (QKD)

• Post-Quantum Cryptography

QKD is the study of using quantum phenomena to establish secret keys which may be used as part
of another cryptography scheme. Post-quantum cryptography refers to the study of cryptographic
algorithms which are safe against quantum attacks.

9 .2 P r o b a b ilit y a n d E n tro p y

Cryptography is a mathematical subject and as such it needs to adhere to a strict set of rules and
definitions. Indeed, one cannot prove whether a cryptographic scheme is ‘secure’ if we do not have
a precise mathematical notion of what constitutes ‘security’ —it is obviously not enough to say
‘this looks hard to hack, so it probably cannot be done *shrug*’.

With this in mind, we recall some concepts from (discrete) probability theory that we will need in
this lecture:

A probability is simply a function that maps a set of events to a number between 0 and 1. Sym-
bolically

Pr : Ω 7→ [0, 1]

where the letter Ω (capital Omega) denotes any event.

A random variable is simply the variable whose outcomes are the possible events. If X is a random
variable with outcomes x then we write

Pr[X = x]

to denote the probability that X is the event Ω.

166 Q C a m p 2 0 2 3
C H A P T E R 9 . C R Y P T O G R A P H Y

Example 1:
When throwing a fair six sided dice, an event is any collection of subsets from the set
{1, 2, 3, 4, 5, 6} of possible outcomes. Let X be the random variable which denotes a
possible event (outcome) from a dice throw.

Then we have

1
Pr[X = 1] = Pr[X = 3] = Pr[X = 6] =
6
2
Pr[X = 1 or 4] =
6
1
Pr[X ≤ 3] = Pr[X = 1, 2 or 3] =
2

Of course, if X is a random variable with outcomes x (here x and y could be anything, like the
faces of a dice or a coin, whether the sky is cloudy or sunnny, etc.), then

Pr[X = x] = 1
X

which is to say that the probability of all possible events happening must be 1.

We may also consider the probabilities from more than one random variable at a time, so called
joint probability distributions.

Pr[X = x, Y = y]

Example 2
Let X be the random variable denoting the outcome of a fair six sided dice and Y be the
random variable denoting the outcome of a fair coin.

We have

1 1
Pr[(X, Y ) = (1, Head)] = ·
6 3
1
Pr[X ≤ 3, Y = Head or Tails] = ·1
2

When dealing with more than one random variable, we also define conditional probability distri-
butions. Conditional probability distributions are very important as they may suggest whether an
event is dependent on another, or not.

Let X be a random variable with outcomes x and Y be a random variable with outcomes y. Then

167 Q C a m p 2 0 2 3
C H A P T E R 9 . C R Y P T O G R A P H Y

the probability that X = x conditioned on Y = y is

Pr[X = x, Y = y]
Pr[X = x|Y = y] =
Pr[Y = y]

Example 3:
On any given day in Singapore, it is cloudy with probability p. On cloudy days, it rains
with probability 0.8. Then on any given day

Pr[it is raining AND it is cloudy] = Pr[it is cloudy] · Pr[it is raining|it is cloudy] = p · 0.8

When dealing with probabilistic processes, it is often useful to understand what the expected
outcome will be. The expectation of random variable X is denoted by E[X] and is simply a
weighted average of the outcomes

E[X] = x · Pr[X = x]
X

Example 4
The expected value of a roll of four dices is
6
X 1 1 2 6
x· = + + · · · + = 3.5
x=1
6 6 6 6

Given an outcome x of a random variable X we would also like to have a way to quantify how
‘surprised’ we are at seeing x. If Pr[X = x] is very high, we should not be very surprised — in the
extreme if Pr[X = x] = 1 we should be 0 surprised. Conversely if Pr[X = x] is small, we should
be somewhat surprised —in the extreme if Pr[X = x] = 0 we should be ∞ surprised.

For this purpose, we define the surprisal associated to random variable X

1
S(x) = log2 .
Pr[X = x]

Indeed if Pr[X = x] is high, then 1

Pr[X=x] is small and so its logarithm will also be small. The
opposite is the case when Pr[X = x] is low.

If we want to quantify how surprised we are on average by a random variable X we can take the
weighted average (or expected value) of the surprisal of each outcome of X. This quantity is known

168 Q C a m p 2 0 2 3
C H A P T E R 9 . C R Y P T O G R A P H Y

as the entropy of X and is crucial in the field of information theory

1
H(X) = E[S(X)] = Pr[X = x] · S(x) = Pr[X = x] · log2
X X

x x
Pr[X = x]
.

The entropy is a measure of uncertainty. It tells us how ‘random’ a random variable is: the larger
the entropy, the more ‘random’ the random variable. Conversely, when the entropy is 0 the random
variable is not random at all!

For a random variable X with two outcomes {0, 1} (like a coin flip) such that Pr[X = 0] =
p and Pr[X = 1] = 1 − p, the entropy is

1 1
H(X) = p · log2 + (1 − p) · log2 .
p 1−p
If p = 0 or if p = 1 then the entropy attains its minimum of H(X) = 0. If p = 0.5 then
the entropy achieves a maximum value of H(X) = 1.

We are now ready to go back to cryptography.

9 .3 C la s s ic a l C r y p t o g r a p h y

9 .3 .1 S e c re t-K e y C ry p to g ra p h y

Figure 9.1: Symmetric Encryption Flow Diagram

Secret-Key Cryptography, which can also be called symmetric cryptography refers to any encryp-
tion scheme where both the parties, the sender (Alice) and the receiver (Bob) use same key to

169 Q C a m p 2 0 2 3
C H A P T E R 9 . C R Y P T O G R A P H Y

both encrypt and decrypt. See Fig. 1. Before formally defining the security conditions of a scheme,
let us introduce some historical examples of cryptographic schemes:

Historical examples
In this section, we introduce some examples of symmetric cryptography, where Caesar cipher,
Substitution cipher, Transposition cipher, and One-Time Pad are included.

• Caesar Cipher
The oldest well-known cryptography is the Caesar ciphers, also known as shift ciphers, or
wheel ciphers. Caesar ciphers take the plaintext and shift it by some key value. For example,
if the key-value were 6, then all “a”s in the plaintext would be shifted to “g”s, and so forth
(see Table 9.1).

a b c d e f g h i j k l m n o p q r s t u v w x y z

↓ ↓ ↓ ↓

g h i j k l m n o p q r s t u v w x y z a b c d e f

Table 9.1: Caesar Cipher Mapping with Key-Value 6

Therefore, with our plaintext example from before of “Hello Q-Camp”, and the key-value of
6, we obtain the ciphertext “Nkrru W-Igsv”.

Figure 9.2: Caesar Cipher Paper-Wheel Tool

Figure 2 is a cipher wheel; if you want, you can print them out, label them, and cut them

170 Q C a m p 2 0 2 3
C H A P T E R 9 . C R Y P T O G R A P H Y

out. The cipher wheel can be used as a tool to help with encrypting/decrypting messages.
By placing the smaller wheel inside the large one, and rotating it, you can see the mappings
of each character.

• Substitution Cipher
Substitution ciphers are similar to Caesar ciphers in that they involve mapping the characters
of the plaintext onto other characters. However, instead of shifting all the characters by the
same number of characters, every character can be mapped onto any other character. See
Table 9.2 for an example from plaintext“Hello Q-Camp” to the ciphertext “Fpaab U-Krce”.

a b c d e f g h i j k l m n o p q r s t u v w x y z

↓ ↓ ↓ ↓

r z k v p o x f w l y a c q b e u g n j t d i m h s

Table 9.2: Substitution Cipher Mapping. Under this mapping, The plaintext “Hello Q-Camp”
turns into the ciphertext “Fpaab U-Krce”.

Note that the mapping shown in Table 9.2 does not follow any set shift or pattern. Unlike
the Caesar cipher, which only requires us to know the shift; substitution ciphers must use
the mapping itself as the key. This means that the key size will be significantly larger.

Again, this is not the only way to implement a substitution cipher. It is also possible to map
a single letter onto a group of letters, which gives some improved security.

Security of an encryption scheme

Definition A symmetric scheme is composed of three elements:

• A key generation algorithm, Gen, which takes a security parameter (a number n) and outputs
a key k from a set of keys K, written Gen(n) = k.

• An encryption algorithm, Enc, which takes two arguments: the key k and a message m from
a set of messages M and outputs a cipher c from a set of ciphers C, written Enck (m) = c.

• A decryption algorithm, Dec, which takes two arguments, the key k and the cipher c and
outputs a message m0 , written Deck (c) = m0 .

Definition We say that cryptographic scheme composed of Gen, Enc, Dec is robust if

Deck (Enck (m)) = m

which is to say that the decryption of an encrypted message returns the original message. This
seems like an obvious requirement but it is important in the case where the encryption or decryption
algorithms may be probabilistic functions.

171 Q C a m p 2 0 2 3
C H A P T E R 9 . C R Y P T O G R A P H Y

Definition We say that a cryptographic scheme (Gen, Enc, Dec) is perfectly secret if for every
probability distribution with which the message m is selected

Pr[M = m|C = c] = Pr[M = m].

Intuitively this equation requires that the probability of getting message m if you know the ci-
phertext c is the exact same as the probability of simply guessing the message. Namely, that the
ciphertext c contains no information about the message m.

It is worth noting that none of the historical examples above satisfy the secrecy definition, and can
be easily broken with any (not so) modern computer.

An equivalent definition of perfect secrecy for a scheme (Gen, Enc, Dec) is based on the following
thought experiment:

1. Adversary creates two messages m0 and m1 of the same length,

2. An honest party encrypts either m0 or m1 at random (without sharing which) to obtain

ciphertext c,

3. The ciphertext c is given to the adversary, who outputs a guess of what message was used to
create it.

Definition Scheme (Gen, Enc, Dec) is perfectly secret if

1
Pr[Guess is correct] = .
2
Namely, the best strategy for the adversary to distinguish whether the cipher came from message
m0 or message m1 is to guess at random!

The One-Time Pad

Let us now introduce an encryption scheme that satisfies the definition of perfect secrecy, the
one-time pad.

First, let us define bitwise addition. Recall that a bit b is either a 0 or a 1. the bitwise exclusive-or
of two bits a and b is denoted by a ⊕ b and works as follows:
⊕ a=0 a=1
b=0 0 1
b=1 1 0
In particular note that b ⊕ b = 0 always. If a strings of n bits is denoted by b = b1 , ..., bn , then

a ⊕ b = (a1 ⊕ b1 ), ..., (an ⊕ bn ).

In particular note that b ⊕ b = 0 always.

The one-time pad (OTP) works as follows:

1. The key generation algorithm outputs a random string of bits k of length n, that is

Gen(n) = k = k1 , k2 , ..., kn

172 Q C a m p 2 0 2 3
C H A P T E R 9 . C R Y P T O G R A P H Y

2. The encryption algorithm takes in a message m of at most n bits and outputs

Enck (m) = k ⊕ m = m1 ⊕ k1 , ..., mn ⊕ kn .

3. The decryption algorithm takes in a cipher c and outputs

Deck (c) = c ⊕ k.

Theorem The OTP is robust and perfectly secret.

Proof:
For robustness, we need to show that for any message m, we have that Deck (Enck (m)) = m, that
is, decryption of an encrypted message yields the original message. Using the definition of the
OTP, we have

Deck (Enck (m)) = Deck (k ⊕ m)

= k ⊕ (k ⊕ m)
=k⊕k⊕m
= m,

where in the last step we used the fact that k ⊕ k = 0 and that 0 ⊕ m = m. This concludes the
proof of robustness.

For secrecy we need to show that if an adversary submits two messages m0 and m1 to the OTP,
receives a cipher c of one of the two messages and is asked to guess which message was used, then
Pr[Guess is correct] = 12 .

Let mb for b = 0 or 1 be the message that was secretly encoded, such that c = Enck (mb ). By
construction, the key k is a random string of n bits, the cipher c = k ⊕ mb is also a random string
of n bits. This implies that both possible ciphers k ⊕ m0 and k ⊕ m1 look completely random to
the adversary and so their best strategy is to guess randomly between the two messages, and so
we have Pr[Guess is correct] = 12 . This concludes the proof.

It is worth noting that the OTP has not seen a lot of use in practical cryptography even though it
boasts the highest level of security. There are two reasons for this.

• It is very impractical to use as the length of the key needs to be as large as the length of the
message. In practice this is a big problem as messages can be arbitrarily long and generating
a very long random key is very hard using classical computers.

• As the name suggests, each key in the one time pad can only be used once! This is because
if you encode two different messages m0 and m1 with the same key, and the two resulting
ciphers c0 = m0 ⊕ k and c1 = m1 ⊕ k are intercepted, then information about the messages
can be leaked to the eavesdropper. This is because

c0 ⊕ c1 = (m0 ⊕ k) ⊕ (m1 ⊕ k) = m0 ⊕ m1

173 Q C a m p 2 0 2 3
C H A P T E R 9 . C R Y P T O G R A P H Y

as the key is cancelled out!

In fact, it turns out that perfect secrecy is too strong a requirement in practice. It has been shown
that for a cryptosystem to be perfectly secret, the keys, ciphers and messages need to all be the
same length, which is a very impractical requirement as we have just discussed!

Currently deployed cryptosystems use a slightly weaker notion of secrecy which allows for the use
of shorter keys and longer messages. The weaker secrecy is based on the notion of computational
complexity, and simply requires that a cryptosystem is secure against attackers with a limited
amount of time. In contrast, perfect security does not put a time limit on the attackers. More
formally:

Definition Suppose that an adversary is playing the guessing game with scheme (Gen, Enc, Dec)
in which he submits two messages m0 and m1 and is asked to guess which message corresponds to
the cipher Enck (mb ) where mb is the randomly chosen message.

We say scheme (Gen, Enc, Dec) is secret against polynomial time bounded adversaries if
1
Pr[Guess is correct] ≤ + ε,
2
where ε is a (exponentially) small positive number.

In the above polynomial time adversary simply means that an adversary is bounded to attacks
that are bounded in a time that is a polynomial function of the length of the key used for Enc and
Dec.

P u b lic - K e y C ry p to g ra p h y

Figure 9.3: Asymmetric Encryption Flow Diagram

Public-Key Cryptography or asymmetric cryptography, refers to any encryption scheme where two
different keys are generated; a public key and a private key. Public keys are shared with everyone
but may only be used for encryption whereas private keys are never to be shared because they are
used for decryption.

174 Q C a m p 2 0 2 3
C H A P T E R 9 . C R Y P T O G R A P H Y

This kind of scheme is unique when compared to those described thus far, and is relatively new;
only dating back to the 1970s. Asymmetric encryption solves the key-distribution problem faced
by symmetric encryption schemes, but not without trade-offs.

Examples
A commonly used public-key cryptosystem is RSA, which is named after its three developers Ron
Rivest, Adi Shamir, and Leonard Adleman. The RSA Encryption Scheme is often used to encrypt
and then decrypt electronic communications.

The idea of RSA is based on the fact that it is difficult to factorize a large integer. The public
key consists of two numbers where one number is the multiplication of two large prime numbers.
The private key is also derived from the same two prime numbers. So if somebody can factorize
the large number, the private key is compromised. Thus, encryption strength totally depends on
the key size. If we double or triple the key size, the strength of encryption increases exponentially.
RSA keys can be typically 1024 or 2048 bits long, but experts believe that 1024 bit keys could be
broken soon. But till now it seems to be an infeasible task.

The RSA protocol works as follows:

Suppose Alice asks Bob to encrypt email messages before sending her. Since computers represent
text as numbers, e.g., 01 for “A”, 02 for “B”, the encrypted message is a very big number.

• Alice’s Setup:

– Picks two prime numbers.

– Calculates the product n = pq.
– Calculates m = (p-1)(q-1).
– Picks numbers e and d so that ed has a remainder of 1 when divided by m.
– Announces her public key (n, e).

• Bob encrypts the message M for Alice:

– Finds Alice’s public key (n, e).

– Finds the remainder C when M e is divided by n.
– Sends ciphertext C to Alice.

• Alice receives and decrypts ciphertext C:

– Uses her private key (n, d).

– Finds remainder R when C d is divided by n.
– R matches the message M that Bob wants to send to Alice!

Below gives a specific example, where Bob aims at encrypting the message M = 14.

• Alice’s setup:

– p = 11 and q = 3
– n = pq = 11 × 3 = 33.

175 Q C a m p 2 0 2 3
C H A P T E R 9 . C R Y P T O G R A P H Y

– m = (p - 1)(q - 1) = 10 × 2 = 20.
– If e = 3 and d = 7, then ed = 21 has a remainder of 1 when divided by m = 20.
– Announces (n, e) = (33, 3).

• Bob encrypts a message M = 14

– (n, e) = (33, 3).

– When 143 = 2744 is divided by 33, the remainder is C = 5.
– Sends ciphertext C = 5 to Alice.

• Alice receives and decrypts ciphertext C:

– (n, d) = (33, 7).

– When 57 = 78125 is divided by 33, the remainder is R = 14.
– R = 14 = M, the original message from Bob!

After introducing the protocol and example of RSA, let’s next talk about the concern of RSA.
The security of RSA relies on the practical computational difficulty of factoring the product of two
large prime numbers. One may think that what if we have a really powerful computer? As far as
is known, no classical algorithm is known that can factor integers in polynomial time. However, if
we have an ideal quantum computer, then Shor’s algorithm shows that it could defeat RSA. Thus,
this is one of the powerful motivators for the design and construction of quantum computers, and
the study of new quantum-computer algorithms.

Finally, I hope, after this lecture, you can answer the three questions mentioned on the first page
of the lecture notes :)

9 .4 Q u a n tu m C ry p to g ra p h y

Quantum cryptography may refer to two different fields, quantum key distribution (QKD) and
post-quantum cryptography.

Quantum computers are heralded to be powerful enough to break currently deployed cryptographic
infrastructures used in the current world. The famous Shor algorithm is a quantum algorithm
known to efficiently solve the prime factorization problem. In fact, it is the only known algorithm,
classical or quantum, to be able to do this. The prime factorization problem is the primitive for
the public-key cryptographic scheme RSA, so Shor’s algorithm ability to solve prime factorization
threatens the security of RSA encryption.

Post-quantum cryptography deals with this exact scenario. This field encompases the study of cryp-
tographic schemes (classical or quantum) which are safe against any attacks (classical or quantum).
Such schemes include classical examples like lattice based cryptography and hash based cryptography
and quantum examples like homomorphic encryption.

176 Q C a m p 2 0 2 3
C H A P T E R 9 . C R Y P T O G R A P H Y

On the other hand, quantum key distribution deals with the distribution of keys for subsequent
cryptographic schemes, using quantum phenomena to guarantee the secrecy of the distributed keys.

In this chapter we will mainly discuss quantum key distribution.

Q u a n tu m K e y D is t r ib u t io n

The task of QKD is for two honest parties, Alice and Bob, to establish a secret key in the presence of
an eavesdropper, Eve. The established secret key is to be used for subsequent secure communication
via e.g. another cryptographic protocol.

The advantage of quantum key distribution over classical key distribution is that quantum phe-
nomena can provide Alice and Bob with access to a source of monogamous correlations. This
means that using quantum particles, Alice and Bob can generate randomness between them whilst
having provable certainty that an eavesdropper cannot be ‘sharing in’ the randomness. This is
enough for a key distribution protocol because intuitively, a secret key is nothing but a random
string of bits agreed upon by Alice and Bob and unknown to Eve.

The QKD protocol and security definitions

In QKD, Alice and Bob wish to establish a secret key in the presence of a malicious eavesdropper.

Figure 9.4: The QKD scenario. Alice and Bob are in possession of some shared quantum states.
They perform a series of local operations (in their respective labs) followed by some public com-
munication. At the end of the protocol they output respective keys KA , KB .

Before providing a protocol and definitions of what constitutes a secure QKD protocol vs an
insecure QKD protocol, we need to formally discuss the ‘power’ of the eavesdropper Eve.

• Is Eve passive and can only listen to the public announcements by Alice and Bob?
• Can Eve perform quantum operations on a quantum computer? Are these operations noisy

177 Q C a m p 2 0 2 3
C H A P T E R 9 . C R Y P T O G R A P H Y

or error-corrected?

• Does Eve have control of the measurement devices of Alice, Bob, or both?

• Can Eve intercept and alter the quantum states shared between Alice and Bob?

• Does she have a quantum memory?

To achieve the maximum level of security, one needs to assume that the eavesdropper is powerful.
For QKD, we assume Eve possesses a powerful quantum computer and can perform any operations
on her states without introducing errors. Further, she is assumed to have a share of the three
partite state shared amongst Alice, Bob and herself.

We also need to discuss what kind of power Eve has over the devices and quantum states used
by Alice and Bob. Does Eve control the measuring instruments? Does she prepare and distribute
the quantum states used by Alice and Bob? Here we need to make a distinction between Device-
Dependent QKD (DD-QKD) and Device-Independent QKD (DI-QKD).

Definition: In DD-QKD, the devices and states of Alice and Bob are characterised. That is, Alice
and Bob know what operations their measuring instruments perform and they know what states
they share.

In DD-QKD security can be guaranteed by the property of entanglement, which you have seen
in previous lectures. To be precise, it has been shown that entanglement is necessary for the
distillation of a secret key. Interestingly enough, it is not known whether entanglement is also
sufficient for this task.

Definition: In DI-QKD, the devices and states of Alice and Bob are un-characterised. That is,
Alice and Bob do not know what operations their measuring instruments perfrom they do not
know what states they share.

In the definition of DI-QKD it is imperative to assume that whilst Alice’s and Bob’s devices are
un-characterised and may have even been prepared and distributed by the eavesdropper, that they
DO NOT communicate with Eve. More precisely, we assume that the devices do not leak (a lot
of) information to Eve, otherwise no security can be guaranteed.

It is surprising property of quantum phenomena that any secret key can be generated if Alice and
Bob have no control or knowledge over the states they share and the operations they perform.
Indeed, security of DI-QKD can be guaranteed based on non-locality, which you have also seen in
previous lectures. In fact, it has been shown that non-locality is both necessary and sufficient to
distill secret keys.

Now we are ready to discuss a general QKD protocol (either DD or DI), which is usually composed
of the followin gfive different stages:

1. Measurements: In this first stage, Alice and Bob independently choose from a set of mea-
surement settings, say setting x for Alice, setting y for Bob. They then perform said measure-

178 Q C a m p 2 0 2 3
C H A P T E R 9 . C R Y P T O G R A P H Y

ments on their respective shares of a quantum state. Since measurements are probabilistic,
the measurements will yield some outcomes a for Alice and b for Bob. The probabilities of
the outcomes from the inputs are Pr[a, b|x, y]. These are known as the correlations between
Alice and Bob. It is the task of Alice and Bob to distill a key from these correlations whilst
ensuring Eve does not have much information about the correlations.

2. Parameter Estimation: In this stage Alice and Bob will publicly compare (a subset of)
their input-output pairs to see if the correlations Pr[a, b|x, y] are what they expect, or if a
third party has tampered with the experiments. If the eavesdropper is detected, the protocol
is aborted.

3. Key Sift: In this step, Alice and Bob produce raw keys from the inputs and outputs of the
measurements.

4. Error Correction: Here, Bob forms a guess for Alice’s raw key using his own raw key and
some public information announced by Alice. This step serves to correct any possible errors
introduced in the protocol via the eavesdropper or simply by experimental error. If Bob’s
guess for Alice’s raw key is not good enough, the protocol is aborted.

5. Privacy Amplification: In this final stage, Alice and Bob perform an operation on their
respective raw keys to further remove any residual information the eavesdropper might have.
Alice then outputs her final key KA , and Bob outputs his, KB .

A ‘sensible’ QKD protocol should satisfy the following three requirements:

• Completeness: If all parties are behaving honestly (that is, no eavesdroppers or cheaters),
the probability that the protocol aborts should be small. Formally, if we let Ω be the event
that the protocol does not abort, such that Ωc (Omega complement) is the even that the
protocol does abort, then

Pr[Ωc ]hon ≤ εcom ,

where εcom (epsilon-complete) is a small positive number and the subscript hon denotes the
honest behaviour.

• Correctness: Provided that the protocol does not abort, we require that Alice’s and Bob’s
outputs are the same except with small probability. Formally:

Pr[Ω and KA 6= KB ] ≤ εcorr ,

where KA and KB are the keys output by Alice and Bob, Ω is the event of not aborting, and
εcorr (epsilon-correct) is a small positive number.

• Secrecy: To guarantee that the keys output by Alice and Bob (which are guaranteed to be
the same by correctness if the protocol does not abort) are secret, they need to be perfectly
random and completely uncorrelated to the eavesdropper. Formally, let ρKA E∧Ω be the

179 Q C a m p 2 0 2 3
C H A P T E R 9 . C R Y P T O G R A P H Y

quantum state at the end of the protocol which contains Alice’s key KA and all of Eve’s
information E provided the protocol did not abort (denoted by the symbol ∧Ω). We require

ρKA E∧Ω ≈ τKA ⊗ ρE∧Ω ,

where τKA is a completely random state containing Alice’s key, ρE∧Ω is an arbitrary quantum
state containing Eve’s information provided the protocol did not abort. The symbol ⊗
denotes that states τKA and ρE∧Ω are separated, i.e. they are uncorrelated.

Usually, the Completeness and Correctness of a QKD protocol are guaranteed by the error correc-
tion step of a protocol. The crux of the security proof therefore lies in showing that a protocol is
secret, namely that the key is random and uncorrelated to the eavesdropper.

To prove secrecy, one also needs to have an understanding of the types of attacks that an eaves-
dropper can perform:

• Collective attacks require the eavesdropper to perform the exact same attack at every round
of the protocol. That is, Eve behaves identically throughout every round of measurements.
Collective attacks model the situation where Eve does not have any quantum memory and
so must produce a guess for Alice’s and Bob’s outputs at every round of the protocol by
performing a measurement on her quantum state for that round.

• General attacks on the other hand are much more general as they allow the eavesdropper
to have a quantum memory and can wait until the very end of the protocol to output her
guess for the secret key. In particular, in general attacks Eve can wait until all the public
information has been exchanged to maximize her side information before she has to measure
her quantum states to produce her guess for the secret key.

Clearly, general attacks are much more powerful than collective attacks and so proving secrecy
against them guarantees a much higher level of security. Nevertheless, this is often a much harder
proof.

180 Q C a m p 2 0 2 3
Friday

181
C h a p te r 10

Q u a n tu m C h e m is t r y

Lecture by Chee Chong Hian

No Lecture Notes. Recommended quantum chemistry textbooks for further reading:

1. LibreTexts Chemistry, Chem1201, Mount Royal University, (2023)

2. Piela, L., Ideas of quantum chemistry, 2 vols, 3rd edn., Elsevier (2020)
3. A. Szabo and N. S. Ostlund, Modern Quantum Chemistry: Introduction to Advanced Electronic
Structure Theory, Courier Corporation (1996)

182
C h a p te r 11

Q u a n tu m J o u r n a l C lu b

11.1 L is t o f p a p e r s

1. Quantum optics. C. C. Gerry and P. L. Knight, Quantum superpositions and Schrödinger

cat states in quantum optics. American Journal of Physics, 65, 964-974 (1997).

2. Quantum information. N. Gisin and S. Popescu, Spin flips and quantum information for
anti-parallel spins. Physical Review Letters, 83, 432 (1999).

3. Quantum game theory. J. Eisert, M. Wilkens and M. Lewenstein. Quantum games and
quantum strategies. Physical Review Letters, 83, 3077 (1999).

4. Quantum biology. S. Gerlich, S.Eibenberger, M. Tomandl, S. Nimmrichter, K. Hornberger,

P. J. Fagan, J. Tüxen, M. Mayor and M. Arndt, Quantum interference of large organic
molecules. Nature Communications, 2, 263 (2011).

183