Sold to

[email protected]

THE GOAL SHOULD BE, NOT TO IMPLANT IN THE STU-

D E N T S ’ M I N D E V E R Y F A C T T H AT T H E T E A C H E R K N O W S

N O W ; B U T R AT H E R T O I M P L A N T A W AY O F T H I N K I N G T H AT

E N A B L E S T H E S T U D E N T, I N T H E F U T U R E , T O L E A R N I N

O N E Y E A R W H AT T H E T E A C H E R L E A R N E D I N T W O Y E A R S .

O N LY I N T H AT W AY C A N W E C O N T I N U E T O A D V A N C E F R O M

O N E G E N E R AT I O N T O T H E N E X T.

E D W I N J AY N E S

T H E I M P O R TA N T T H I N G I N S C I E N C E I S N O T S O M U C H T O

O B TA I N N E W F A C T S A S T O D I S C O V E R N E W W AY S O F T H I N K -

ING ABOUT THEM.

S I R W I L L I A M L AW R E N C E B R AG G
JA KO B S C H W I C H T E N B E R G

PHYSICS
FROM
FINANCE

NO-NONSENSE BOOKS
First printing, May 2020

Copyright © 2020 Jakob Schwichtenberg

With Illustrations by Boris Quade, Hamburg

All rights reserved. No part of this publication may be reproduced, stored in, or
introduced into a retrieval system, or transmitted in any form or by any means
(electronic, mechanical, photocopying, recording, or otherwise) without prior
written permission.

unique id: 9bae9ef47751953cfc1f57b8c4cc9713fccaca56812257a6a10f5dfdc01c06c8

Each copy of Physics from Finance has a unique ID which helps to prevent illegal sharing.

book edition: 1.8

Dedicated to my parents
Preface

There’s a deep and beautiful connection between finance

and physics. But so far, this connection was used almost ex-
clusively by economists to get a better grip on the financial
market.
economists

The main idea of this book, however, is to use the connec- physics finance
G
tion between finance and physics to translate in the opposite
this book
direction. Specifically, the goal of this book is to use a toy
version of the financial market to explain fundamental mod-
els of nature like general relativity or electrodynamics.1 1
Already at this point,
be assured that no
previous knowledge of
Maybe this sounds like a stupid idea. After all, finance is finance theory is neces-
an extremely difficult subject. Wouldn’t it be easier to learn sary to understand this
book.
physics directly? Well, I’m convinced that it makes sense to
spend a few minutes to think about finance first. Otherwise
I wouldn’t have spent the time to write this book. But let me
explain.

Modern physics often appears like a conglomerate of several

seemingly unrelated models.2 These models are usually 2
Quantum mechanics,
discussed completely independently. There are books on general relativity,
special relativity,
general relativity, books on quantum mechanics, books electrodynamics, . . .
on electrodynamics and students read them in different
8

3
This is known as the
fiber bundle perspec- academic terms.
tive. Fiber bundles are
a mathematical con-
cept which provides a However, it turns out that the key ideas at the heart of all
unified framework for
all modern theories of these models are exactly the same. Therefore, by studying
physics. a small number of key ideas, we can understand them all
4
Some popular science from a common perspective.3
books try to explain
some of these key ideas
using analogies. How- Now the bad news is that these key ideas are somewhat ab-
ever, these analogies stract. We can’t dumb them down to the level of apples and
don’t represent the
main features of the oranges. This is why they are not explained to beginning
ideas faithfully. This students and why they are rarely discussed in popular sci-
leads to lots of prob- ence books.4 We can, however, understand them by talking
lems when readers try
to read more technical about a toy model of the financial market.
texts afterwards. On
the other hand, analo-
gies are undoubtedly This toy model is not trivial and we need some time to think
extremely powerful it through. But afterwards, you will not only be able to
tools. It can even be understand the key ideas at the heart of modern physics but
argued that all fun-
damental advances in also see the connections between the various fundamental
physics were due to models.
some analogy. (This is
discussed extensively
in the fantastic book
Surfaces and Essences
by Douglas Hofstadter
and Emmanuel Sander.) Financial Toy Model / Quantum Mechanics
And, in some sense,
this book makes heavy
use of analogies, too. *
But the main difference Electrodynamics
between the analogies
commonly used in
popular science books  &
and the finance analo- General Relativity Weak and Strong Interactions
gies discussed in this
book is that they are, I
believe, as close to the
real thing as it gets. In
That’s why I’m confident that learning a thing or two about
particular, this means
that you won’t run into the financial market is one of the best investments of your
confusing contradic- time that you’ll ever make.
tions when you read
more advanced books
afterwards. With all that said, let me make a few things clear.
 9

B This is not a book on finance theory. We will not try

to build a realistic model of the financial market. In-
stead, we will only use those parts of the financial market
which help us to understand fundamental physics.
B After reading this book you won’t know everything.
There are lots of details missing and many important
concepts are not even mentioned. However, this book
can do two things for you. First of all, it can help you to
see the big picture which often gets lost in more detailed
treatments. In other words, you will be able to see the
whole forest and not just individual trees. And secondly,
it lays the foundation for further studies. By the end of
this book you will be equipped with a completely new
set of tools which allows you to explore modern physics
from a beautiful common perspective.
B This is not a popular science book, but it’s also not a
textbook. It’s something in between. While the following
chapters contain mathematical equations, the number of
mathematical arguments is kept to a minimum and no
attempt at mathematical rigor is made. I hope that this
way laypersons and students alike can benefit from the
following explanations. In particular:
– If you’re a somewhat advanced student of physics, this
book allows you to understand many things you prob-
ably already know from a completely new perspective.
– But even if you have no formal physics education, this
book will allow you to understand how nature works
at a fundamental level. The only prerequisites are a
rough understanding of calculus (e.g. what a deriva-
tive is) and complex numbers. Undoubtedly, some
details will be harder or even impossible to understand
without any foreknowledge. But if you feel stuck, you
can simply glance over these details and move on. Ev-
ery important idea will be explained multiple times
from various perspectives, with and without equations.
10

B I did my best to keep this book as short as possible. In

particular, I removed all unnecessary tangents and sec-
tions discussing advanced concepts because I believe
they would only distract you from the truly fundamen-
tal ideas. In addition, I hope the shortness of this book
demonstrates that you don’t need to spend months of
your life to understand the essentials of modern physics.
B The shortness of this book also means that after reading it
you will probably have more questions than you started
with. However, I think this is the best thing a book on
physics can accomplish because it means your curiosity
has been sparked. In Chapter 6, we will discuss in detail
which books you should read if you want to dive deeper.

So if you think that this sounds like a solid plan, without

any further ado, let’s dive in. Hopefully you’ll enjoy reading
this book as much as I have enjoyed writing it.

Karlsruhe, January 2019 Jakob Schwichtenberg

PS: If you find an error, I would appreciate a short email to

[email protected].

Acknowledgments

Special thanks to Luc Longtin whose comments, ideas and

corrections have made this book so much better. Moreover,
I want to thank Florian Colbatzky, Andreas Pargner and
Jacob Ayres for carefully proofreading the manuscript and
Jerry D. Logan, Nam-Kyu Park, Robert Welters and Don
Washburn for reporting several typos. I’m also indebted to
Paul Tremper for many insightful discussions that inspired
me to write this book.
Contents

Part I The Physics of Finance

1 Finance Intuitively 19
1.1 Symmetry . . . . . . . . . . . . . . . . . . . . . . 20
1.2 Gauge Symmetry . . . . . . . . . . . . . . . . . . 21
1.3 Internal Space . . . . . . . . . . . . . . . . . . . . 23
1.4 Connections . . . . . . . . . . . . . . . . . . . . . 25
1.5 Gauge Invariant Quantities . . . . . . . . . . . . 26
1.6 Curvature . . . . . . . . . . . . . . . . . . . . . . 31
1.7 Dynamical Curvature . . . . . . . . . . . . . . . . 32

2 Finance Mathematically 35
2.1 Arbitrage Dynamics . . . . . . . . . . . . . . . . 51
2.2 Trader Dynamics . . . . . . . . . . . . . . . . . . 56
2.2.1 Trader Dynamics and Arbitrage . . . . . 62

Part II The Physics of Nature

3 Elementary Particles 71

4 Quantum Mechanics 77
4.1 Charge Space . . . . . . . . . . . . . . . . . . . . 81
4.2 Electrodynamics . . . . . . . . . . . . . . . . . . . 87
4.3 Interactions in Quantum Mechanics . . . . . . . 90
12

4.4 Isospin Space and Color Space . . . . . . . . . . 92

4.4.1 Symmetries of Isospin and Color Space . 97
4.4.2 Isospin and Color Connections . . . . . . 97
4.4.3 Weak and Strong Interactions . . . . . . . 100

5 Special Relativity and General Relativity 107

5.1 Special Relativity . . . . . . . . . . . . . . . . . . 109
5.2 General Relativity . . . . . . . . . . . . . . . . . . 117
5.2.1 Gravity and Curvature Dynamics . . . . 121
5.3 Summary and a Few Loose Ends . . . . . . . . . 124

6 Closing Words 129

6.1 Reading Recommendations . . . . . . . . . . . . 132
6.2 Outlook . . . . . . . . . . . . . . . . . . . . . . . . 136

One Last Thing

Part V Appendices

A Statistics 141
A.1 Probability Distribution . . . . . . . . . . . . . . 141
A.2 Mean . . . . . . . . . . . . . . . . . . . . . . . . . 142
A.3 Standard Deviation . . . . . . . . . . . . . . . . . 143

B Metrics 147

C Group Theory 151

C.1 Examples . . . . . . . . . . . . . . . . . . . . . . . 154
C.2 Geometry of Symmetries . . . . . . . . . . . . . . 156
C.3 Lie Algebras . . . . . . . . . . . . . . . . . . . . . 157

D Fiber Bundles 163

D.1 Representation Theory . . . . . . . . . . . . . . . 169
D.2 Associated Bundles . . . . . . . . . . . . . . . . . 173
D.3 Parallel Transport . . . . . . . . . . . . . . . . . . 176
D.4 Connections . . . . . . . . . . . . . . . . . . . . . 178
D.5 Curvature . . . . . . . . . . . . . . . . . . . . . . 183
 13

E Taylor Expansion 187

Bibliography 193

Index 197
This page is intentionally left blank
 Part I
The Physics of Finance

"We all wish that we had these three laws that explain 99% of all behavior.
In fact, economists have 99 laws that explain maybe 3% of economic
behavior."

Andrew Lo
1

Finance Intuitively

The main idea of this book is that we can understand many

of the most important insights in modern physics as dis-
coveries about the arena modern physics takes place in. In
particular, physicists discovered that there is not just space-
time but also internal spaces. Spacetime, together with these
internal spaces, yields the arena that we use to make sense
of modern physics.

Arguably the most important insight was that this arena is

not some kind of static background structure but a dynam-
ical part of nature itself. Describing the interplay between
elementary particles and this arena is what modern physics
is all about.

influence
#
Elementary Particles Arena of Physics
c
influences
20 physics from finance

So to understand modern physics we need to answer a few

key questions:

B What’s an internal space?

B How can the structure of an internal space be non-trivial?
B Where does such a non-trivial structure come from?
B How does the non-trivial structure affect objects in the
system?

Admittedly all of this sounds horribly abstract.

1
We will only talk
about aspects of the But, as mentioned already in the preface, we can answer all
financial market which
are directly relevant these questions by thinking about the financial market. This
to modern physics. is a useful analogy because, while financial transactions take
The following connec-
tion between financial place at locations in the real world (spacetime), we need
markets and funda- abstract internal spaces to keep track of them.
mental physics was
first put forward by
Malaney and Wein- Maybe you don’t care about financial markets at all. But let
stein, as summarized me assure you that it makes sense to think about the follow-
in Ref. [Malaney, 1996]
and later popularized ing simplified version of the financial market. As soon as we
in Refs. [Ilinski, 1997, understand how we can describe what is going on here con-
Young, 1999, Malda- sistently, we automatically understand how modern physics
cena, 2016]. The ideas
were further devel- works, too.1
oped in [Vazquez and
Farinelli, 2009]. After
talking about the key We will start by talking about symmetries. Symmetries
concepts in the context are something which we can grasp intuitively and there
of the financial mar- is a close connection between the internal spaces and the
ket, we will use them
directly to understand symmetries of a system.
how nature works.
The main difference
between what we do in
the financial toy model
and in modern physics 1.1 Symmetry
is that we use different
internal spaces.

First of all, what is a symmetry?

 finance intuitively 21

Imagine a friend stands in front of you and holds a per-

fectly round ball in her hand. Then you close your eyes,
your friend performs a transformation of the ball and after-
ward you open your eyes again. If she rotates the ball while
your eyes are closed, it is impossible for you to determine
whether or not she did anything at all. Hence, rotations are
symmetries of the ball. But if she moves the ball to another
location, you will immediately notice it. Therefore, transla-
tions are not symmetries of the ball.

2
In mathematical terms,
Moreover, if she holds a cube, only very special rotations we usually collect
all symmetries of an
can be done without you noticing it. In general, all trans- object and call this set
formations which, in principle, change something but lead of transformations,
together with a rule
to an indistinguishable result are symmetries. Formulated which allows us to
differently, a symmetry takes us from one state to a different combine them, a group.
one, which happens to have the same properties.2 In other words, a
group consists of
all transformations
With this in mind, let’s talk about an important symmetry of which leave a specific
object unchanged and
the financial market. Talking about this symmetry will lead a rule which allows
us directly to the internal space of our financial toy model. us to combine two
transformations in
such a way that we
get another symmetry
transformation. This is
discussed in a bit more
1.2 Gauge Symmetry detail in Appendix C.

The key observation we need is that prices do not have any 3

Maybe you wonder
absolute meaning. We can shift them arbitrarily, as long as what happens when
different countries use
all prices get shifted equally.3 different currencies.
Even in that case, prices
still have no absolute
meaning. We will talk
about this in detail
below.
22 physics from finance

Imagine, for example, that a trader sells three tomatoes at

€1 each and then uses this money to buy six apples at €0.50
each.

The end result of such a process is completely unchanged if

the government decides to print lots of euro notes such that
the value of each euro drops by a factor of ten. Afterwards,
the trader gets €10 for each tomato but then needs to pay €5
per apple. So again, our trader starts with three tomatoes
and ends up with six apples.

4
At least in our toy
model changing the
value of the currency
has no effect, since This is the case because fiat money has no fixed absolute
we imagine that, as a value. Thus, we can rescale a given currency without any
result, all prices and
wages are automatically physical effect.4
adjusted. Of course,
in the real world there In the previous section we learned that whenever we find
could be psychological
effects since people get a transformation which leaves a given object unchanged,
used to certain prices.
 finance intuitively 23

5
The freedom to rescale
we call the transformation a symmetry. In this sense, we currencies is an ex-
ample of what we
say that rescaling a given currency is a symmetry of the call a gauge symme-
financial market.5 try in physics. If you
want to learn more
about gauge symme-
At first glance, this may seem like an unimportant detail. tries, try Demystifying
But we will see below that this symmetry is an extremely Gauge Symmetry by J.
Schwichtenberg avail-
useful tool when we want to describe the financial market able at https://arxiv.
mathematically.6 org/abs/1901.10420.

6
To spoil the surprise:
Now, what does this symmetry have to do with the internal whenever we want
space of the financial market?7 to write down laws
describing a system, we
know that they must
respect its symmetries.
In particular, when
1.3 Internal Space we want to describe
the financial market
consistently, we need
to take the rescaling
For simplicity, let’s focus on one specific commodity, say, symmetry into account.
copper. We can imagine that all possible prices of copper The rescaling of a
local currency is not
live in an internal space that we call the money space. Each allowed to have any
point in this space corresponds to one specific price. effect on the dynamics
of the system. This is
a powerful constraint
which will allow us
to derive the correct
equations.

7
Recall that one of
our main goals in
this chapter is to
understand what an
internal space is.

But, as we’ve discussed in the last section, it doesn’t mat-

ter where exactly we are in money space. We can move
the price of copper around freely. The price could be any
number between €0.0 . . . 01 and €∞.8 To take all these pos- 8
Here €0.0 . . . 01 de-
sibilities into account simultaneously, we imagine that there notes an arbitrarily
small but non-zero
positive number.
24 physics from finance

is an abstract line which represents our money space. Each

point on this line corresponds to a specific price for the
commodity in the (local) currency. The symmetries of our
system are shifts from one point to another.

Since different countries possibly use different currencies

and have different prices, it’s instructive to imagine that
9
Take note that each there is such a line above each country.9
line here is, in principle,
infinitely long. But, of
course, we can’t draw
that.

10
Don’t worry if this
construction still seems
horribly abstract.
Below we will discuss
concretely how this
picture of the financial
market helps us to
describe it.
Moreover, take note
that this construction is
completely analogous
to what we do in Together, all these lines with all possible locations (here all
modern physics. Later
we will discuss how countries) yield the arena which we use to describe what is
and why in physics going on in our toy model.10
we also attach internal
spaces to each point of
spacetime.
 finance intuitively 25

Internal spaces are not only important to help us keep track

of what is happening but because they can also directly
influence the dynamics. To understand this, we need to
make our toy model only slightly more realistic.

1.4 Connections

We now need to talk about what happens when countries

use different currencies. When we allow that countries use
different currencies, it still has to be true that prices alone
do not have any meaning. A large number for the price of
copper in one country (e.g. 10 Indian rupees per kilogram)
and a tiny number in another country (e.g. 0.5 euros per
kilogram) does not necessarily mean that copper is cheaper
in the latter country.11 11
At the time of writ-
ing, January 2019, 10
Indian rupees are worth
Whether the price in a given country is cheap or expensive approximately 0.12
depends on the exchange rates of the local currency. If I euros.

only tell you that copper trades in Japan for 100 yen per
kilogram, it’s impossible for you to decide if it makes sense
to ship your copper to Japan and sell it there. But if I also
tell you that 100 yen are worth 0.8 euros, you know immedi-
ately that this is a good price compared to, say, 0.5 euros per
kilogram in Germany.12 12
As of January 2019,
1 yen is worth 0.008
euros.
For our geometrical picture, this means that when countries
use different currencies we not only need a line above each
country but also something which glues them together.
Our exchange rates are the glue which allow us to calculate
how much a given currency is worth in terms of another
currency.
26 physics from finance

1.5 Gauge Invariant Quantities

All this may seem rather boring or even trivial. But be as-
sured that thoughts like this are really at the heart of mod-
ern physics. To understand how this comes about, we need
to talk about dynamics.

To describe what is going on in a given system, we need an

equation of motion. An important idea is that our equation
cannot depend on anything which changes if a currency
gets rescaled. In particular, this means that our equation
of motion cannot depend on the prices of copper directly.
These prices depend on the local currency and therefore
change as soon as a country rescales its currency.
 finance intuitively 27

This is already a powerful insight.13 Naively, we may have 13

In fact, we will use
thought that any equation describing our system certainly it below to derive one
of the most important
involves the prices of copper. But we just learned that any equations.
such equation wouldn’t be useful because prices do not
have any direct meaning.

But what does then have an effect on the dynamics?

Well, opportunities to make money certainly have an effect.

If such an opportunity shows up somewhere, traders will
use it to earn money.

One way to make money is to buy copper in one country

and sell it in a different country at a higher price. Spotting
these kinds of opportunities is not so simple when countries
use different currencies.

In order to be sure that an opportunity is lucrative, we need

to be able to calculate whether the value of the currency we
will receive when we sell the copper is greater than that of
the currency we used to buy the copper. Only when what
we finally receive is more valuable than what we started
with, do we really earn any money.

For example, if we start by buying copper for 10 euros and

end by selling it for 10000 yen, this does not tell us anything
unless we calculate how many euros we get for our yen.14 14
As of January 2019,
10000 yen is worth 80.2
euros. So this would be
A crucial point is that such an opportunity to earn money a good deal.
does not depend on local conventions. For example, if Japan
decides to drop a zero from their currency, you would only
get 1000 yen for your copper. But you could still trade these
for 80.2 euros, because the euro-yen exchange rate is ad-
justed as soon as Japan decides to modify its currency.

In physics, we call a quantity which does not depend on

28 physics from finance

local conventions gauge invariant. In our example, the

amount of money we can earn through such a copper trade
is gauge invariant.

This is an important observation because if we want to

write down an equation describing the system, the gain
factor which quantifies the opportunity would be a great
15
We will define the candidate to appear in it.15
gain factor later in
mathematical terms.
What other quantities should appear in our equation of
motion? Maybe the exchange rates?

Again, this cannot be the case because exchange rates do

depend on local conventions. We can, however, construct a
quantity closely related to the exchange rates which does
not depend on local conventions and therefore possibly
plays a role in our equation of motion.

All opportunities to make money are crucial for the dynam-

ics within our system. So far, we have only talked about
how we can earn money by trading copper. It turns out that
sometimes it’s possible to earn money purely by trading one
16
In general, arbitrage currency for another. This is known as currency arbitrage.16
describes an opportu-
nity to make a risk-less
profit.
For example, let’s imagine the exchange rates are as follows:

DM/P = 1
P/F = 2
F/L = 10
DM/L = 10 , (1.1)

where DM denotes Deutsche Mark, F Francs, P Pounds and

L Lira.

A trader is able to earn money by exchanging currencies.

 finance intuitively 29

If he starts with 1DM, he can trade it for 1P, then use the
pound to trade it for 2F, then trade these for 20L and finally
trade these back for 2DM.

17
We are assuming
the financial market
works perfectly. Every
time a country decides
to change its local
currency, the exchange
rates get adjusted
immediately and
perfectly. But then, it’s
natural to wonder: if
everything works so
Take note that again it’s crucial that we end up with the perfectly, where do
arbitrage opportunities
same currency that we started with because we can’t other- come from? This is
wise be sure that we really earned money. Moreover, take something we will talk
note that such an arbitrage possibility is a real feature of the about below. At this
point, our exchange
system and does not depend on local conventions. rates are completely
passive and their only
job is to keep track
We can check this explicitly. First of all, our exchange rates of local conventions.
do change when, for example, Italy decides to drop a zero Later we will introduce
from their currency L → L̃ = 10L.17 As a result of this new dynamical actors
(banks) which adjust
the exchange rates
actively.
30 physics from finance

change in Italy, our exchange rates change as follows

DM/P = 1
P/F = 2
F/L = 10 → F/ L̃ = F/(10L) = 1
DM/L = 10 → DM/ L̃ = DM/(10L) = 1 . (1.2)
While the exchange rates are different, the amount of money
our trader earns is unchanged by such a rescaling!

If he starts again with 1DM, he can still trade it for 1P, then
for 2F, then trade these for 2 L̃ and finally trade these for
2DM. The final result is the same as before.

What we’ve therefore discovered is a second method to earn

money. The gain factor which quantifies how much money
 finance intuitively 31

we can earn does not depend on local conventions and

therefore quite likely plays a dominant role in our equation
of motion.18 18
We will define the
corresponding gain
factor in mathematical
terms below.

1.6 Curvature

Before we move on, let’s return for a moment to our geo-

metrical picture. Above, I’ve mentioned that the exchange
rates are the glue which ties together the lines representing
the local money spaces in the various countries. Our total
internal space consists of all these lines glued together. An
extremely useful interpretation is that arbitrage opportu-
nities correspond geometrically to a non-zero curvature of
this internal space.

Intuitively, whenever we move in a loop and end up with a

different state than the one we started in, we say the space
we move in is curved.19 19
This is discussed
in more detail in
Appendix D.5.
32 physics from finance

For the moment, simply imagine that if there is an arbi-

trage possibility, our local internal spaces (lines) are glued
together non-trivially. This is a useful picture because when-
ever a trader performs a financial transaction, he not only
moves through spacetime but also through our internal
space. His behavior is directly influenced by the curvature
of the space, analogous to how a marble rolls differently on
a curved surface.

Before we move on and discuss things like this in detail,

there is one more feature of the toy model that we need to
discuss.

1.7 Dynamical Curvature

In a real financial market there are, of course, not only

traders but also other actors. Most importantly, there are
banks which determine the exchange rates and allow
traders to exchange one currency for another. But these
banks are not purely sitting in the background passively.
Instead, they are dynamically shaping what is going on in
the system. In particular, they adjust the exchange rates all
the time. They do this, for example, to get rid of arbitrage
opportunities because otherwise they would quickly run out
of money.

This means that there is a continuous interplay between

the two kinds of actors. The behavior of traders is directly
influenced by the pattern of arbitrage opportunities within
the system. In turn, the banks modify the exchange rates all
We will make this
20
the time depending on the behavior of the traders.20
more concrete below.

This idea allows us to get a first glimpse at what all this has
 finance intuitively 33

to do with physics.

First of all, in physics, we follow exactly the same steps to

find quantities which do not depend on local conventions.
With these quantities at hand, writing down the equations
governing the dynamics of the system is only a small addi- 21
In mathematical
tional step. In addition, as we will discuss later, the gain fac- terms, the gain factors
describing currency
tors we discussed in the context of the financial toy model arbitrage opportunities
are mathematically completely analogous to the quantities represent the curvature
of our internal space.
used to describe fundamental physics!21 Moreover, the in- In physics, we call the
terplay between traders and exchange rates is completely curvature (at least if it’s
analogous to the interplay between elementary particles and a dynamical actor) a
gauge field.
force fields.22
22
Examples of force
Before we start talking about physics, it is extremely in- fields are the gravi-
tational field and the
structive to rephrase everything that we have just learned in electromagnetic field.
mathematical terms. This allows us to understand many of
the most important concepts of modern physics and to write
down one of the most important equations.
2

Finance Mathematically

First of all, a short warning: It’s easily possible that you will get
bored or confused by reading the following sections. But please
don’t let that discourage you from reading the rest of the book.
When you feel lost, simply skim through the remaining sections
and then move on to the next chapter. You will be able to un-
derstand the rest of the book even if you don’t internalize all the
mathematical details in this chapter. Moreover, you can always
come back later and catch up on the details.

In addition, take note that this section is inspired by the excellent

essay titled "The symmetry and simplicity of the laws of physics
and the Higgs boson" by Juan Maldacena.1 Therefore, if you don’t 1
The essay is available
understand one of the explanations below, you can consult this at https://arxiv.org/
abs/1410.6753.
essay to read a second perspective.

With this in mind, let’s dive in.

Mathematically, we imagine that our countries live on a

two-dimensional lattice. Each point on the lattice is labelled
36 physics from finance

by 2-numbers: ~n = (n1 , n2 ).

In other words, each country can be identified by a vector ~n

which points to its location.

We can move from one country to a neighboring country by

using a basis vector ~ei , where i denotes the direction we are
moving in. For example, ~e1 = (1, 0).
 finance mathematically 37

We denote the exchange rates between the country labelled

by the vector ~n and its neighbor in the i-direction by Ri (~n).
For example, if the country at the location labelled by ~n
uses Deutsche Marks and its neighbor in the 1-direction
uses Francs, R1 (~n) tells us how many Francs we get for each
Deutsche Mark.

Analogously, R2 (~n) describes the exchange rate between the

country at ~n and the country at ~n +~e2 :
38 physics from finance

And R1 (~n + ~e2 ) describes the exchange rate between the

country at ~n +~e2 and the country at ~n +~e2 +~e1 :

2
We will see below
that this makes our
formulas much more
simple. In particular,
the products in the
transformation rules
can be replaced by
sums of the logarithms.
In mathematical terms
what we do here is
to work with the Lie
algebra instead of with
the corresponding Lie
group. While a Lie
group is, in general,
a complicated object,
the corresponding Lie
algebra is always a
simple vector space. Lie In physics, we usually introduce the corresponding loga-
algebras are discussed
in Appendix C.3. rithms of the exchange rates2

3
The natural logarithm Ri (~n) ≡ e Ai (~n) , (2.1)
is the inverse of the
exponential function.  
This means, where Ai (~n) ≡ ln Ri (~n) .3
eln( x) = x .
The next ingredient that we need is a notation for gauge
transformations. In our toy model a gauge transformation
is a rescaling of a local currency and directly impacts, for
example, the exchange rates. We use the notation f (~n) to
denote a rescaling of the currency in the country at ~n by a
factor f .
 finance mathematically 39

In addition, we again introduce the corresponding loga-

rithm
f (~n) ≡ ee(~n) . (2.2)

In general, when we perform such a gauge transformation,

f (~n), in the country labelled by ~n and also another gauge
transformation in the neighboring country in the i-direction,
f (~n +~ei ), the corresponding exchange rate gets modified by
the ratio of these two factors:

f (~n +~ei )
Ri (~n) → Ri (~n) . (2.3)
f (~n)
40 physics from finance

In terms of the logarithms this equation reads

Eq. 2.1 A (~n) f (~n +~ei )

Ri (~n) = e i → Ri (~n)
f (~n)

y
Eq. 2.2
ee(~n+~ei )
= e(~n) e Ai (~n)
e
eA
= e A− B

y
eB
= e Ai (~n)+e(~n+~ei )−e(~n) (2.4)

and we can conclude that

Ai (~n) → Ai (~n) + e(~n +~ei ) − e(~n) . (2.5)

We learned above that an important aspect of the system is

whether arbitrage opportunities exist. An arbitrage possibil-
ity exists when we can trade currencies in such a way that
we end up with more money than we started with. But we
can only make such a statement when the initial currency
and the final currency are the same. Only then can we be
certain whether or not the final amount of money is larger
than the amount we started with. Therefore, we need to
trade money in a loop.

Defining the gain factor as the ratio of the final amount of

money to the initial amount of money (in the same local
currency), the total gain we can make by following a specific
4
We will discuss why loop can be quantified by4
this formula is correct
below. 1 1
G (~n) = Ri (~n) R j (~n +~ei ) . (2.6)
Ri (~n +~e j ) R j (~n)

When this gain factor is larger than one, we can earn money
by trading money following the loop. If it is smaller than
one, we lose money.

To understand Eq. 2.6 for the gain factor in the case of a

loop process, imagine that we start with 1DM. We trade
P
it for Pounds and R1 (~n) = 1 DM tells us that we get in
 finance mathematically 41

total 1P. Afterwards, we trade our Pounds for Francs and

R2 (~n +~e1 ) = 2 PF tells us that we get in total 2F. Afterwards,
1 F
we trade our Francs for Lira. R1 (~n +~e2 ) = 10 L tells us that
we get 10 Lira for each Franc. Hence, we have to calculate
L
2F/R1 (~n +~e2 ) = 20L. Finally, we use that R2 (~n) = 10 DM
tells us that we need to pay 10L for each Deutsche Mark and
therefore calculate 20L/R2 (~n) = 2DM.5 The gain factor in 5
The crucial point
this case is: is that our exchange
rates Ri (~n) always tell
final amount 2DM us how much of the
= = 2, (2.7) currency in the neigh-
initial amount 1DM
boring country in the
which is what is obtained by substituting values for the i-direction we get for
each unit of the local
exchange rates in Eq. 2.6. currency in the country
at ~n. Hence, we some-
times have to divide by
the corresponding ex-
change rate to calculate
the resulting amount of
a new currency.

Once more we introduce the corresponding logarithm

G (~n) ≡ e Fij (~n) (2.8)

and again, we can rewrite our equation (Eq. 2.6) in terms of 6

To spoil the surprise:
the logarithms6 in physics Fij is directly
related to components
of the magnetic field.
For example, F12 =
− B3 .
42 physics from finance

1 1
G (~n) = Ri (~n) R j (~n +~ei )
Ri (~n +~e j ) R j (~n)

y
Eq. 2.8, Eq. 2.1
Fij (~n) Ai (~n) A j (~n+~ei ) 1 1
e =e e
e Ai (~n+~e j ) e A j (~n)
eA
= e A− B

y
eB
= e Ai (~n)+ A j (~n+~ei )− Ai (~n+~e j )− A j (~n) . (2.9)

We can read off here that

7
A single (logarithm of
an) exchange rate Ai (~n)
Fij (~n) = A j (~n +~ei ) − A j (~n) − [ Ai (~n +~e j ) − Ai (~n)] . (2.10)
is gauge dependent
and can therefore be A crucial consistency check is that G and Fij are unchanged
set to zero simply by by gauge transformations. An arbitrage possibility is some-
changing local money
coordinate systems. thing real and thus cannot depend on local choices of the
money internal system. Quantities like this are usually
called gauge invariant. So in words, G and Fij encode what
is physical in the structure of exchange rates.7 Moreover,
an important technical observation is that Fij (~n) is antisym-
metric: Fij (~n) = − Fji (~n), which follows directly from the
8
Antisymmetry means definition (Eq. 2.9).8
that we get a minus
sign when we exchange
the indices i ↔ j. Using So far, we have only talked about spatial exchange rates.
Eq. 2.9, we can calculate However, there are also temporal exchange rates, i.e. interest
Fji (~n) = rates. A clever trick to incorporate this is to introduce time
Ai (~n +~e j ) − Ai (~n) as the zeroth-coordinate like we do in special relativity.9 In
− [ A j (~n +~ei ) − A j (~n)] other words, in addition to specific locations (countries) our

= − A j (~n +~ei ) − A j (~n) lattice contains copies of these locations at different points


 in time.
− [ Ai (~n +~e j ) − Ai (~n)]
= − Fij .

9
We will talk about
special relativity in
Section 5.1.
 finance mathematically 43

This means a point on this spacetime lattice is specified by

2 + 1 coordinates ~n = (n0 , n1 , n2 ) and the zeroth component
indicates the point in time.

Then, Eq. 2.9 reads10 10

Once more, let me
spoil the surprise:
the non-vanishing
Fµν (~n) = Aν (~n +~eµ ) − Aν (~n) − [ Aµ (~n +~eν ) − Aµ (~n)] , (2.11) components of Fµν (~n)
with either µ = 0 or
where previously ij ∈ {1, 2} and now µ, ν ∈ {0, 1, 2}. ν = 0 are directly
related to what we
call the electric field in
physics. For example,
F01 = E1 . An important
difference is that in
physics, we’re typically
dealing with three
An important step to make our model more realistic is spatial dimensions and
hence the indices µ, ν
to allow not just one possible location per country, but in
run from 0 to 3 and the
principle, infinitely many. Graphically, we can imagine that corresponding spatial
we add more and more points to our lattice. These points indices, i, j run from 1
to 3.
represent possible locations of our traders.

Each possible location corresponds to a vector ~x = a~n,

where ~n is a vector with integer components, ~n = (ni , n j ) T .
These integers get larger and larger in value, at any specific
location, as a gets smaller and smaller. This is because we fit
in more and more points (locations) within a given space in-
terval. However, the length of a vector ~x remains fixed and
finite in size. Usually, a is known as the lattice spacing. The
limit, in which there is a continuum of possible locations, is
known as the continuum limit.
44 physics from finance

With this notation at hand, we can specify what the contin-

uum limit means mathematically. In the continuum limit,
the distance between neighboring lattice points goes to zero
11
Take note that math- a → 0 while ~x has to remain a fixed finite vector.11 More-
ematically this implies
over, since our bookkeepers Ai (~n) connect lattice points on
that ~n goes to infinity.
But this is nothing you our discrete lattice, we need to replace them with book-
should worry about keepers which connect points in our continuous space
because the only thing
we care about after Ai (~n) → a Ãi (~x ). As we will see in a moment, the factor
we’ve taken the limit is a makes sure that the values of our bookkeepers stay fixed,
a finite ~x. as we take the limit a → 0.

So, specifically, the transformation in Eq. 2.5

Ai (~n) → Ai0 (~n) ≡ Ai (~n) + e(~n +~ei ) − e(~n) (2.12)

now reads

a Ãi (~x ) → a Ãi0 (~x ) ≡ a Ãi (~x ) + e(~x + a~ei ) − e(~x )

a
y =1
a
e(~x + a~ei ) − e(~x )
= a Ãi (~x ) + a
a
y

e(~x + a~ei ) − e(~x )

 
=a Ãi (~x ) + . (2.13)
a

In the limit a → 0, the second term on the right-hand side

12
This is how the yields a difference quotient12
derivative of a function
is defined in general. e(~x + a~ei ) − e(~x ) ∂e
lim ≡ . (2.14)
a →0 a ∂xi
13
We can now see And by looking at Eq. 2.13 we can therefore conclude that13
why we introduced
the additional factor ∂e
a in the definition of
Ãi (~x ) → Ãi0 (~x ) ≡ Ãi (~x ) + . (2.15)
∂xi
Ãi . This factor makes
sure that a drops out Analogously, the gain factor in Eq. 2.9,
completely from our
equation and we get a Fij (~n) = A j (~n +~ei ) − A j (~n) − [ Ai (~n +~e j ) − Ai (~n)] , (2.16)
quantity which remains
finite in the limit a → 0.
 finance mathematically 45

becomes

Fij (~x ) = a à j (~x + a~ei ) − a à j (~x ) − [ a Ãi (~x + a~e j ) − a Ãi (~x )]
a

y
=1
a
à j (~x + a~ei ) − à j (~x ) Ãi (~x + a~e j ) − Ãi (~x )
= a2 − a2
a a
(2.17)

and we can conclude that we have to replace Fij → a2 F̃ij to

get a quantity with a fixed value in the limit a → 0

à j (~x + a~ei ) − à j (~x ) Ãi (~x + a~e j ) − Ãi (~x )

a2 F̃ij (~x ) = a2 − a2
a a
a2

y

à j (~x + a~ei ) − à j (~x ) Ãi (~x + a~e j ) − Ãi (~x )
F̃ij (~x ) = −
a a
(2.18)

On the right-hand side, we again have difference quotients

in the continuum limit

à j (~x + a~ei ) − à j (~x ) ∂

lim = Ã
a →0 a ∂xi j
Ãi (~x + a~e j ) − Ãi (~x ) ∂
lim = Ã . (2.19)
a →0 a ∂x j i

We therefore find that

∂ ∂
F̃ij (~x ) = Ã − Ã . (2.20)
∂xi j ∂x j i

Moving forward, we will drop the tilde above our symbols

to unclutter the notation. Moreover, if we include temporal
exchange rates, we have to replace our indices i, j with µ, ν
which each run from 0 to 2.

To summarize, in the continuum limit Eq. 2.5 becomes

∂e
Aµ (t, ~x ) → Aµ (t, ~x ) + (2.21)
∂xµ
46 physics from finance

and Eq. 2.11 becomes

∂Aν ∂Aµ
Fµν (t, ~x ) ≡ − . (2.22)
∂xµ ∂xν

In the previous chapter we noticed that we can not only

earn money by trading currencies, but also by trading goods
like copper. Depending on the local prices it can be lucrative
to buy copper in one country, bring it to another country,
sell it there, and then go back to the original country to
compare the final amount of money with the amount of
money we started with.

From the definition of the gain factor as the ratio of the final
amount of money to the initial amount of money (in the
same local currency), the gain factor for such a process is
given by
p(~n +~ei )
gi (~n) = , (2.23)
p(~n) Ri (~n)
where p(~n) denotes the price of copper in the country lo-
cated at ~n. To understand the definition in Eq. 2.23, imagine
that we start with 10DM and the price for one kilogram of
copper in Germany is p(~n) = 10DM. This means that we
can buy exactly 1 kilogram of copper. Then we can go to
the neighboring country and sell our copper for, say, 30F
since p(~n +~e1 ) = 30F. Afterward, we can go back to Ger-
many and exchange our 30F for 15DM since, say, Ri (~n) = 2.
Therefore, we will have made 5DM in total; and so the gain
factor in this case is:
final amount 15DM
= = 1.5 . (2.24)
initial amount 10DM
This is exactly what is obtained by substituting values for
the prices and exchange rate in Eq. 2.23. Again, a gain factor
 finance mathematically 47

larger than one means that we earn money and a gain factor
smaller than one means that we lose money.14 14
If you are unsure
which quantity goes
in the numerator
and which in the
denominator, ask
yourself: would it
increase our profit if
the given quantity is
larger? We write in
the numerator if the
answer is yes. If the
answer is no, we write
in the denominator. For
example, a higher price
Once more, we introduce the corresponding logarithms of copper in France
certainly increases
gi (~n) ≡ e Hi (~n) our profit. Hence
p(~n + ~ei ) is written
p(~n) ≡ e ϕ(~n) (2.25)
in the numerator.
Similarly, a higher price
and Eq. (2.23) then reads in terms of the corresponding of copper in Germany
logarithms would lower our profit
and therefore, we write
p(~n +~ei ) in the denominator.
gi (~n) =
p(~n) Ri (~n)
y

Eq. 2.25, Eq. 2.1

e ϕ(~n+~ei )
e Hi (~n) = ϕ(~n) A (~n)
e e i
y

= e ϕ(~n+~ei )− ϕ(~n)− Ai (~n) , (2.26)

which implies

Hi (~n) = ϕ(~n +~ei ) − ϕ(~n) − Ai (~n) . (2.27)

The gain factor gi (~n) and its logarithm Hi (~n) quantify di-
rectly how lucrative copper trade is for a trader based at
location ~n and doing business with the country at ~n +~ei . We
can imagine that the more lucrative a trade is, the more will
traders make use of it to earn money.

This motivates us to define the copper current

 
Ji (~n) ≡ qHi (~n) = q ϕ(~n +~ei ) − ϕ(~n) − Ai (~n) , (2.28)
48 physics from finance

where q is a proportionality constant that quantifies how

strongly the traders in a given system react to an opportu-
nity to earn money by trading copper.

Completely analogously to what we did above, we can

generalize this formula for situations that involve time by
replacing i ∈ {1, 2} with µ ∈ {0, 1, 2}):
 
Jµ (~n) = q ϕ(~n +~eµ ) − ϕ(~n) − Aµ (~n) . (2.29)

Moreover, in the continuum limit Eq. 2.29 becomes

 ∂ 
Jµ (t, ~x ) = q ϕ(t, ~x ) − Aµ (t, ~x ) . (2.30)
∂xµ

Let’s talk about the three quantities Jµ (~n) (with µ ∈ {0, 1, 2})
in more detail.

The gain factor Hi (~n) tells us how lucrative a trade is in

which copper is transported from the country at ~n to the
neighboring country at ~n +~ei . Hence, Ji (~n) is a measure of
the amount of copper that flows between the two countries.
The trade related to H0 (~n) does not involve the exchange of
copper between neighboring countries. Instead, H0 (~n) tells
us how much we can earn by buying copper and selling it at
a later point in time in the same country. Therefore, J0 (~n) is
a measure of the amount of copper in the country at ~n.

This is an important idea since, so far, we had nothing in

our description that contained any information about the
amount of copper at a certain location or about how copper
flows through the system. Copper is represented in our de-
scription in somewhat abstract terms by its price. However,
the amount of copper in a given country is not proportional
 finance mathematically 49

to the local price of copper. The local price depends on the

local money coordinate system and therefore cannot directly
represent a measurable quantity like the amount of copper.
Instead, the amount of copper in a given country must be
represented by a gauge independent quantity like J0 (~n). The
same is true for the flow of copper Ji (~n).

Maybe a second perspective helps to understand this better.

The main idea is that the distribution and flow of copper
is somehow encoded in the pattern of prices in the various
countries at different times. The prices themselves contain
no information about the flow and distribution of copper
since they depend on local conventions. Instead, everything
we need to know is encoded in how prices change as we
move from country to country or to a different point in
time.

For example, if the price in a given country goes down

rapidly over time, this signals that they have "too much" 15
As discussed above,
copper. In other words, a rapidly sinking price is an indi- we can’t compare the
prices directly but need
cator that lots of copper is located in the corresponding to take into account that
country. The gain factor H0 (~n) describes accurately how the price differences can be
the result of changes in
price in the country located at ~n changes over time.15 Using the currency. For this
these ideas, we defined J0 (~n) ≡ qH0 (~n) which describes reason, the definition
how much copper is located in the country at ~n. The pro- of H0 (~n) contains
an additional term
portionality constant q is necessary to describe how given − A0 (~n) which adjusts
amounts of copper (e.g., 3 kilograms or 2 cubic meters) are the price differences
appropriately.
related to specific changes of the price H0 (~n).

Similarly, if the price in one country is much higher than in

a neighboring country, we can assume that they desperately
need cooper and hence that there is net influx of copper.
Therefore, a price difference in neighboring countries is an
indicator that copper flows from the low-price country to
the high-price country. The price difference between the
country at ~n and its neighbor at ~n + ~ei is described accu-
50 physics from finance

16
Once more, we rately by Hi (~n).16 Therefore, we use Ji (~n) ≡ qHi (~n) to
need to take into
describe the flow of copper from ~n to ~n +~ei .
account that price
differences can be the
result of different local
conventions. That’s
why the definition
of Hi (~n) contains
an additional factor
− Ai (~n). Now that we’ve introduced a mathematical notation to
describe our toy model, we can also write down specific
rules which describe how the system behaves. We need
two kinds of rules. On the one hand, we need rules which
describe how arbitrage opportunities show up and evolve
over time. On the other hand, we need rules which describe
how our traders behave and in particular, how they react
to the presence of arbitrage opportunities. The following
diagram summarizes what we will discover in the following
sections.

inhomogeneous Maxwell equation

&
Traders
b Arbitrage Opportunities
diffusion equation with covariant derivative

We will start with the rules describing arbitrage opportuni-

ties (the inhomogeneous Maxwell equation) and then move
on to derive the rules which describe how traders react to
them (the diffusion equation with covariant derivative).
 finance mathematically 51

2.1 Arbitrage Dynamics

Of course, it’s possible to consider various kinds of dy-

namics within our financial toy model. Different kinds of
dynamics require different kinds of laws. However, in the
following we discuss a very particular set of rules which
correspond to what are known as Maxwell’s equations. So,
as mentioned in the preface, our goal is not to derive equa- 17
Additionally, take
tions which describe financial markets accurately, but to use note that no telepor-
tation of copper is
the financial market to understand concepts like Maxwell’s allowed.
equations.

To derive these, we start with the crucial assumption that

copper is conserved, i.e. no copper is destroyed or pro- 18
We only consider
duced. In particular, this means that whenever the amount elementary copper
trade loops. In these
of copper decreases in a given country it must have gone elementary loops
somewhere. Conversely, whenever the amount of copper copper always flows
"from left to right".
increases in a country it must have come from somewhere.17 For example, Ji (~n) is
proportional to the
In mathematical terms this means:18 amount of copper
that flows from ~n to
~n +~ei . A positive Ji (~n)
change of copper in country ~n = total net flow therefore represents
an outflow of copper.
which we can write in mathematical terms as Analogously, the flow
! involved in the process
2 2 which yields Ji (~n −~ei )
J0 (~n +~e0 ) − J0 (~n) = − ∑ Ji (~n) − ∑ Ji (~n −~ei ) also goes from left
i =1 i =1 to right. However,
a positive Ji (~n − ~ei )
where we sum over all neighboring countries. We can also means a net inflow.
write this as Therefore, we need a
! relative minus sign.
2 2 If we have a positive
J0 (~n +~e0 ) − J0 (~n) + ∑ Ji (~n) − ∑ Ji (~n −~ei ) = 0 . (2.31) quantity on the left-
i =1 i =1 hand side the amount
of copper increases and
and we can visualize this equation as follows: this must correspond to
a net inflow.
52 physics from finance

In the continuum limit, Eq. 2.31 becomes the famous conti-

19
Here and in the nuity equation19
following, we use
Einstein’s summation 2
convention, i.e. there is ∂0 J0 + ∑ ∂i Ji = 0
an implicit sum over all i =1
y

indices that appear in

pairs. This allows us to 2
write ∑2µ=0 ∂µ Jµ as ∂µ Jµ . ∑ ∂µ Jµ = 0
µ =0
y

summation convention
∂µ Jµ = 0 . (2.32)

Now, what we really want is an equation that lets us un-

derstand how arbitrage opportunities show up and evolve
dynamically. Moreover, we already know that Jµ and Fµν
are good quantities to describe our system since these do
not depend on local conventions. The quantities Jµ contain
information about the positions and flow of copper, while
Fµν represent arbitrage opportunities.

However, the first naive guess to write Fµν on one side and
 finance mathematically 53

20
This is an important
general result. Every
Jµ on the other side of an equation fails because Jµ has only
time we have a sum
one index but Fµν has two indices. In addition, there is one over something sym-
further piece of information that we can use: Eq. 2.32. metric in its indices
multiplied with some-
thing antisymmetric in
If taking the derivative of the right-hand side of an equation the same indices, the
result is zero:
yields zero, the left-hand side has to be zero, too. There is,
however, no reason why ∂µ Fµν should be zero and this is ∑ aij bij = 0
ij
another hint that our first naive guess is wrong.
if aij = − a ji and
bij = b ji holds for all
The crucial observation is that i, j. We can see this by
writing

∂ν ∂µ Fµν = 0 , (2.33) ∑ aij bij =

ij

1 
since Fµν is antisymmetric but partial derivatives commute ∑
2 ij
aij bij + ∑ aij bij
ij
∂µ ∂ν = ∂ν ∂µ .20 This suggests that we should try
We are free to rename
our indices i → j and
∂ν Fµν = µ0 Jµ , (2.34) j → i, which we use in
the second term

where we have introduced the proportionality constant ∴ ∑ aij bij =

ij
µ0 which encodes how strongly the pattern of arbitrage 1 
opportunities reacts to the presence and flow of copper. ∑
2 ij
aij bij + ∑ a ji b ji
ij

Then we use the sym-

This equation has exactly one free index (µ) on both sides metry of bij and an-
and, most importantly, both sides yield zero if we calculate tisymmetry of aij , to
switch the indices in
the derivative the second term, which
yields
∂µ ∂ν Fµν = µ0 ∂µ Jµ ∴ ∑ aij bij =
ij
y

Eq. 2.32 and Eq. 2.33

0=0 X 1
2 ∑
aij bij
ij

Eq. 2.34 is the famous inhomogeneous Maxwell equation. + ∑ a ji b ji
ij |{z} |{z}
=− aij =bij

For our discrete system it reads 1

2 ∑
= aij bij
ij
2 2

− ∑ aij bij
∑ Fµν (~n) − ∑ Fµν (~n −~eν ) = µ0 Jµ (~n) . (2.35) ij
ν =0 ν =0
=0
54 physics from finance

Moreover, Eq. 2.35 with µ = 0 reads

2 2
∑ F0ν (~n) − ∑ F0ν (~n −~eν ) = µ0 J0 (~n)
ν =0 ν =0

y
F00 (~n) = 0
2 2
∑ F0i (~n) − ∑ F0i (~n −~ei ) = µ0 J0 (~n) , (2.36)
i =1 i =1
21
An object is antisym- where we used that F00 (~n) = 0, since Fµν is antisymmetric.21
metric if we get a minus Here, we have µ0 J0 (~n) on the right-hand side, which is
sign when we swap its
indices: Fµν = − Fνµ . proportional to the amount of copper located at ~n. Thus, for
This automatically µ = 0, Eq. 2.34 gives us information about the pattern of
implies that all compo-
nents for which the two exchange rates around a country in which copper is present.
indices are equal will In the continuum limit, Eq. 2.36 becomes
vanish. :
∂i F0i = µ0 J0 , (2.37)
F00 = F11 = F22 = 0
because the antisymme- which is known as Gauss’s law.
try implies F00 = − F00 ,
F11 = − F11 and
F22 = − F22 . This can Similarly, for µ → i ∈ {1, 2} in Eq. 2.35, we get equations
only be true if F00 = 0, that give us information about the pattern of exchange rates
F11 = 0 and F22 = 0.
that are present whenever copper flows22
2 2
∑ Fiν (~n) − ∑ Fiν (~n −~eν ) = µ0 Ji (~n) . (2.38)
ν =0 ν =0
22
Above we investi- In the continuum limit, this equation becomes23
gated the temporal
component of the equa- ∂0 Fi0 + ∂ j Fij = µ0 Ji , (2.39)
tion µ = 0. Now we
investigate the spatial which is known as the Ampere-Maxwell law.
part µ → i ∈ {1, 2}.
23
Again, Einstein’s
summation convention
is used. Here this
means that there is
an implicit sum over
the index j because it In addition to the inhomogeneous Maxwell equation
appears twice in the (Eq. 2.34), there is the homogeneous Maxwell equation:
second term.
∂λ Fµν + ∂µ Fνλ + ∂ν Fλµ = 0 . (2.40)

This equations follows automatically from the definition of

the gain factor Fµν in terms of the exchange rates Aµ .
 finance mathematically 55

∂λ Fµν + ∂µ Fνλ + ∂ν Fλµ

y
Eq. 2.22
= ∂λ (∂µ Aν − ∂ν Aµ ) + ∂µ (∂ν Aλ − ∂λ Aν ) + ∂ν (∂λ Aµ − ∂µ Aλ )

y
= ∂λ ∂µ Aν − ∂λ ∂ν Aµ + ∂µ ∂ν Aλ − ∂µ ∂λ Aν + ∂ν ∂λ Aµ − ∂ν ∂µ Aλ

y
∂µ ∂ν = ∂ν ∂µ
= ∂λ ∂µ Aν − ∂λ ∂ν Aµ + ∂µ ∂ν Aλ − ∂λ ∂µ Aν + ∂λ ∂ν Aµ − ∂µ ∂ν Aλ

y
=X ν − µ + λ − ∂λX µ −
∂λX AX
∂µX X
∂λX A
∂ν ∂µ A
∂ν X ∂ A +X∂λX A
∂ν ∂µ A
∂ν
 XX µXν
X   XX λ

y
=0 X

Now we move on and try to write down rules which de-

scribe how our traders behave.
56 physics from finance

2.2 Trader Dynamics

Before we try to understand how traders react to arbitrage

opportunities, we need to understand how they behave
24
In physical terms, when there are none.24
we say that we first
want to write down
rules describing free But what do we actually know about traders? Of course,
traders, before we try if there is a risk-less opportunity to make money, they will
to describe interacting
traders. certainly use it. This is the easy part which we want to
understand in a second step. But what do they do if there
are no such obvious opportunities?

The best answer I can give you is: we don’t know. The indi-
vidual psychology of each trader is far too complex and it is
simply impossible to know all possible factors which influ-
ence their decisions. But this doesn’t mean that we can’t de-
scribe how traders move through our system. In fact, many
of the important advances in modern physics are methods
25
For example, max- to describe systems without knowing all the details.25 And
imum entropy as an after all, it’s one of the main goals of economists to produce
optimal guessing strat-
egy but also the renor- theories of human behavior that gloss over the details of
malization group and individual psychology.
quantum mechanics.

On a "microscopic" level there are an incredibly large num-

ber of factors influencing the decisions of each trader all the
time.

The situation is therefore analogous to what happens to

pollen grains in water. These grains constantly collide with
water molecules, but there is no way we can describe all
the water molecules influencing the pollen grains. If we
ignore this microscopic level and zoom out, our pollen
grains simply perform random walks.

Therefore, our working assumption will be that traders

 finance mathematically 57
26
Maybe you’re dis-
appointed that we
aren’t trying something
move randomly. more ambitious. But
be assured that we
are steadily moving
Of course, if we zoom in we can deduce how traders react towards the edge of
to individual signals (Behavioral Psychology) analogous to modern physics. One
of the main realizations
how we can deduce how pollen grains collide with individ-
in modern physics was
ual water molecules. that the laws describing
nature at a more fun-
damental level do not
This doesn’t really help us because there are too many help you to describe
factors influencing trader decisions and too many water nature if you zoom out.
molecules colliding with our pollen grains. From a macro- Quantum mechanics
does not help us to de-
scopic perspective, the random walk hypothesis is really the scribe how a ball rolls
best we can do. down a ramp because a
ball consists of far too
many quantum parti-
What we therefore need, is an equation describing our ran- cles and there is no way
dom walking traders.26 Since we are not making any special to take the quantum
rules governing their
assumptions, the equation we are looking for is exactly the individual behavior into
same equation describing pollen grains in water. account. Analogously,
there are too many
actors and factors in an
This equation which describes random moving objects is economy. Therefore,
known as the diffusion equation and reads27 the rules governing the
behavior of individual
actors do not help us to
∂ρ(t, x ) ∂2 ρ(t, x ) describe the economy
=D , (2.41)
∂t ∂x2 as a whole.

where ρ is the probability density of the quantity we want 27

For simplicity, we
to describe and D is a constant known as diffusion con- restrict ourselves to one
spatial dimension. In
stant.28 three spatial dimen-
sions, the diffusion
equation reads
Previously, we argued that J0 (t, x ) tells us how much copper
∂ρ(t, ~x )
is located at a particular location.29 But now, we are assum- = D ∇2 ρ(t, ~x ) ,
∂t
ing that our traders move copper around randomly. This
where ∇2 is the Lapla-
means that we need to describe our model in probabilistic cian operator.
terms.
28
We will talk about
these two new notions
To understand what this means, let’s imagine that there below.
exists just one unit of copper. Our goal is to describe how
this unit of copper is moved around. In this case, we need to
29
We discussed this in
the text below Eq. 2.29.
58 physics from finance

describe it using a probability density.

For example, we can imagine that at one point in time t0 ,

the function ρ(t, x ) is exactly one for one particular location
x0 :
ρ(t0 , x0 ) = 1 and zero otherwise. (2.42)
In other words, this means that we are 100% certain to find
our unit of copper at t0 at the location x0 .

At a later point in time t1 , the probability density possibly

looks like this:

ρ(t1 , x0 ) = 0.1
ρ(t1 , x1 ) = 0.2
ρ(t1 , x2 ) = 0.7
ρ ( t1 , x ) = 0 for all other locations x (2.43)

This means that we are 70% certain to find the unit of cop-
per at the location x2 , 20% certain to find it at x1 and only
10% certain to find it at x0 . Formulated differently, if we
prepare the system equally many times, we expect to find
our unit of copper at this point in time at x2 in 70% of cases,
at x1 in 20% of cases and at x0 in only 10% of cases. More-
over, we don’t expect to find it anywhere else. Similarly, if
there are multiple units of copper in the system, ρ(t, x ) tells
us how we expect them to be distributed throughout the
system.

To summarize: J0 previously told us how much copper is

located at a particular location. Now, we use ρ to describe
how much copper we can expect to find at a particular
location if we repeat our experiment multiple times. This
change to a probabilistic perspective is necessary because
we assume that our traders move copper around randomly.
While we therefore can’t be certain what will happen, we
can make probabilistic predictions.
 finance mathematically 59

Now, how do we get from the random-walk hypothesis to

Eq. 2.41?30 30
If you don’t care
about how we get
from the random walk
As before, let’s focus on one unit of copper. At each time hypothesis to the
step ∆t the unit’s location x changes by a step l: diffusion equation, you
can skip the rest of this
x (t + ∆t) = x (t) + l (t) . (2.44) subsection.

Our task is to write down an equation which allows us

to calculate the probability density ρ( x, t + ∆t) given the
probability density at a previous time step ρ( x 0 , t). Then,
we can start with a specific density at one point in time (for 31
In more technical
example one in which all copper is located at one particular terms: given initial
conditions, we can
point) and calculate how copper gets distributed through then calculate how the
the system.31 If a unit of copper moves from x 0 at time t to x system evolves as time
passes.
at time t + ∆t, it must have moved a step l = x − x 0 . What’s
the probability that it moves such a step?

To answer this question, we need to introduce the proba-

bility distribution of our random walk.32 We denote the If you’re unfamiliar
32

probability distribution which tells us the probability for with probability distri-
butions, have a look at
each possible step as χ(l ). In particular, χ(ls ) tells us the Appendix A.1.
probability that our unit of copper moves a specific step ls .

The probability to find our unit of copper at x at time t + ∆t

is therefore33
Z ∞
ρ( x, t + ∆t) = ρ( x 0 , t)χ( x − x 0 )dx 0
−∞
new variable l ≡ x − x 0
y

Z ∞
= ρ( x − l, t)χ(l )dl , (2.45)
−∞
where we sum over all possible previous locations x 0 . In Take note that the
33

change of variable
words, this formula means that we sum over all possibilities brings in a minus sign,
how our unit of copper can end up at x at t + ∆t. In prin- but that this minus
sign gets canceled by
ciple, we can reach x from any location x 0 . But we need to another minus sign that
take into account that the step from x 0 to x only happens results from taking into
with a certain probability, as indicated by χ( x − x 0 ). More- account the limits of
the integral, which get
over, the probability density that indicates how likely is to reversed for the new
variable.
60 physics from finance

find our unit of copper at x 0 at t is ρ( x 0 , t). Hence, the prob-

ability to find the unit of copper at x at t + ∆t is given by
the sum over all possible locations x 0 at t weighted by the
probability density values that tell us how likely it is that it
really was at x 0 and that the step from x 0 to x happens.

Now we need to make some (rather mild) assumptions

about χ(l ) in order to move forward. While we do not need
to specify the probability distribution χ(l ) itself, we need
to make an assumption about the mean and standard de-
34
The notions of mean viation of the steps l.34 From the mean and the standard
and standard devia- deviation, we can tell how far from its original position we
tion are discussed in
Appendix A. can expect to find each given unit of copper. But since we
do not want to single out any particular direction, we as-
sume the probability for steps in each direction are equal.
This implies that the mean of l is zero because steps to the
35
We sum over the left happen on average equally often as steps to the right:35
probability χ(l ) for
Z ∞
each step l.
lχ(l )dl = 0 . (2.46)
−∞

But, of course, we do not assume that copper doesn’t get

moved around at all. Instead a mean of zero only implies
that steps in each direction happen equally likely and there-
fore average out. This also means that the standard devia-
tion Z ∞
l 2 χ(l )dl = a (2.47)
−∞
 finance mathematically 61

is non-zero.

The standard deviation encodes information about the total

distance each unit of copper gets moved around on average
at each step.36 The constant a is some constant which de- 36
As mentioned above
pends on the system in question and usually needs to be already, the meaning
of the standard devi-
determined experimentally. In other words, a is simply the ation is discussed in
name we give to the standard deviation. In addition, χ(l ) Appendix A.
has to be normalized properly37 37
This is not an as-
Z ∞ sumption but a consis-
χ(l )dl = 1 . (2.48) tency requirement.
−∞

This requirement ensures that the probabilities for all pos-

1
sible steps sum up to 100% = 100 100 = 1. With these 38
We can imagine that
formulas at hand, we can rewrite Eq. 2.45. The crucial idea our trades happen
between neighboring
is that we observe the system from far away and the stepsize individual locations
is therefore small compared to the scale on which ρ varies while we are interested
in the general behavior
noticeably.38 of the global market.

This allows us to Taylor expand ρ( x − l, t):39 39

The idea behind the
Z ∞ Taylor expansion is ex-
plained in Appendix E.
ρ( x, t + ∆t) = ρ( x − l, t)χ(l )dl
−∞
y

Taylor expansion
Z ∞ 
∂ρ 1 2 ∂2 ρ 
≈ ρ( x, t) − l + l χ(l )dl
−∞ ∂x 2 ∂x2
y

Z ∞
∂ρ ∞
Z
≈ ρ( x, t) χ(l )dl − lχ(l )dl
| −∞ {z }
∂x −∞
| {z }
=1 (Eq. 2.48) =0 (Eq. 2.46)

1 ∂2 ρ ∞
Z
+ l 2 χ(l )dl
2 ∂x2 −∞
| {z }
= a (Eq. 2.47)
y

a ∂2 ρ
≈ ρ( x, t) + . (2.49)
2 ∂x2
Moreover, we assume that the trades happen much quicker
than the timescale at which we observe ρ. This allows us to
62 physics from finance

Taylor expand ρ( x, t + ∆t) on the left-hand side

∂ρ
ρ( x, t + ∆t) ≈ ρ( x, t) + ∆t . (2.50)
∂t
Using Eq. 2.49 and Eq. 2.50 we therefore find

∂ρ a ∂2 ρ
ρ( x, t) + ∆t ≈ ρ( x, t) +
∂t 2 ∂x2

y
ρ( x,
  t)
∂ρ a ∂2 ρ
∆t ≈
∂t 2 ∂x2

y
∂ρ a ∂2 ρ
≈
∂t 2∆t ∂x2 a

y
D≡
2∆t
∂ρ ∂2 ρ
≈D 2, (2.51)
∂t ∂x
where in the last step, we introduced the diffusion constant
a
D ≡ 2∆t . This is exactly the diffusion equation (Eq. 2.41) we
wanted to derive.

We have therefore shown that the diffusion equation allows

us to predict in probabilistic terms how copper gets moved
around if our traders act randomly.

Now we can move on and think about how we can incor-

porate into our equation how traders react to arbitrage
opportunities.

2.2.1 Trader Dynamics and Arbitrage

40
So far the diffusion
equation only describes An arbitrage opportunity is a risk-less way to make money.
how copper is moved
through the system by So whenever such an opportunity shows up in the market
traders in the absence it’s clear that this will cause a drift among our traders.
of arbitrage opportuni-
ties
How we can incorporate this effect into the diffusion equa-
tion?40
 finance mathematically 63

The answer to this question is really simple and beautiful.

In fact, the idea we will use in the following is regarded by
many physicists as one of the most fundamental insights in
modern physics.

The main observation is that as soon as there are arbitrage

opportunities, we need to be more careful whenever we 41
Here we compare the
calculate derivatives. When we calculate the derivative value of the function
f ( x ) at the location x
of a function, we compare the values of the function at with its value at the
two different locations. This follows if we recall the usual location x + h.
definition of a derivative in terms of a difference quotient41
∂ f (x) f ( x + h) − f ( x )
≡ lim . (2.52)
∂x h →0 h
But whenever there is an arbitrage opportunity, there is nec-
essarily a non-trivial exchange rate Aµ somewhere.42 There- 42
This follows because
fore, comparing the values of a function at two different arbitrage opportuni-
ties are described by
locations is non-trivial. We need to take the exchange rate (Eq. 2.22):
Aµ between the two locations into account. Otherwise we ∂Aν ∂Aµ
Fµν (t, ~x ) ≡ − .
are comparing apples with oranges (e.g. dollars to euros). ∂x µ ∂x ν
Therefore, an arbitrage
opportunity exists if
Luckily, we already know how to do this. The idea we need
Fµν 6= 0 and this implies
is exactly the same one that we used already in Eq. 2.27. Aµ 6= 0 for at least one
Our goal there was to compare the prices of copper at two of the components µ.

different locations and we found

 
Jµ (~n) = q ϕ(~n +~eµ ) − ϕ(~n) − Aµ (~n) , (2.53)

which in the continuum limit becomes

 
Jµ (~x ) = q ∂µ ϕ(~x ) − Aµ (~x ) , (2.54)

because the first two terms yield a difference quotient.

This is how, in general, the derivative of a function looks

in the presence of non-trivial exchange rates Aµ (~x )! It is
conventional to call this the covariant derivative

Dµ ≡ ∂µ − Aµ (~x ) , (2.55)
64 physics from finance
43
The adjustment hap-
pens like this and not
by multiplying the
function with the ex- In words, this means that we need to adjust the values of
change rate because we the function using the appropriate exchange rate.43
are working, as before,
with the corresponding
logarithms. This is already everything we need to know.

44
In physics, this To get the correct equation describing our traders in the
procedure is known as presence of non-trivial exchange rates, we simply have
minimal coupling.
to replace the ordinary derivative ∂µ with the covariant
45
Reminder: the ordi- derivative Dµ .44
nary one-dimensional
diffusion equation
reads (Eq. 2.41) Concretely, the diffusion equation in the presence of arbi-
∂t J0 (t, x ) = D∂2x J0 (t, x ) . trage opportunities reads45

Dt J0 (t, x ) = DDx2 J0 (t, x ) ,

y
Eq. 2.55
(∂t − At (t, x )) J0 (t, x ) = D (∂ x − A x (t, x ))2 J0 (t, x ) (2.56)

where D is still the diffusion constant and Dx and Dt denote

46
In physical terms, the covariant derivatives.46
the new terms in the
diffusion equation
describe that there is a That’s it. To describe how traders behave when arbitrage
force which pushes our opportunities exist, we simply have to consistently take non-
traders in a particular
direction. trivial exchange rates into account. We do this by acknowl-
edging that whenever we calculate the derivative of some
function, we need to include the appropriate exchange rate.
This is necessary because when we calculate a derivative, we
are evaluating the function at two different locations. This
leads us to the conclusion that we need to replace our ordi-
nary derivatives with covariant derivatives ∂µ → Dµ , where
Dµ = ∂µ − Aµ . If we do this, we automatically end up with
the correct equations.

There are two important lessons that we learned in this

chapter. Firstly, we saw why gauge symmetry is useful. By
 finance mathematically 65

thinking about gauge symmetry we learn which quantities

are independent of local conventions. Secondly, we derived
the correct equation describing the interplay between the
flow of goods like copper and arbitrage opportunities by
using the ideas that we should describe the system with
gauge invariant quantities, that copper is conserved and that
traders move randomly in the absence of risk-less opportu-
nities to make money.

With this in mind, we are ready to understand how nature

works at the fundamental level.
66 physics from finance
 Part II
The Physics of Nature

"When learning about the laws of physics you find that there are a large
number of complicated and detailed laws, laws of gravitation, of electricity
and magnetism, nuclear interactions, and so on, but across the variety of
these detailed laws there sweep great general principles which all the laws
seem to follow."

Richard P. Feynman
 finance mathematically 69

Many of the most important discoveries in modern physics

can be translated into the language of geometry and this
allows us to understand them from a common perspective.

Specifically, one of the main discoveries of the last century

was that our arena in physics consists of spacetime and
three internal spaces known as charge space, isospin space
and color space. This is important because there is a direct
connection between each of these spaces and a specific
fundamental interaction.

So far, four fundamental interactions have been discov-

ered:47 47
Sometimes the notion
of fundamental forces
is used instead of fun-
damental interactions.
B At large (cosmological) scales, gravity is the most impor- But this can lead to
tant of the four interactions. Gravity is responsible for the misunderstandings
because the effects
movement of planets and galaxies. caused by fundamental
interactions can be
B At tiny scales, strong and weak interactions are responsi- quite different from
what we know from
ble for most of the interesting phenomena. For example, macroscopic forces. For
protons are held together by the strong force and nuclei example, weak interac-
decay through weak interactions. tions lead to the decay
of nuclei and it is quite
difficult to bring this
B In between these two extremes it’s electromagnetic inter- into agreement with the
actions which hold sway. macroscopic notion of a
force.

While gravity is directly connected to spacetime, weak in-

teractions are connected to isospin space, strong interactions
to color space and electromagnetic interactions to charge
space.

To understand all this, it makes sense to rip the total arena

apart and study its individual constituents. In particular,
we will discuss properties of the spacetime part and of the
internal parts of the arena separately.
70 physics from finance

48
This is analogous to
what we discussed for In addition, before we discuss global properties of the arena,
the financial market
previously. The inter- it makes sense to study how it looks like locally. This is
nal space above one a useful approach because, while the arena as a whole is
country is simply a
line. But globally these extremely complicated, locally its structure is quite simple.48
lines can be glued to-
gether non-trivially and
therefore, in general,
our arena as a whole is Arena of Physics
quite complicated.
internal internal internal external

s v  (
Color Space Isospin Space Charge Space Spacetime

locally locally locally locally

   
SU(3) SU(2) U(1) R1,3

   
Color
O Isospin Electric Charge Energy-Momentum
O O O

   
Strong Interactions Weak Interactions Electromagnetic Interactions Gravity

Understanding all this may sound like a lot of work. But

the good news is that everything that we will discuss in the
following sections is completely analogous to what we dis-
cussed in the previous chapters. In other words, if you’ve
understood the main ideas discussed in the previous chap-
ters, you are perfectly prepared to understand how nature
works at the fundamental level. Moreover, the general dis-
cussion for each of the individual parts is extremely similar;
only the details are different.

But before we discuss the properties of the arena, we should

talk about the objects we want to describe.
3

Elementary Particles

During the last century, physicists discovered a whole zoo

of elementary particles. For all we presently know, these
particles cannot be further decomposed and have no sub-
structure. That’s why we call them elementary particles.

These elementary particles are the building blocks that all

of the more complex structures (atoms, molecules, proteins)
we see around us consist of. For example, a hydrogen atom
consists of a proton and an electron.
72 physics from finance

But the proton is not an elementary particle. In fact, it con-

1
To be a bit more sists of quarks.1
precise: a proton
consists of two up
quarks and one down
quark.

2
Take note that there
are lots of additional
elementary parti-
cles which cannot be
discovered so easily
(muons µ, tauons τ,
charm quarks, strange
quarks, top quarks,
bottom quarks). They Quarks and electrons are, as far as we know, elementary
are heavy cousins of
the electron and the particles.2
quarks that make up a
proton (up quark and
down quark). Since The particle zoo itself is not that interesting. Instead, what is
they are heavier, they interesting is the way the particles interact with each other.
decay into their lighter
cousins shortly after
they are produced. In the toy model discussed in the previous chapter, we
They were discovered wanted to describe traders who carry copper around. Now
in collider experiments
in which high-energetic our goal is to describe elementary particles which carry
particles collide. When what we call charges around.3
two protons collide at
higher energies, they
sometimes create heav- We can distinguish between different types of elementary
ier particles. (We will particles because they carry different charges. If two ele-
understand roughly
how this works below.) mentary particles carry the same charges, we say that they
In this sense, colliders are the same type of elementary particle, e.g., electrons.
are essentially large mi-
croscopes which allow
us to look deeper at To understand what we mean by charges, we need to recall
matter. The higher the one main idea. One of the deepest insights during the last
energy of the colliding
particles, the deeper we century was that the arena of physics is not a static back-
can look. ground structure. Instead, the arena itself is a dynamical
actor which changes all the time. Most importantly, the
3
Take note that charge
does not necessarily arena is directly influenced by the behavior of all elemen-
mean electric charge.
 elementary particles 73

tary particles and, in turn, these particles are influenced by

the structure of the arena.

influence
#
Elementary Particles Arena of Physics
c
influences
4
This is analogous to
Formulated a bit more precisely, the way elementary par- what we discussed in
the previous chapter.
ticles influence each other is by modifying the geometry of There our traders
the arena. are influenced by the
geometry of the internal
money space and, in
For example, if two elementary particles influence each turn, the geometry
other electromagnetically, this happens because they change of the money space
changes dynamically
the structure of the internal charge space and at the same depending on the
time feel the structure of this charge space.4 behavior of the traders.
5
We can imagine that a
But the crucial point is that elementary particles influence trader who carries a lot
of copper around has a
and get influenced by different parts of the arena. This is much bigger effect than
what we use to label elementary particles. a trader who carries
only a small amount.

In other words, we identify elementary particles based on 6

There is a beautiful
which part(s) of the total arena they modify and how strong connection between
these labels and the
their effect is on the corresponding geometries.5 symmetries of internal
spaces. This connection
is known as Noether’s
Concretely:6 theorem and we will
talk a bit more about
it in Section 5.3. More-
B We say that particles which modify charge space carry over, take note that
there is another in-
electric charge.
credibly important
B We say that particles which modify isospin space carry label called spin. While
energy-momentum
isospin. is the Noether charge
following from invari-
B We say that particles which modify color space carry
ance under spacetime
color. (Take note that there is a reasonable explanation translations, spin is the
for the somewhat strange label "color" but it has nothing Noether charge follow-
ing from invariance
under rotations.
74 physics from finance

to do with the colors we perceive in everyday life. We

will talk more about this when we discuss color space
below. )
B We say that particles which modify spacetime carry
energy-momentum. However, it is conventional to use
the label "mass" for elementary particles because energy-
momentum depends on the standpoint (reference frame)
from which we observe the particle. (Mass is directly re-
lated to the "length" of the energy-momentum vector; and
it is the quantity that is invariant (same in all reference
frames). We will discuss this further in Section 5.3.)

With this in mind, here is a table listing a few of the elemen-

tary particles whose interactions we want to describe:

electric charge isospin color mass

left-chiral electron-neutrino νe 0 1/2 0 ???
left-chiral electron e L −e -1/2 0 me
right-chiral electron e R −e 0 0 me
red left-chiral up quark urL 2/3 e 1/2 ( 12 , √1 ) mu
2 3
blue left-chiral up quark ubL 2/3 e 1/2 −1 )
(0, √ mu
3
g
green left-chiral up quark u L 2/3 e 1/2 ( −21 , √1 ) mu
2 3
red left-chiral down quark drL -1/3 e -1/2 ( 12 , √1 ) md
2 3
blue left-chiral down quark dbL -1/3 e -1/2 (0, √ −1 ) md
3
g
green left-chiral down quark dL -1/3 e -1/2 ( −21 , √1 ) md
2 3
red right-chiral up quark urR 2/3 e 0 ( 12 , √1 ) mu
2 3
blue right-chiral up quark ubR 2/3 e 0 (0, √ −1 ) mu
3
g
green right-chiral up quark u R 2/3 e 0 ( −21 , √1 ) mu
2 3
red right-chiral down quark drR -1/3 e 0 ( 12 , √1 ) md
2 3
blue right-chiral down quark dbR -1/3 e 0 (0, √ −1 ) md
3
g
green right-chiral down quark dR -1/3 e 0 ( −21 , √1 ) md
2 3

Here e = 1.602 × 10−19 C is the magnitude of the electric

 elementary particles 75

7
You can find them, for
charge of an electron and me = 9.109 × 10−31 kg is its mass. example, on Wikipedia.
Moreover, take note
The masses of neutrinos have not been measured and right- that the labels "left-
chiral neutrinos have not been discovered yet. chiral" and "right-
chiral" refer to another
important charge
The remaining values for the masses are not listed because particles can carry
they are not important for our purpose.7 In fact, the specific known as chirality.
But unfortunately,
values of the charges of the various elementary particles are discussing chirality
not important for the modest goal of this book. would lead us too
far astray. For our
purposes, it is sufficient
The only message to take away is that there are different to realize that the
particles which carry different charges. right-chiral electron
is a different particle
since it carries different
Take note that there are other elementary particles. There charges. "Right-chiral
electron" is simply
are two additional generations which consist of heavy the name we give to
cousins of the particles listed here. In other words, these this particle. We will
additional generations of elementary particles have exactly talk about the closely
related concept of spin
the same charges, except for a larger mass. For example, in Section 5.3. You
the muon has exactly the same charges as the electron but can find a complete
discussion in
a larger mass. Moreover, all the particles listed in the table,
Jakob Schwichten-
together with their heavier cousins of the other two gener-
berg. Physics from Sym-
ations, share a common property, as a result of which they metry. Springer, Cham,
are named, collectively, fermions. This property is that no Switzerland, 2018b.
ISBN 978-3319666303
two identical fermions can locally be found in the same
quantum state. 8
Formulated more
technically, the par-
ticles listed in the
In addition, other particles, collectively named bosons, share table all carry spin
a different property, in that an indefinite number of them 1/2. (A fermion is
defined as a particle
can be found in the same quantum state. Among these are with half-integer spin
the Higgs boson, and so-called gauge bosons called pho- 1/2, 3/2, 5/2 etc.) Par-
tons, W-bosons, Z-Bosons, gluons (and maybe gravitons). ticles with spin 1 are
known as gauge bosons
But the role of these particles in nature is a different one. and they appear as
While fermions are associated with matter, gauge bosons are quantized excitations
of our bookkeepers Aµ ,
associated with the fundamental interactions.8 Wµ , Gµ which we will
talk about below. (A
In case you’re wondering: no one knows why exactly these boson is defined as a
particle with integer
elementary particles exist and why they carry the labels spin 0, 1, 2 etc.) In quan-
tum field theory, we
therefore say that these
gauge bosons mediate
their corresponding
interaction.
76 physics from finance

they do. This is one of the big open questions in modern

physics. We only know that these particles exist because we
can observe them in experiments.

No one predicted their existence on theoretical grounds. For

example, when the muon (a heavy cousin of the electron)
was discovered, Nobel laureate I. I. Rabi famously quipped,
"Who ordered that?"

Nevertheless, we can successfully describe the behavior and

interactions of the various members of this particle zoo.
We will do this again in two steps. First, we will determine
how our particles behave when they are alone and only
afterwards will we discuss how particles interact with each
9
PS: If you want to other. 9
learn more about color
charge, try http://
jakobschwichtenberg.
com/
what-are-quantum-numbers/.
4

Quantum Mechanics

In Section 2.2, we discussed how we can describe the move- 1

Recall that this is a
ment of copper through our system even though we can’t reasonable assumption
as long as there are no
take all factors influencing trader decisions into account. arbitrage opportunities.
The main idea was that if we don’t know all these details, If there are risk-less
opportunities to earn
we have to assume that copper gets moved around ran- money, it’s clear that
domly.1 our traders will use
them to make money.
But we only included
Using this assumption we were able to derive the so-called this in a second step
diffusion equation (Eq. 2.41) once we have under-
stood how traders
∂ρ(t, ~x ) behave when there are
= D ∇2 ρ(t, ~x ) . (4.1) no such opportunities.
∂t
Now we want to describe elementary particles.

Luckily, the situation is not so different. It was one of the

biggest discoveries in modern physics that there seems to
be a fundamental randomness in the behavior of elementary
particles.

The origin and meaning of this randomness is something

78 physics from finance

that physicists have now argued about for more than a

2
Just for the record: century and is not what I want to dive into here.2
deriving the laws of
quantum mechanics
from the assumption The equation at the heart of quantum mechanics is known
that elementary parti- the Schrödinger equation and reads
cles perform random
walks is known as the
∂ψ(t, ~x )
stochastic interpre- = iD ∇2 ψ(t, ~x ) . (4.2)
tation. One way to ∂t
explain the randomness
is by noting that in The only difference between Schrödinger’s equation and the
quantum field theory diffusion equation is the imaginary unit i on the right-hand
the vacuum (ground
state) is not trivial. In- side. In other words, Schrödinger’s equation is a diffusion
stead, for each quantum equation with an imaginary diffusion constant.
field there are vacuum
fluctuations which
cannot be turned off. This small change has huge implications.
In other words, even
in the state with the
lowest possible energy, For example, while solutions of the diffusion equation
quantum fields still ρ(t, x ) are real, solutions of the Schrödinger equation ψ(t, ~x )
wiggle a bit around.
are, in general, complex functions. This follows automati-
Another popular way
of looking at it is to say cally because, to get the same expression on both sides of
that virtual particles the equation, we must get an imaginary unit from some-
continuously pop in ∂ψ(t,~x )
and out of existence. where. Specifically, on the left-hand side, ∂t must yield
With this picture in something times i.
mind, we can imagine
that our elementary
particles scatter con- We can interpret any real function in a straightforward man-
tinuously with these ner, but for a complex function the interpretation is less di-
vacuum fluctuations/
virtual particles. In this rect. It is clear that the result of any physical measurement
sense, the situation is is a real number; together with appropriate units. Thus, the
somewhat analogous to
complex function ψ(t, ~x ) itself cannot be directly related to
pollen grains in water
which collide with the a real physical quantity; its relationship to things we can
water molecules all the measure has therefore to be somewhat indirect.
time.

In contrast, we are able to interpret ρ(t, ~x ) directly as the

probability density which tells us how much copper we can
expect to find at a particular location.

So what’s the meaning of ψ(t, ~x )?

 quantum mechanics 79

While ψ(t, ~x ) is, in general, complex, it’s quite easy to use it

to construct a function which is certainly always real:3 3
A complex number
multiplied by its com-
ρ(t, ~x ) = ψ? (t, ~x )ψ(t, ~x ) . (4.3) plex conjugate always
yields a real number. To
see this, we note that
The superscript star symbol, ?, on a function denotes com-
we can write a complex
plex conjugation; that is, it denotes the same initial function number as a sum of its
except for the imaginary unit i replaced everywhere by −i it real part plus i times its
imaginary part
appears in the initial function. The function ρ(t, ~x ) is inter-
z = a + ib
preted as the probability density in quantum mechanics,
analogous to how we interpreted ρ(t, ~x ) in the financial toy where a is the real part
of z, b is its imaginary
model. part and i is the imag-
inary unit which is
defined by its prop-
We can understand how this interpretation comes about
erty i2 = −1. Using
by noting that there is also a continuity equation in quan- this notation we can
tum mechanics which can be derived directly from the calculate

Schrödinger equation:4 z? z =
( a + ib)? ( a + ib) =
D  
∂t (ψ? ψ) = ∂i ψ∂i ψ? − ψ? ∂i ψ . (4.4) ( a − ib)( a + ib) =
i
a2 + iab − iab − i2 b2 =
This equation has exactly the same structure as the continu-
a2 + b2
ity equation (Eq. 2.32). Any equation of this form is called a
(Take note how the
continuity equation. We get an equation of this form when- imaginary unit i drops
ever there is a conserved quantity. out and we end up with
a purely real number.)

We therefore write it as 4
We will derive this
explicitly below.
∂t ρ = ∂i ji . (4.5)

This is why we interpret ρ ≡ ψ? ψ as the probability den-

sity.5 5
The quantity
D 
ji ≡ ψ∂i ψ? − ψ? ∂i ψ
To make the connection to our previous continuity equation i

even more explicit, we rewrite Eq. 4.5 as is known as the proba-

bility current.
∂0 j0 = ∂i ji .

Now, where does Eq. 4.4 come from?6 6

As usual, if you’re
not interested in the
derivation, feel free
to skip the following
paragraph.
80 physics from finance

One puzzle piece is to multiply the Schrödinger equation

(Eq. 4.2) with the complex conjugated function ψ?

∂ψ(t, ~x )
ψ? (t, ~x ) = iDψ? (t, ~x )∇2 ψ(t, ~x ) . (4.6)
∂t
The second puzzle piece we need is the complex conjugate
of the Schrödinger equation (Eq. 4.2)

∂ψ? (t, ~x )
= −iD ∇2 ψ? (t, ~x ) , (4.7)
∂t
which we multiply with ψ:

∂ψ? (t, ~x )
ψ(t, ~x ) = −iDψ(t, ~x )∇2 ψ? (t, ~x ) . (4.8)
∂t
7
To unclutter the The sum of Eq. 4.6 and Eq. 4.8 reads7
notation, from now on
we write ψ instead of ∂ψ ∂ψ?
ψ(t, ~x ) etc. ψ? +ψ = iDψ? ∇2 ψ − iDψ∇2 ψ? . (4.9)
∂t ∂t
8
Hint: product rule. We can write this more compactly by noting that8

∂ψ ∂ψ? ∂
ψ? +ψ = (ψ? ψ) . (4.10)
∂t ∂t ∂t
In addition, we note that
 
ψ ? ∇2 ψ − ψ ∇2 ψ ? = ∇ ψ ? ∇ ψ − ψ ∇ ψ ? (4.11)

Therefore, we can write Eq. 4.9 as

∂ ?  
(ψ ψ) = iD ∇ ψ? ∇ψ − ψ∇ψ? . (4.12)
∂t
9
Take note that This is exactly the continuity equation (Eq. 4.4).9
1 −i2 i
= = −i = −i
i i i

In summary, we have learned that the fundamental equation

of quantum mechanics is extremely similar to the diffusion
equation. The only difference between the two is that there
 quantum mechanics 81

is an imaginary unit i in the Schrödinger equation. It is re-

ally this little difference which makes quantum mechanics
a bit more difficult than our financial toy model. In particu-
lar, we can’t interpret solutions of the Schrödinger equation
directly. Instead, we end up with the interpretation that the
absolute square of the solutions yield probability densities.

Maybe you wonder why macroscopic objects behave so

regularly if the elementary particles they consist of move
around randomly?

This is possible because, while individual elementary par-

ticles seem to move around randomly, we can predict how
they behave on average. Otherwise we couldn’t write down
equations like the diffusion equation and Schrödinger’s
equation.

Moreover, take note that macroscopic objects consist of

an incredibly large number of elementary particles. The
random behavior of the individual constituents therefore
averages out and this allows us to describe macroscopic
objects using simple laws like Newton’s second law.10 10
In quantum mechan-
ics, this can be shown
explicitly. This result is
Of course, there is a lot more to quantum mechanics than known as the Ehrenfest
just the Schrödinger equation and its solutions. But, dis- theorem.
cussing quantum mechanics in detail requires at least a
few hundred pages.11 For the modest goal of this book we If you want to learn
11

already have everything we need to move forward. more about quantum

mechanics, try

Jakob Schwichten-
berg. No-Nonsense
Quantum Mechan-
4.1 Charge Space ics. No-Nonsense
Books, 2018c. ISBN
978-1719838719

One of the big surprises in modern physics was that the

internal part of the arena of physics is much bigger and
82 physics from finance

more complex than anyone would have expected. While

charge space is relatively simple and was discovered more
than a century ago, isospin space and color space are more
complicated and were discovered much later.

But the good news is that we can understand all of the

essential features by focusing on the internal charge space.
Once we’ve discussed charge space thoroughly, we can
understand the structure of the more complicated color and
isospin spaces almost immediately.

To understand charge space, we need to recall a few things

which we have discussed already in the context of our fi-
nancial toy model.

In this toy model, we discovered an internal space "above"

each possible location. Each such internal space is simply a
line going from 0.0 . . . 01 to ∞. Each point on this line rep-
resents a possible price of copper. The total internal space
is given by all these lines taken together. We also discussed
that when countries use independent local currencies, we
need something which glues the various lines together. In
our toy model, the exchange rates were the glue. Moreover,
we argued that there is a symmetry because each country
can rescale its currency freely. This is possible because as
soon as a country rescales its currency, the exchange rates
are adjusted immediately and there is no effect on the dy-
namics of the system. But we also discovered that the struc-
ture of the internal space can have a direct impact on the
dynamics. This happens whenever there are imperfections
in the exchange rates such that there are currency arbitrage
opportunities. Whenever such a risk-free opportunity to
earn money exists, traders will use it to earn money. There-
fore, the dynamics of the system is crucially shaped by such
arbitrage opportunities.
 quantum mechanics 83

Now, in order to understand charge space and thus how el-

ementary particles influence each other via electromagnetic 12
Don’t worry if you
interactions, we only need to repeat the same story using don’t know any quan-
tum mechanics. The
different actors and a slightly different internal space. following discussion
is aimed at people
without previous
Let’s start with the internal space. The existence of the inter- knowledge of quan-
nal space we call charge space follows directly from one of tum mechanics. Of
the basic facts of quantum mechanics.12 course, we can’t discuss
quantum mechanics
properly here. But
In quantum mechanics, we describe what is going on us- if you want to learn
more about quantum
ing a complex function ψ(t, ~x ), called the wave function. mechanics, you might
However, everything we can measure is a real number. This enjoy reading
means that the relationship between quantities, O, we can Jakob Schwichten-
measure and the wave function ψ is always of the form berg. No-Nonsense
Quantum Mechan-
O ' ψ? ψ. A direct consequence of this observation is that
ics. No-Nonsense
we have some freedom in our wave functions, analogous to Books, 2018c. ISBN
the freedom we discovered in the price of copper. In par- 978-1719838719

ticular, we always have the freedom to multiply our wave

functions by a phase factor

ψ → ψ̃ = eie ψ , (4.13)

since

O ' ψ? ψ ⇒ ψ̃? ψ̃ = (eie ψ)? eie ψ

(eix )? = e−ix
y

−ie ? ie
=e ψ e ψ
y

= ψ? e−ie eie ψ
e−ie eie = 1
y

?
=ψ ψ=O X (4.14)

What we’ve discovered here is completely analogous to

what we discussed for the financial toy model. In the finan-
cial toy model, we used the prices to describe copper and
discovered that we have the freedom to rescale the prices by
multiplying them by a real number (Eq. 2.2)

p → ee p . (4.15)
84 physics from finance

In quantum mechanics, we describe particles, say an elec-

tron, using a wave function and have the freedom to multi-
ply it by a phase factor (Eq. 4.13). The only difference is the
little i in the exponential function.

This observation allows us to understand what the internal

space of quantum mechanics is. In the financial toy model,
the internal space was a line ranging from 0.0 . . . 01 to ∞
which represents all possible prices of copper. We can imag-
ine that we determine the structure of this space by starting
with one specific price and then applying all allowed sym-
metry transformations. This yields the space of all allowed
values.

To get the internal space of quantum mechanics, we do

exactly the same thing. We imagine that someone hands us
13
A wave function is a a specific wave function Ψ:13
complex function which
can always be written ψ(t, ~x ) = R(t, ~x )eiϕ(t,~x) , (4.16)
in this form.
where R(t, ~x ) is a real function and eiϕ(t,~x) is a specific phase
factor. We can then determine the structure of the corre-
 quantum mechanics 85

sponding internal space by applying all allowed symmetry

transformations (Eq. 4.13). A crucial observation is now that
this does not yield a line ranging from 0.0 . . . 01 to ∞ but a
circle! Something of the form eiϕ(t,~x) evaluated at a specific
location ~x and at a specific moment in time t is a complex
number of unit length since

|eiϕ(t,~x) |2 = (eiϕ(t,~x) )? eiϕ(t,~x) = e−iϕ(t,~x) eiϕ(t,~x) = 1 . (4.17)

Geometrically all complex numbers of length one lie on a

circle with radius one in the complex plane.

We can understand this using Euler’s formula

z = eiφ
= cos(φ) + i sin(φ)
≡ Re(z) + iIm(z).

What this means is that in quantum mechanics, we have a

little circle "above" each possible location.
86 physics from finance

Each point on these circles represents a possible phase fac-

tor. Completely analogously to what we discussed for the
prices of copper, these phase factors do not have any direct
14
This is discussed in meaning because we can shift them however we want.14
a bit more detail in However, analogous to how the prices are, after all, essential
Appendix C.2.
to describe copper, the phase factors are essential to describe
elementary particles. The main observation is again that the
relationship is a bit more indirect than what we would’ve
naively thought.

That’s really the only new thing we needed to understand.

All that is left to do to derive a theory describing electro-
magnetic interactions is to apply all the lessons we have
learned in the financial toy model to our new internal
(charge) space.
 quantum mechanics 87

4.2 Electrodynamics

First of all, we can imagine that we want to allow for the 15

The interplay between
phase factor to be independently shifted at each location. copper and arbitrage
opportunities is de-
scribed by (Eq. 2.34)
Of course, this is again possible but we then need some-
thing which keeps track of the local shifts. In other words, ∂ν Fµν = µ0 Jµ .
if we allow the phase factor to be shifted at each location
independently, we need something which glues the circles
together in such a way that the dynamics of the system re-
mains unaffected by such local shifts. This is analogous to 16
In other words, it
the exchange rates in our financial toy model which get ad- is conventional to
give special names to
justed every time a local currency is changed. Again, we certain components
denote the glue by Aµ . But, in physics, it is conventional to of the electromagnetic
field tensor Fµν . Since,
call Aµ the electromagnetic potential. µ, ν ∈ {0, 1, 2, 3} the
field tensor can be
The most interesting part of the story is once more how the written as a (4 × 4)
matrix:
structure of the internal space affects the dynamics within
Fµν =
the system. And again, we only need to recall what we have  
F00 F01 F02 F03
learned in the financial toy model. There we argued that F
 10
F
F11
F21
F12
F22
F13 

F23 
=
20
F30 F31 F32 F33
currency arbitrage opportunities have a nontrivial effect on  
0 − E1 /c − E2 /c − E3 /c
the dynamics. These are described by the object (Eq. 2.22)  E /c
 1
 E /c
0
B3
− B3
0
B2 
− B1 

2
E3 /c − B2 B1 0
∂Aν (t, ~x ) ∂Aµ (t, ~x ) We can write the field
Fµν (t, ~x ) ≡ − (4.18)
∂x µ ∂x ν tensor like this, because
it is antisymmetric:
and arise whenever there is copper present in the system.15 Fµν = − Fνµ . This
follows automatically
Completely analogously, we now introduce from the definition in
terms of the potential
∂Aν (t, ~x ) ∂Aµ (t, ~x ) Aµ . The antisymmetry
Fµν (t, ~x ) ≡ − (4.19) implies that there must
∂x µ ∂x ν be zeroes on the diago-
for the electromagnetic potential Aµ . nal since, for example,
F11 = − F11 , which is
only true if F11 = 0.
In this context, we call Fµν the electromagnetic field tensor Moreover, we have, for
example, F12 = − F21
and its components are usually called the electric field and
and therefore there are
the magnetic field.16 in total only 6 indepen-
dent components.
88 physics from finance

Moreover, the equation (Eq. 2.34)

∂ν Fµν = µ0 Jµ (4.20)

holds again. But this time, Jµ does not describe the flow
of copper but the flow of electric charges through the sys-
tem. In other words, this equation describes the interplay
between electric charges and the electromagnetic field.

That’s it. We’ve successfully derived electrodynamics.

Admittedly, we have glossed over lots of details. Never-

theless, it’s true that we’ve successfully derived the correct
structure describing the interplay between electric charges
and the electromagnetic field solely by applying the lessons
17
If you want to under- we learned in the financial toy model.17
stand electrodynamics
in more detail, you
might enjoy my book Let’s summarize what is going on here.18
Jakob Schwichten-
berg. No-Nonsense In quantum mechanics, we use complex functions although
Electrodynamics. No- everything we can measure in experiments has real values.
Nonsense Books, 2018a.
ISBN 978-1790842117 This means that we can multiply the functions we use in
quantum mechanics to describe what is going on with (unit)
18
Don’t worry if you’re
confused. We will phase factors without changing anything.
discuss everything we
just discovered multiple
times from various
This gauge symmetry is completely analogous to the gauge
perspectives. symmetry we discussed in the financial toy model. We
therefore concluded that we also have an internal space,
analogous to the money space in the toy model. The only
difference is that now the internal space above each point is
a circle and not a line. Again, the total internal space is all
these local internal spaces taken together.

The main idea is then that this internal space is essential to

describe the interactions of elementary particles analogous
to how the internal money space is essential to describe how
copper gets moved around in our toy model. In particular,
 quantum mechanics 89

there are elementary particles which have a direct influ-

ence on the structure of the internal space. We call these
particles electrically charged particles. Wherever a charged
particle is present, the structure of the internal space gets
modified such that the various circles are glued together
non-trivially.19 19
This is analogous to
how currency arbitrage
opportunities show up
In turn, the behavior of charged objects gets modified by the whenever exchange
non-trivial structure of the internal space.20 rates favor a loop in
currency exchanges.

The picture that emerges this way is the following. Par- 20

In our financial toy
ticles carrying electric charge modify the structure of the model, the behavior
of traders carrying
internal space which we call charge space. In turn, all elec- copper gets modified
trically charged particles feel the structure of this internal as soon as arbitrage
space. Maybe it helps to imagine that our internal space opportunities show up.

gets curved due to the presence of the electric charges and

therefore the path of electrically charged particles gets mod-
ified similarly to how the path of a marble is different when
it rolls on a curved surface as opposed to when it rolls on
a flat surface. And this is really how electrically charged
particles influence each other. In other words, this is how
electromagnetic interactions work. Electrically charged par-
ticles do not bump into each other. Instead, they respond
to each other more indirectly by modifying the structure of
charge space.21 21
Again, don’t worry
if this is not perfectly
clear at this point. We
While this story may still seem somewhat abstract and will talk about all the
confusing, it is worth thinking about it in detail because all ideas mentioned here
multiple times below.
fundamental interactions work like this. The story is always
the same, only the details are different.

Now, how can we implement the idea outlined above in

quantum mechanics?
90 physics from finance

4.3 Interactions in Quantum Mechanics

22
One way to under-
stand this is by noting Luckily, everything works completely analogously to what
that products of the
form ψi? ( x )∂ x Ψ( x ) are we already discussed in Section 2.2.1 for the financial toy
not invariant under model.
local transformations
Ψ( x ) → eie( x) Ψ( x ) The main idea is once more that if there is a non-zero elec-
ψi? ( x ) → e−ie( x) ψi? ( x ) tric field strength Fµν 6= 0, we must be careful whenever
as we can check explic- we calculate derivatives. This is necessary because when we
itly:
calculate the derivative of a function, we compare the values
ψi? ( x )∂ x Ψ( x ) →
  of the function at two different locations. And as soon as
ψi? ( x )e−ie( x) ∂ x eie( x) Ψ( x ) there is a non-zero electromagnetic field strength Fµν 6 = 0,

= ψi? ( x )∂ x Ψ( x ) we know immediately that there will be a non-trivial elec-

+ ψi ( x ) i∂ x e( x ) Ψ( x ) tromagnetic potential Aµ somewhere. But when we want to
 
?

compare the value of a function at two locations, we need

6= ψi? ( x )∂ x Ψ( x ) .
to take this potential into account. This is analogous to how,
However we can get
something invariant by
in the financial toy model, the exchange rates are essential
replacing the derivative in determining the flow of copper (copper current) between
∂ x with the appropriate neighboring countries.
covariant derivative
Dx = ∂ x − iA x ,
There is only one small difference which has to do with the
where the bookkeeper imaginary unit i in the exponent of our gauge transforma-
A x adjusts automati-
cally tions in quantum mechanics, eie( x) . To take the complex
A x → A x + ∂ x e( x )
structure of the internal space of quantum mechanics into
account, we have to add an imaginary unit to our covariant
whenever we perform
a transformation of derivative formula.22
the wave function (see
Eq. 2.21). Let’s check
this explicitly:
To cut a long story short: to take the effects of a non-zero
electromagnetic field strength into account, again we need
ψi? ( x )(∂ x − iA x )Ψ( x ) → to replace our ordinary derivatives with covariant deriva-

= ψi? ( x )e−ie(x) ∂ x − iA x
tives

∂µ → Dµ ≡ ∂µ − iAµ . (4.21)

− i∂ x e( x ) eie(x) Ψ( x )
= ψi? ( x )∂ x Ψ( x ) In other words, this is what the derivative of a function
 
?
+ ψi ( x ) i∂ x e( x ) Ψ( x ) looks like in the presence of a non-trivial electromagnetic
− iψi? ( x ) A x Ψ( x )
 
− iψi? ( x ) ∂ x e( x ) Ψ( x )
= ψi? ( x )(∂ x − iA x )Ψ( x ) .
 quantum mechanics 91

potential Aµ .

Therefore, the Schrödinger equation (Eq. 4.2)

∂0 ψ(t, ~x ) = iD∂2i ψ(t, ~x ) (4.22)

becomes
   2
∂0 − iA0 ψ(t, ~x ) = iD ∂i − iAi ψ(t, ~x ) . (4.23)

This is what the Schrödinger equation looks like in the

presence of a non-zero electromagnetic potential. It’s really
that simple and completely analogous to what we discussed
for the financial toy model.

A small detail that we need to talk about before we move on

is that different particles experience the influence of an elec-
tromagnetic field differently. To take this into account, we
introduce a new constant q in front of the electromagnetic
potential
   2
∂0 − iqA0 ψ(t, ~x ) = iD ∂i − iqAi ψ(t, ~x ) . (4.24)

This new constant q is known as the electric charge of the

particle in question. The constant q is a measure of how
strongly the particle reacts to the presence of the electro-
magnetic potential Aµ . 23 23
In the financial toy
model, q encodes how
strongly traders react to
the prices of copper and
arbitrage opportunities.

Now, before we move on and discuss the remaining two

internal spaces (isospin space and color space) and space-
time itself, it may be useful to take a step back and read
Appendix D. In this appendix, everything we discussed so
far is rephrased a bit more systematically using the proper
mathematical notions. But, of course, you can also move on
to the next chapter and read Appendix D afterwards.
92 physics from finance

In either case, we are ready to move on to the remaining

two internal spaces.

4.4 Isospin Space and Color Space

Unfortunately, to properly describe isospin space, color

space and the corresponding weak and strong interactions,
we need quantum field theory. This is because we need
to be able to describe how different types of particles can
pop in and out of existence. And to describe systems in
which the number of particles is not conserved, we need the
24
To accurately de- formalism of quantum field theory.24
scribe electromagnetic
interactions at high
energies, we also need Nevertheless, at this point we are capable of getting at least
quantum field theory a glimpse of what is going on.
because then the parti-
cle number is no longer
conserved. But for
electromagnetic inter-
actions, we can develop
a theory which is valid
at lower energies (this In the previous sections, we discussed how we can under-
is what we did above) stand electromagnetic interactions by making use of the
while for strong and
weak interactions this internal space we call charge space. The main idea there
is not possible because was that particles which carry electric charge curve this
they are short ranged
and therefore their
charge space and in turn, are influenced by the curvature of
effect is only visible at the charge space. This is how electromagnetic interactions
high energies. take place.

In this section, we will do exactly the same thing again but

this time we will understand how we can describe weak
interactions using isospin space and how we can describe
strong interactions using color space. In fact, it’s really only
the space which is different. The main ideas stay the same.

This means that particles carrying isospin curve isospin

 quantum mechanics 93

space and in turn, are influenced by the curvature of isospin

space. This is how weak interactions between elementary
particles take place. Similarly, particles carrying color curve
color space and feel the structure of color space. This is how
strong interactions take place.

Therefore, our only task is to understand isospin space and

color space. We just need to recall what we discussed when
we talked about charge space and then go one step beyond
that.

But before we dive in, here’s one short comment. A ques-

tion you may have is: why exactly these spaces? Unfortu-
nately, the answer is: nobody knows yet. This is an open
question in physics. We only know that these spaces are
the right ones because we can observe their implications in
experiments. But we don’t know why only these specific
structures show up.25 25
A popular idea
among theoretical
physicists is to add
Now, what is isospin space and color space? new internal spaces
or to unify the known
spaces into one higher-
In the previous sections, we talked about internal spaces dimensional internal
first and only afterwards about how they are related to in- space. These ideas are
teractions. But isospin space and color space are a bit more known as extended
gauge models and
abstract and it is convenient to discuss the corresponding Grand Unified Theories.
interactions first.

A first crucial fact we need is that in quantum theories there

is not only a fundamental randomness in the movement of
particles but also in their interactions. In a quantum theory,
we can only calculate the probability that a certain interac-
tion takes place. This is analogous to how we were previ-
ously only able to calculate the probability to find a certain
amount of copper or electric charge at a particular location.
But the randomness in the movement of particles and the
randomness in interactions are, of course, not completely
94 physics from finance

independent. How strongly two elementary particles inter-

act depends crucially, for example, on the distance between
them. And, if the movement of the particles is random, we
can only calculate the probability to find them at a particu-
lar distance from each other. Therefore, we can’t be certain
26
In the illustration, about the strength of the interaction.26
an electron performs
a random walk before
it interacts with an
electron-neutrino, νe .
The dashed line indi-
cates the interaction
and the shaded area
illustrates the region
in which an interac-
tion becomes highly
probable.

A second crucial fact which led to the discovery of isospin

27
We will discuss what space and color space is that weak and strong interactions
this really means in a treat certain particles as if they were merely different states
moment.
of the same particle. For example, weak interactions are
28
The color names blind to the electric charge of particles and cannot distin-
red, blue, green have guish between an electron and an electron-neutrino; thus, in
nothing to do with
the colors we see in the internal isospin space, these two particles are perceived
everyday life and are as different states of a single particle.27 Similarly, strong
only used as convenient
labels. Richard Feyn- interactions treat red up-quarks, blue up-quarks and green
man famously called up-quarks as different states of the same object.28 For the
his colleagues "idiot second example, this seems a bit more obvious thanks to the
physicists" for choosing
this confusing name, name choices, but the logic is really exactly the same.
"which has nothing to
do with color in the
normal sense.". If we combine these two observations, we can understand
what color space and isospin space are.
 quantum mechanics 95

The meaning of isospin space is that it encodes how much

of a given state is an electron and how much of it is an
electron-neutrino.

In other words, if there is an electron/neutrino state in our

system, the corresponding axes of isospin space encode the
"electron-ness" and "neutrino-ness" of the state.

29
Maybe it would help
to call the electron-
neutrino an up-electron
and the ordinary
electron a down-
electron. Similarly, we
could call a muon-
neutrino an up-muon
and the regular muon
a down-muon, and a
strange-quark an up-
charm quark and the
regular charm quark
a down-charm quark
Similarly, if there is an up-quark/down-quark state in our etc. This way, it would
system, the axes of isospin space describe the "up-ness" and become clear that
we are dealing with
"down-ness" of the states.29 different states of the
same thing. We could
then say that the axes of
isospin space in general
describe the relative
"up-ness" and "down-
ness" of the state at
hand. The color names
described below are a
bit more helpful in this
regard. However, we
don’t use these kinds of
names for the neutrinos
etc. because historically
the isospin structure
was discovered much
later than most of the
fermions.
96 physics from finance

Above each location, we have a copy of isospin space, anal-

ogous to how we have a copy of price space in the financial
toy model. The total isospin space consists of all these local
internal isospin spaces glued together. A specific state of
our system corresponds to exactly one point in each of these
spaces. A specific point in isospin space corresponds to a
specific probability that we find an electron or an electron-
neutrino at a given location.

Analogously, we have above each location a copy of color

space and the total color space consists of all these local
internal color spaces taken together. The axes of this space
encode the redness, blueness and greenness of the state in
question.

A specific point in color space corresponds to specific proba-

bilities to find, say, a given up-quark as a red up-quark, blue
up-quark or green up-quark.

This is certainly somewhat confusing and it takes some time

to get used to this kind of thinking. But we will see in a
moment that the rest of the story is completely analogous to
what we discussed previously.
 quantum mechanics 97

4.4.1 Symmetries of Isospin and Color Space

First of all, as soon as we are dealing with a space of any

sort, we can ask: what are its symmetries? The most natural
assumption is always that each space possesses symmetries.
We saw this previously for the price space and for charge
space. And again the same is true for isospin space and
color space. The symmetries of charge space are simply 30
This is discussed in
rotations in the complex plane. Mathematically, we describe a bit more detail in
Appendix D.
these rotations using the group U (1).30
31
SU (2) consists of
Now, the symmetries of isospin space and color space are, all unitary (2 × 2)
matrices with unit
in some sense, simply higher-dimensional analogs of these determinant. The S
rotations. The symmetry group of isospin space is known as denotes "special" which
means determinant
SU (2) and the symmetry group of color space is SU (3).31 one. The U denotes
"unitary", which means
that the group elements
fulfill the condition
g? g = 1. Analogously,
4.4.2 Isospin and Color Connections SU (3) consists of all
unitary (3 × 3) matrices
with unit determinant.
In this sense, these
Again, the key idea is that we allow arbitrary local con-
symmetries are the
ventions. In the finance toy model, we allowed countries next simplest thing
to use local currencies. Here, we allow each location to we can try after U (1).
(Note that the group
choose freely which axis is called the electron axis and SU (1) also exists
which is called the neutrino axis. These are just labels and but consists only of
the trivial element 1
local choices do not matter as long as we take them into
which does nothing
account consistently. This is again where bookkeepers be- at all.) Discussing
come important. In the context of weak interactions, it is group theory and these
groups in detail would
conventional to use the symbol Wµ for the bookkeepers. lead us too far astray,
but, if you’re intersted,
you can find a detailed
While the local choice between electron and neutrino axes
discussion in
doesn’t make any difference, we need to be careful when
Jakob Schwichten-
we compare states at two different locations. The relative berg. Physics from Sym-
"electron-ness" and "neutrino-ness" between the states at metry. Springer, Cham,
Switzerland, 2018b.
two locations is important and cannot depend on local ISBN 978-3319666303
98 physics from finance

32
This is analogous conventions.32
to how a price differ-
ence in two different
countries is something
physical. (At least, if
we convert the prices
to a common currency
before we compare
them.)

33
There is an important
point we are glossing
over here. Since our
internal spaces are now
higher-dimensional, The bookkeepers Wµ keep track of the local conventions
we need multiple types
of bookkeepers. For and always adjust so that the relative "electron-ness" and
isospin space, we need "neutrino-ness" stay unaffected by local conventions.33
three bookkeepers
Wµ1 , Wµ2 , Wµ3 . In math-
ematical terms, this Formulated a bit more technically, the bookkeepers Wµ
follows because the are (again) the connections which allow us to consistently
adjoint representation
of SU (2) is three- transport a given state to a different location.34 This trans-
dimensional. portation is necessary whenever we want to compare a state
at one location with another state at another location.35
34
The mathematical
concept of connection
is discussed in more We can’t compare them directly because the description of
detail in Appendix D.4.
states can depend on local conventions. But, by transporting
35
For example, when- one state to the location of a second state, we make sure that
ever we calculate the we compare them using the same local conventions.
derivative.

Analogously, the redness, blueness and greenness of a given

state can be redefined arbitrarily at each location.
 quantum mechanics 99

But if we want to compare states at two different locations,

the relative redness, blueness and greenness become impor-
tant.

36
Color space is a lot
more complicated than
isospin space and we
need eight bookkeepers
Therefore, we once more need bookkeepers Gµ to keep track to keep track of all
local conventions. In
of these local choices.36 mathematical terms
this follows because the
The final puzzle piece we need to talk about is what all this adjoint representation
of SU (3) is eight-
really has to do with weak and strong interactions. dimensional.
100 physics from finance

37
Take note that only
for electromagnetic
interactions does the
4.4.3 Weak and Strong Interactions
field strength tensor
have the simple form
(Eq. 2.22)
Now, the main idea is once more completely analogous
∂Aν ∂Aµ
Fµν ≡ − . to what we discussed in the financial toy model and for
∂x µ ∂x ν
electromagnetic interactions. The freedom to choose lo-
The field-strength
tensor for weak and cal conventions is a nice feature but not really essential.
strong interactions Nevertheless, it is possible that our bookkeepers become in-
contains an additional
term: dispensable. In the financial toy model this is the case when
∂Aν ∂Aµ there are currency arbitrage opportunities (Fµν 6= 0). The ex-
Wµν ≡ −
∂x µ ∂x ν istence of such an opportunity is a real feature of the model
a b c
+ gTa f bc Aµ Aν , which influences the dynamics and implies directly that
where g is a constant there has to be a non-trivial exchange rate Aµ somewhere.
which characterizes the
Analogously, if there is a non-zero electromagnetic field
strength of the corre-
sponding interaction, strength in the model Fµν 6= 0, we know immediately that
a are the so-called
f bc the electromagnetic potential Aµ has to be non-vanishing
structure constants
and Ta is a generator somewhere.
of the corresponding
symmetry group. This
Completely analogously, our isospin bookkeepers Wµ and
term was only absent
for the electromagnetic color bookkeepers Gµ become indispensable when there
field-strength tensor is a non-zero weak field strength and a non-zero strong
Fµν because the struc-
ture of charge space field strength somewhere.37 Don’t let the somewhat un-
is extremely simple. fortunate names "weak" and "strong" confuse you. We are
Mathematically, we get
really dealing with new types of interactions and not simply
an extra term like this
whenever the symmetry weak electromagnetic or strong electromagnetic interactions.
group is non-abelian. For example, the weak force is responsible for the decay of
Physically, this extra
term describes self- nuclei while the strong force holds protons and neutrons
interactions of the together.38 These are phenomena which cannot be explained
corresponding gauge
by electromagnetic interactions.
bosons. We will see
something very sim-
ilar when we discuss There are particles which can modify the structure of the to-
the tensor describing
gravity. tal isospin and color space such that non-zero field strengths
show up. We say that a particle which modifies the struc-
38
Protons and neutrons ture of isospin space carries isospin. A particle which mod-
consist of quarks.
ifies color space carries color. Again, this is analogous to
how only particles carrying electric charge modify the struc-
 quantum mechanics 101

ture of charge space. And only particles which carry electric

charge, isospin charge or color charge are influenced by the
structure of the corresponding space.

This is how particles interact with each other. 39 39

Hopefully this dia-
gram starts to make a
bit more sense.

influence
#
Elementary Particles Arena of Physics
c
influences

But this geometrical picture is not only important because it

lets us understand all these fundamental interactions from
a common perspective. We can also use it to understand an
important feature of weak and strong interactions.

To understand this feature, we need to ask: what does it

really mean that bookkeepers are indispensable here?

Let’s assume we are 100% certain that we are dealing with

an electron at a particular location. Moreover, let’s assume
that we have fixed all local conventions and we try to keep
them as simple as possible.40 The main point is that if there 40
In the financial toy
is a non-zero weak field strength somewhere in the sys- model this would mean
that we try to use the
tem, we necessarily have a non-trivial bookkeeper Wµ . same (global) currency
That is, even if we try to choose our local isospin spaces everywhere. However,
this turns out to be
extremely cleverly such that our pure electron state will cor- impossible if there is an
respond to a pure electron state everywhere (if we move it arbitrage opportunity
around in time and space), we will fail because otherwise since this means that
there must be a non-
the bookkeeper Wµ would be zero everywhere. We know trivial exchange rate
that Wµ has to be non-zero somewhere since there is a non- somewhere.
vanishing field strength. This implies that even if we start
with a pure electron state and try to make sure that it cor-
responds to a pure electron state everywhere in time and
space, we will get a state with a non-vanishing electron-
102 physics from finance

neutrino component somewhere.

In physical terms, this means that a pure electron state can

be transformed into an electron-neutrino state through weak
interactions. Formulated more precisely, if there is a non-
zero weak field strength, there is a non-zero probability to
find that our electron has turned into an electron-neutrino at
a later point in time.

Do you wonder about the conservation of electric charge?

After all, an electron carries electric charge while an electron-
neutrino does not. Rest assured: in all particle interactions
observed so far, electric charge was found to be conserved.
Therefore, if our electron is transformed into a neutrino, the
electron’s electric charge must go somewhere.

To understand how this works, we need to recall that the

non-zero field strength which affects the electron in question
is caused by a second particle. But, as usual in physics, this
41
action = reaction second particle is, in turn, influenced by the electron.41 This
42
Take note that this second particle is exactly where the electron’s electric charge
is only possible if our ends up.
connections are able to
carry electric charge.
And this is indeed The picture which emerges this way is the following. We
what we observe in
nature. Two of the three start with two particles: an electron at some location A and,
particles associated say, an electron-neutrino at a different location B. Each of
with weak interactions these two particles modifies the structure of isospin space
do indeed carry elec-
tric charge. This also and is influenced by it.
indicates that there
is a close connection
between weak and This is how they "talk" to each other. What we discussed
electromagnetic interac- above means that there is a non-zero probability that after
tions. In fact, in modern some time we will find an electron-neutrino at location A
physics we usually
talk about electroweak and an electron at location B. 42
interactions. To prop-
erly understand this
requires an understand-
ing of spontaneous
symmetry breaking
which is somewhat
beyond the scope of this
book.
 quantum mechanics 103

This is a real physical process mediated by weak interac-

tions which we can observe in nature.

Completely analogously, an up-quark can turn into a down-

quark and vice versa.
43
The logic is really
the same as before.
The main reason why
we treat different
states related to weak
interactions as truly
different particles is
that the corresponding
symmetry (SU (2))
is broken in nature.
As a consequence,
the two states inside
each doublet (the
electron-neutrino and
the electron or the up-
quark and the down-
quark) have different
And exactly the same picture emerges if we consider strong masses. This is why
interactions. The only difference is that it’s the "color" of the we give these states
completely different
quarks which gets modified.43 Specifically, this means, for names. In contrast, the
example, that a red up-quark gets transformed into a blue symmetry associated
up-quark or a green up-quark. But again, we must make with strong interactions
(SU (3)) is intact, which
sure that everything important is conserved. This means is why we treat the
that color charge must go somewhere and a different quark different color states as
different versions of the
at some other location must therefore provide the required same particle. (A green
color and absorb the color which our first quark "lost". quark has the same
mass as a blue quark
and a red quark.)
104 physics from finance

Take note that the various interactions are not mutually ex-
clusive. Instead, they happen simultaneously all the time.
A particle which carries electric charge, isospin charge and
color charge (e.g. a red up-quark) is influenced by the struc-
ture of color space, isospin space and charge space simulta-
neously! Therefore, if we want to calculate the probability of
finding a given particle in a particular final state, we need to
44
Unfortunately, a take lots of things into account. As mentioned before this is
proper discussion of possible using quantum field theory. One of the main goals
quantum field theory
is far beyond the scope in quantum field theory is indeed to calculate the proba-
of this book. But if you bilities of finding a given set of particles in a different state
want to learn more
about it, you might after some time.44
enjoy my book

Jakob Schwichten- There are two final comments to make before we move on
berg. Physics from Sym- to the final part of the arena of physics and the interaction
metry. Springer, Cham,
associated with it.45 Firstly, maybe you wonder, since the
Switzerland, 2018b.
ISBN 978-3319666303 effect of weak and strong interactions is to turn particles
45
Spacetime and grav-
into different particles, what exactly do electromagnetic
ity! interactions do?

The answer is: electromagnetic interactions change the

46
The connection be- momentum of particles.46 If we start with two electrons,
tween charge space, each with some given momentum, we end up again with
which encodes the free-
dom to rotate the phase two electrons but they possibly carry different momenta.
of wave functions, and
momentum can be
understood as follows.
The momentum of
a particle described
by a wave function
ψ = Aeiϕ is directly
proportional to the
derivative of the phase
ϕ of the wave function.
If there is a non-zero
electromagnetic field,
the momentum of the
particle becomes pro-
portional to ∂µ ϕ − qAµ
where q denotes the
electric charge of the The momentum one particle loses is transported via the
particle.
 quantum mechanics 105

electromagnetic connection Aµ to the location of the second

particle.47 47
Also take note that
the momenta carried
by the final particles
Similarly, when an electron scatters with a muon, we again in weak and strong
get as a final state an electron plus a muon but with differ- interactions are, in
general, different from
ent momenta. the momenta carried by
the initial particles. Yet,
the total momentum is
always conserved.

And, of course, electrons and muons can also interact elec-

tromagnetically with quarks since all these particles carry
electric charge:

Secondly, maybe you wonder why we only observe electro-

magnetic interactions in everyday life but never weak and
strong interactions? The reason for this is that strong and
weak interactions are extremely short-ranged while electro-
magnetic interactions are long-ranged. This means that a
106 physics from finance

particle which carries isospin only influences isospin space

in a small region of spacetime around it. Similarly a particle
which carries color charge only influences color space in a
48
This feature of weak tiny neighboring region of spacetime.48
interactions can be
understood because the
corresponding gauge In contrast, every particle which carries electric charge in-
bosons are massive. fluences charge space in a huge region around it. This is
(In the pictures shown
above you can imagine except, of course, when there is another particle nearby with
that the line between exactly the opposite charge. Then, their effects on charge
the electron and the space cancel each other. This happens, for example, because
neutrino represents an
additional particle Wµ the charge of an electron is −e and the charge of a proton
which mediates the is +e. Thus, we get an electrically neutral composite object
interaction between the
two.) This is a result whose combined effects on charge space cancel.
of the spontaneous
symmetry breaking
of the corresponding
SU (2) symmetry
through the Higgs
mechanism. As a result,
the massive gauge
bosons cannot travel
very far before they
decay into lighter
particles and that’s
why weak interactions
are short-ranged. In
contrast, the gauge
bosons mediating
strong interactions are
massless. Nevertheless,
strong interactions are
short-ranged because
colored objects always
lump together to color-
neutral objects. This
mechanism is known as
confinement but there
is still no completely
satisfactory explanation
for why and how this
happens.
5

Special Relativity and

General Relativity

First of all, a short disclaimer. As before, we will only focus on

very special aspects of special relativity and general relativity.
There is, of course, a lot more one could and should say about
them. But discussing the nuances of special and general relativity
requires at least a few hundred pages and is thus not something
we can realistically do here. We will skip many mathematical
details and lots of important concepts are not mentioned at all.
Nevertheless, the main features of both theories can be understood
beautifully using the same ideas discussed already in the previous
chapters. And this is the only thing we will focus on.

So far, we have talked about internal spaces and how we

can use them to understand how particles interact with
each other. The main idea was that particles which carry
charges modify the structure of the corresponding internal
spaces. At the same time, each such particle is influenced by
108 physics from finance

1
To be influenced the structure of the internal spaces.1 This is how particles
by the structure of a interact even though they do not directly touch each other.
given internal space, a
particle must carry the
charge that is associated Now, there is exactly one additional known space (space-
with that internal space.
For example, only time) and exactly one additional known fundamental inter-
electrically charged action which we haven’t talked about so far. Gravity.2
particles are influenced
by the structure of
charge space. This is At this point you will probably not be surprised if I tell
discussed in a bit more you that there is a direct connection between the structure
detail in Appendix D.
of spacetime and gravity, completely analogous to how
2
There could, of course, there is a direct connection between the structure of charge
be additional internal space and electromagnetic interactions. But historically, un-
spaces and fundamen-
tal interactions. But
derstanding this connection was a huge intellectual step
so far, only gravity, forward and is rightfully regarded as one of the most beau-
electromagnetic, weak tiful insights in physics. It took a genius like Einstein more
and strong interactions
have been observed in than a decade to figure it out. Nowadays, with the power of
experiments. hindsight, we can understand it almost effortlessly in just a
few minutes.

In fact, not only is there a close connection between gravity

and spacetime, but we can understand this interplay com-
pletely analogously to how we have come to understand the
connection between the other fundamental interactions and
their corresponding (internal) spaces.

The main idea is once more that objects carrying the cor-
responding charge, here energy-momentum, modify the
structure of spacetime. In turn, all objects carrying energy-
momentum are influenced by the structure of spacetime.
This is how gravitational interactions take place!

Moreover, analogous to how a non-zero electromagnetic

field strength corresponds to a non-zero curvature of charge
space, a non-zero gravitational field strength corresponds to
a non-zero curvature of spacetime.
 special relativity and general relativity 109

In other words, while a few details are different, the general

story-line is exactly the same.3 3
Take note that even
though general rela-
tivity is more than a
To understand all this a bit better, we start by discussing century old, experts
the local structure of the space in question.4 Afterwards, we still argue about basic
concepts; like which
again need something ("bookkeepers") to glue these individ- geometric structure
ual local spaces together. Then we can talk about situations correctly describes our
in which our individual local spaces are glued together universe. But like the
discussion about the
non-trivially. As before, such situations are characterized correct interpretation
by a non-zero curvature. But now we will talk about the of quantum mechanics,
this is not something
curvature of spacetime and not that of some internal space. we will discuss here.
Finally, we will talk about the Einstein equation which de-
scribes how a non-zero curvature of spacetime arises, anal-
4
The local structure in
the financial toy model
ogous to how Maxwell’s equations describe how a non-zero was simply a line, while
curvature of charge space arises whenever electric charges for quantum mechanics
our local internal space
are present. is a circle.

With this plan in mind, let’s dive in.

5.1 Special Relativity

What do we know about the structure of space and time?

Maybe this seems like a strange question, but there are

many important things we can learn by answering it.

First of all, we usually think of the space we live in as a

relatively boring arena. To describe the location of an object,
we introduce a coordinate system and determine three
numbers ( x, y, z). These numbers indicate the location of the
object with respect to three coordinates axes.
110 physics from finance

Since, in general, we need three such numbers we say that

space is three-dimensional. And as you can easily check
with a ruler, the distance ∆s between any two objects in
this three-dimensional space can be calculated using the
5
From another perspec- Pythagorean theorem5
tive, we can say that we
calculate the length of ∆s2 = ∆x2 + ∆y2 + ∆z2 . (5.1)
the vector
∆x
 
Since we can calculate distances like this, we say that the
~v =  ∆y 
∆z
space we live in is Euclidean.6

pointing from one point

to another: Now, in physics we are not only interested in knowing where
|~v|2 = ~v · ~v
something happens but also when. After all, one of our main
= ∆x2 + ∆y2 + ∆z2 .
goals in physics is to describe how objects move around.
This requires that we take time into account.
6
Maybe you wonder
why we care about Therefore, to properly describe an object in physics, we
such a trivial fact and
actually need four numbers (t, x, y, z). The time coordi-
even introduce a special
name? We will see in a nate t tells us the time at which the object is at the location
moment that the above ( x, y, z). This means we add a fourth axis to our coordi-
relation is not always
true, i.e. there can be nate system. In other words, we now not only use a spatial
non-Euclidean spaces. coordinate system, but also a spacetime coordinate sys-
tem. While ( x, y, z) describes the location of an object in
 special relativity and general relativity 111

space, we say that (t, x, y, z) describes the location of an

event in spacetime. And since we need four numbers to
describe the location of an event, we say that spacetime is
four-dimensional.

Hopefully you are not too bored or confused by these lines

of thought because things are about to get really interesting.

We defined above the spatial distance between two objects

(Eq. 5.1). Now that we’ve added time to our coordinate
system, we can ask: what’s the distance between two events
in spacetime? Naively, we might write down

∆s2 = ∆t2 + ∆x2 + ∆y2 + ∆z2 (5.2)

but this doesn’t make any sense. The differences in the spa-
tial components (∆x, ∆y, ∆z) are measured in meters, while
∆t is measured in seconds. This means we are comparing
apples with oranges in Eq. 5.2. However, we can fix this
problem by introducing a new constant c which has units
meters per second:

∆s2 = c2 ∆t2 + ∆x2 + ∆y2 + ∆z2 . (5.3)

Now, the first term in the sum on the right-hand side has
meters 2
seconds2 = meters2 as it should be.7


units second 7
We will talk about
the meaning of this
constant in a moment.
One of the big discoveries in physics was that this formula
is not the relevant one.
112 physics from finance

Instead, the relevant expression for the (squared) "distance"

8
Historically, the between events is8
Michelson-Morley ex-
periment showed the
constancy of the speed
of light, independent of ∆s2 = c2 ∆t2 − ∆x2 − ∆y2 − ∆z2 . (5.4)
the reference frame in
which it is measured.
This fact, which lies
at the heart of special
In words, this means there is a relative minus sign between
relativity, was formal-
ized by Einstein and the (squared) spatial distance ∆x2 + ∆y2 + ∆z2 and the
Minkowski. But before (squared) temporal distance ∆t2 .9 While this formula may
we address this, we
need to understand the look extremely strange, from a mathematical perspective,
physical implications of it simply tells us that the local structure of spacetime is not
this relation.
what we would’ve naively expected. In some sense this is
analogous to how we discovered that the local structure
of charge space is not a simple line but a circle. Here, we
discover that the local structure of spacetime is not Eu-
9
We will talk about the clidean (Eq. 5.3) but Minkowskian (Eq. 5.4). The main fea-
physical meaning of the
minus sign in Eq. 5.4 in ture of a Minkowski space compared to a Euclidean space
a moment. is the way we define distances. Mathematically, we denote
the four-dimensional Euclidean space as R4 and the four-
dimensional Minkwoski space as R1,3 .

But the minus sign not only encodes an important fact

about the local structure of spacetime, it also has extremely
important physical implications.

To understand this, let’s consider two events called A and

B which are spatially separated by 3 meters. Moreover, we
assume that B is caused by A. For example, we can imagine
that event A is the emission of a light pulse and event B
is the detection at a location that is 3 meters away. Now,
how soon after A can B happen? In other words, what’s the
minimum temporal distance ∆t between the two events?

We can answer this question by looking at Eq. 5.4. The

(squared) spatial distance between the two events ∆x2 +
 special relativity and general relativity 113

∆y2 + ∆z2 = (3 meters)2 is fixed. Therefore

∆s2 = c2 ∆t2 − ∆x2 − ∆y2 − ∆z2

∆x2 + ∆y2 + ∆z2 = (3 meters)2

y
2 2 2
= c ∆t − (3 meters) . (5.5)

Our goal is to find the minimum allowed value of ∆t. A

crucial observation is that ∆t cannot be arbitrarily small and
certainly cannot be zero; that is, if we forbid instantaneous
action at a distance. Since the two events are causally re-
lated, the pulse of light necessarily needs some time (even
if very short) to travel the spatial distance between the (sep-
arate) locations of emission and detection. Therefore, the
minimum allowed value of ∆t is precisely the value for
which ∆s2 is zero.

The main point is that there is a non-zero minimum time

period between two events which happen at two different
locations. This implies that there is a maximum velocity at
which a signal can travel from one point to another.

Using Eq. 5.4 we can learn something important about this

maximum velocity. A signal which travels at the maximum
velocity between the two events, needs the minimum time
interval ∆t to travel the distance ∆x2 + ∆y2 + ∆z2 . Above,
we argued that the minimum time interval ∆t corresponds
to ∆s = 0. Putting this into Eq. 5.4 yields

∆s2min = 0 = c2 ∆t2min − ∆x2 − ∆y2 − ∆z2

y

∆x2 + ∆y2 + ∆z2

c2 = . (5.6)
∆t2min
This is interesting because in general

∆x2 + ∆y2 + ∆z2

p
v≡ (5.7)
∆t
is exactly the velocity at which a signal travels the distance
∆x2 + ∆y2 + ∆z2 .10 Therefore, Eq. 5.6 tells us that a signal
p
10
A velocity is always
defined as a spatial
distance divided by the
time interval ∆t needed
to travel the distance.
114 physics from finance

which needs the minimum amount of time ∆tmin travels at

velocity c.

In other words, the constant c, which we introduced to get

the same units for all terms in Eq. 5.4, encodes the maxi-
mum velocity at which anything can travel in spacetime.
The constant c is usually called the speed of light because,
11
Take note that c is well, light travels at this maximum velocity.11 From ex-
a general constant periments we know that while c is incredibly large, it’s
which often appears
in contexts which not infinite.12 And this is historically how the correct local
have nothing to do structure of spacetime (Eq. 5.4) was discovered.
with light. The name
"speed of light" is
only used for historic The experimental fact that there is a maximum velocity
reasons. In general, which is valid for anyone and anything is the basis of Ein-
c is an upper speed
limit for everything in stein’s theory of special relativity.13
physics and all massless
particles travel at speed
c.

12
The experimental
value of c is
meters
To summarize, the main idea described in this section is
2.9979 × 108 .
second that there is an upper speed limit. This limit implies that
the interplay between time and space works a little dif-
13
Discussing special ferently from what we would expect. Naively, we would
relativity any further
expect that the distance between two objects is simply
here would lead us
too far astray. But for ∆s2 = ∆x2 + ∆y2 + . . . or ∆s2 = (c∆t)2 + ∆x2 + ∆y2 + . . .
the modest goal of this if we are including a coordinate with different units like
book, we already have
everything we need to t.14 However, if there is an upper speed limit, there is actu-
move forward. A few ally a relative minus sign between the temporal and spatial
more mathematical
coordinates: (∆s)2 = (c∆t)2 − (∆x )2 − (∆y)2 − . . ..15 By
details about the
structure of Minkowski including this relative minus sign, we are hard-coding the
space are discussed in upper speed limit into the geometry of space and time.
Appendix B.
14
In mathematical
We can visualize the structure of the resulting space as
terms, a space in which
this is true is called an follows. At each moment in time we imagine that light
Euclidean space. (or anything else moving with the maximum velocity c)
moves away from each particular location. The paths traced
15
The space in which
we calculate distances
this way is known as
Minkowksi space.
 special relativity and general relativity 115

this way define the boundaries of our arena. Every object

starting at this moment in time and at this location has to
remain within these boundaries. Nothing can move outside
of this so-called light cone because this would require that it
moves faster than c.

Since we get such a light cone for each possible location at

each moment in time, we get the following picture:

Before we move on and talk about the global structure of

spacetime and general relativity, there is one additional
thing worth mentioning. In the previous sections we noted
that it can be instructive to analyze a given space using its
symmetries. In particular, this is the case for isospin and
116 physics from finance

16
Mathematically,
the Poincaré group color space. The structure of these spaces can be best un-
P(1, 3) is the semidirect
product of the Lorentz derstood by noting that the symmetries of local isospin and
group SO(1, 3) and the local color space are SU (2) and SU (3) respectively. Since
translation group R1,3 :
Minkwoski space is also quite complicated, it also makes
P(1, 3) = R1,3 o SO(1, 3) . sense to study its symmetries to understand it better. One
The S denotes again type of symmetry of spacetime is certainly rotations. But
"special" which means
that all elements of Minkowski space is four-dimensional and also includes a
SO(1, 3) have deter- time-coordinate. This means there are also transformations
minant one. The O which mix spatial and temporal components. These kinds
denotes "orthogonal"
and means that all of transformations are commonly called boosts. In physical
SO(1, 3) elements fulfill terms, a boost is a transformation to a coordinate system
the condition O T O = 1.
Take note that rotations
moving at a constant velocity with respect to the initial ref-
in three-dimensional erence frame. This is a symmetry because there is no way to
Euclidean space R3 tell whether you are sitting still or moving with a constant
are described mathe-
matically by the group velocity. (For example, if you close your eyes and block out
SO(3). This interpre- all noise in a stable train moving at constant velocity.)
tation comes about be-
cause transformations
which fulfill the con- Another symmetry of spacetime is translations. A spa-
dition O T O = 1 leave tial translation is a shift to another location x → x + e
the length of vectors
unchanged. Moreover, and temporal translation is a shift to a different point in
transformations with time t → t + e. The total symmetry group of Minkowski
determinant one leave
the orientation of the
space therefore consists of rotations, boosts and translations.
coordinate system un- Mathematically, we call this symmetry group P(1, 3).16
changed. Therefore,
taken together, these
two conditions single Now, let’s move on and discuss why Minkowski space is not
out rotations from the enough to describe spacetime on a global scale.
set of all imaginable
transformations. Now,
we consider Minkowski
space R1,3 . "Rota-
tions" in Minkowski
space are denoted
analogously SO(1, 3).
Take note that trans-
lations in some space
are mathematically
equivalent to the space
itself. For example, the
group of translations in
Minkowski space (R1,3 )
is R1,3 . The Poincaré
group is discussed
in detail in my book
Physics from Symmetry.
 special relativity and general relativity 117

5.2 General Relativity

As mentioned at the beginning of this chapter, our goal is to

repeat for spacetime what we discussed previously for the
various internal spaces.

In the previous section, we discussed the local structure

of spacetime. This step is analogous to how we discussed
the local structure of the internal space in our financial toy
model (a line) or in quantum mechanics (a circle). With this
local structure at hand, we can follow exactly the same steps
as in the previous chapters.

First of all, we can argue that it should be possible to use

arbitrary local coordinate systems, analogously to how
we argued that it should be possible to use arbitrary local
currencies. For example, it shouldn’t make any difference
whether we use a rotated coordinate system at a later point
in time or in a different region of spacetime. Maybe you
think it’s a somewhat silly idea to use different coordinate
systems in different regions. But in some sense, this is ex-
actly what we do whenever we use curvilinear coordinates.
We need bookkeepers Γ as soon as we use curvilinear coor-
dinates, e.g., spherical coordinates. When we use spherical
coordinates the coordinate axes at different locations are
orientated differently. This means that the components of
a given vector change if we move it to a different location
even though the vector itself doesn’t change. Therefore, if
we want to compare two vectors at two different locations,
we can’t compare their components directly. Instead, we
need to move one vector to the location of the other vector.
To do this consistently, we need a connection which keeps
track of the local orientation of the coordinate axes through-
out space.
118 physics from finance

Once more, this is only possible if we introduce bookkeep-

ers Γ which keep track of all local conventions. But again,
this is not really essential if we can simply use the same
choice of coordinate system everywhere; unless, of course,
there is something which prevents this.

And indeed it turns out that this is not always possible,

analogous to how in the financial toy model it’s sometimes
impossible to define a global currency. In the financial toy
model, we discovered that whenever there is a currency
arbitrage opportunity there can’t be a global currency. A
global currency would mean that we can eliminate all ex-
change rates Aµ . But this in turn would imply that there is
no arbitrage opportunity Fµν .

Moreover, we argued that geometrically a currency arbitrage

opportunity corresponds to a non-zero curvature of our
internal money space.

One of the main ideas of general relativity is that exactly the

same is true for spacetime.
 special relativity and general relativity 119

If the curvature of spacetime is non-zero, it’s impossible

to use the same Minkowski coordinate system globally.
In other words, we can’t describe global spacetime using
a single Minkowski space, but it is possible to neatly glue
together multiples copies of local Minkowski spaces in a
non-trivial manner.17 17
The only exceptions
where this type of
construction doesn’t
work are regions
where the curvature of
spacetime gets infinite
(ex: singularity at
the center of a black
hole, or the big bang
singularity at the origin
of the universe).

This is analogous to how the total internal money space

consists of multiple lines which are glued together non-
trivially.

Before attempting to understand what all this has to do 18

This is discussed in
with gravity, we need to talk about curvature. The main detail in Appendix D.5.

idea is that when we move an object along a closed path in

a curved space, the final state of the object is not necessarily 19
Take note that this is
equivalent to its initial state.18 exactly how we defined
the gain factor Fµν
which indicates that
The difference between the initial and the final state encodes there is a currency
arbitrage opportunity.
how curved the space is.19
120 physics from finance

We use exactly the same idea to describe the curvature of

spacetime. Specifically this means that we move a vector
around a closed path and compare the final vector with the
initial vector. The difference between the two tells us how
much spacetime is curved in this region.

For example, we can imagine that a given vector Vα moves

from A to B via two different paths:
 special relativity and general relativity 121

Taken together these two paths yield a closed path and we

can calculate20 20
The dots here indicate
further terms which do
not describe the curva-
Vα ( A → C → B) − Vα ( A → D → B) = Rαβν µ V β δx µ δx ν + . . . ,
ture of spacetime, but
(5.8) torsion. Torsion is an-
ν
where Rαβ µ denotes the corresponding (Riemann) curvature other property through
which a given space
tensor21 can have a non-trivial
structure. (A space has
Rαβν µ = ∂α Γ βνµ − ∂ β Γανµ + Γανκ Γ βκµ − Γ βνκ Γακµ . (5.9) non-zero torsion if the
parallelogram that is
formed by two vectors
and Γακν are the corresponding connections.22
does not close.) Usually
this term is neglected
Now, what does all this have to do with gravity? because, so far, no tor-
sion of spacetime has
ever been measured in
experiments. This is
discussed a bit further
5.2.1 Gravity and Curvature Dynamics in Section 5.3. More-
over, if you wonder
why we suddenly have
superscript indices
The main idea in all previous chapters is that we can un-
here, have a look at
derstand how interactions happen if we allow the structure Appendix B.
of the space in question to be non-trivial. In the previous
section, we saw that the structure of spacetime can be non-
21
Since Minkowski
space is four-
trivial too. And if spacetime is curved, we need the connec- dimensional and has
tions Γ to move vectors around consistently. In other words, a non-trivial structure,
the explicit form of
if the curvature of spacetime is non-zero, the connections Γ the curvature tensor is
are essential parts of the system. quite complicated. But
these details are not
really important for the
Now, analogously to how we asked in the financial toy following discussion
model where currency arbitrage opportunities come from, and we therefore only
quote the final result.
we can now ask: where does a non-zero curvature of space-
time come from? 22
The connections
Γακν are usually called
Christoffel symbols.
The answer is: any object which carries non-zero energy-
momentum. Thus, curvature arises from every physical
object and in particular every elementary particle. And
in turn, every object which carries energy-momentum is
influenced by the structure of spacetime.23 23
This is analogous to
how objects carrying
electric charge modify
charge space and are
influenced by it.
122 physics from finance

This is how gravitational interactions happen.

This means that the "charge" associated with gravity is

energy-momentum. Therefore, every particle which carries
energy-momentum takes part in gravitational interactions.
Moreover, the curvature of spacetime Rαβν µ is what we call
24
All this may seem the gravitational field.24
quite intimidating.
But take note that
everything we do here The interplay between energy-momentum and curvature is
is completely analogous described by the Einstein equation. We can derive it by fol-
to what we discussed
already for the financial lowing the same steps that allowed us to derive Maxwell’s
toy model and quantum equations in Section 2.1.
mechanics. The only
difference is that we
are now talking about
spacetime and not some
internal space.

First of all, we need the fact that energy-momentum is con-

25
Using Noether’s served.25 Mathematically, we describe energy-momentum
theorem this can be using the energy-momentum tensor Tµν . The conservation
derived by using the
fact that spacetime of energy-momentum implies that there is a continuity
is homogeneous. A equation26
space is homogeneous
whenever it possesses ∂µ Tµν = 0. (5.10)
translation symmetry.
Moreover, we know already that curvature is described by
26
This is analogous to the curvature tensor Rαβν µ (Eq. 5.9).
Eq. 2.32 where, in the
financial toy model,
Attempting to connect these two equations, we observe that
copper was conserved.
Tµν has two indices while Rαβν µ has four. Therefore, we can’t
simply write
Tµν = const × Rαβν µ . (5.11)
In addition, we need to make sure, since the derivative of
the left-hand side is zero (Eq. 5.10), that the derivative of the
27
All steps here are right-hand side is zero, too.27
analogous to what we
discussed in Section 2.1.
What we therefore need is a quantity, constructed from
Rαβν µ , which has only two indices and yields zero if we take
 special relativity and general relativity 123
28
For electrodynamics
and our financial toy
model, this task was
the derivative. Deriving the correct mathematical object easier because the
which fulfils these conditions is a somewhat complicated derivative of the object
we use to describe
mathematical problem and we will therefore not dive into curvature (∂µ Fµν ) has
the details here. For our purpose, it is enough to know that exactly the properties
it can be done and that the correct result is known as the we need. In contrast,
the Einstein tensor can
Einstein tensor Gµν .28 be written in terms of
the curvature tensor
Rαβν µ as follows:
As soon as we have this tensor, we can write down the
Einstein equation 1
Gµν = Rµν − Rgµν ,
2
Tµν = CGµν . (5.12)
where
The constant C encodes how strongly spacetime reacts to Rµν = Rαµα ν
the presence of energy-momentum.29 R = gµν Rαµα ν ,

and gµν denotes the

Now we have two indices on both sides and taking the metric. The concept of
derivative yields zero a metric is discussed in
Appendix B.
∂µ Tµν = 0 = ∂µ Gµν . (5.13) 29
Specifically, C = 8πG
c4
where G is Newton’s
gravitational constant
and c the speed of light.

30
Reminder: this is
A final aspect that we need to talk about is that, once more, necessary because,
when we calculate
we need to replace our ordinary derivatives with covariant the derivative, we
derivatives whenever there is non-zero curvature.30 When compare the values
of a function at two
spacetime is curved, we need to replace all derivatives with different locations
the following covariant derivative31
f 0 ( a) =

Db v a ≡ ∂b v a + Γ a bc vc . (5.14) f ( a + h) − f ( a)
lim .
h →0 h

In other words, if we want any equation we derived so far In a curved space this
comparison is not as
to be valid in curved spacetime, we need to change trivial as in a flat space.
To move objects around
∂b → Db = ∂b + Γ a bc . (5.15) consistently, we need
to use the appropriate
connections.

By doing this, we take into account how particles react to 31

This is analogous to
the presence of curvature. In other words, while the Einstein what we discussed for
curved internal spaces,
e.g., around Eq. 2.55.
124 physics from finance

equation (Eq. 5.12) describes how spacetime reacts to the

presence of objects carrying energy-momentum, the replace-
ment rule in Eq. 5.15 allows us to describe how particles
react to the structure of spacetime:

Einstein equation
!
Elementary Particles Spacetime
b
covariant derivative

And, as mentioned before, this is how gravitational interac-

tions happen. Objects modify spacetime and are influenced
by it in turn.

5.3 Summary and a Few Loose Ends

The main point of this chapter is that we can describe grav-

ity in a manner analogous to how we describe all other fun-
damental interactions. The only difference is that now it’s
spacetime which gets curved and not some internal space.

One loose end which we need to talk about is the charge

32
A conserved quantity (energy-momentum) associated with gravitational interac-
is a quantity which tions. In general, we can understand the charge associated
does not change in time
with a given space by using Noether’s theorem. Noether’s
∂t Q = 0 → Q = const.
theorem tells us that for each symmetry of the space in
You can find a detailed question, we get a conserved quantity:32
discussion and math-
ematical derivation of
Noether’s theorem in
B The conserved charge following from the U (1) symmetry
Jakob Schwichten-
berg. Physics from Sym- of charge space is what we call electric charge.
metry. Springer, Cham,
Switzerland, 2018b. B The conserved charge following from the SU (2) symme-
ISBN 978-3319666303 try of isospin space is what we call isospin.
 special relativity and general relativity 125

B The conserved charge following from the SU (3) symme-

try of color space is what we call color.

B The conserved charge following from the translational

symmetry of Minkowski space is what we call energy-
momentum. In particular:
– The conserved charge following from invariance under
spatial translations x → x + e is momentum.
– The conserved charge following from invariance under
temporal translations t → t + e is energy.

A crucial point is that since spatial components can get

mixed with the time component, we need to consider them
as a single (overall) conserved charge which we call energy-
momentum.33 This allows us to keep track of exactly how 33
Transformations
which mix spatial com-
energy and momentum get mixed in different coordinate
ponents with the time
systems. For example, we can always achieve that the mo- component are what
mentum of a given particle is zero by switching to a co- we call boosts. A boost
is a transformation to
ordinate system which moves with (system attached to) a coordinate system
the particle. Mathematically, energy-momentum is a four- which moves with a
vector pµ = ( E/c, p1 , p2 , p3 ), where we have introduced the constant velocity with
respect to the initial
constant c once more to get the same units for all compo- coordinate system.
nents.34 Moreover, take note that the quantity we call mass
m corresponds to the length of the energy-momentum four
34
We did the same
thing when we defined
vector: pµ · pµ = E2 /c2 − p21 − p22 − p23 ≡ m2 c2 . the vector which
describes locations
in spacetime xµ =
Another interesting point is that there are additional sym- (ct, x1 , x2 , x3 ) such that
metries of Minkowski space. Therefore, we get additional
xµ · xµ =
conserved charges. In particular:35
c2 t2 − x12 − x22 − x32 .

B The conserved charge following from the rotational 35

Using Noether’s the-
symmetry of Minkowski space is what we call angular- orem for the invariance
under boosts does not
momentum. Upon closer inspection, this conserved quan- yield a useful con-
tity consists of two parts: orbital angular momentum servation law. This is
discussed explicitly, for
and spin. We can interpret spin as some kind of internal
example, in my book
Physics from Symmetry.
126 physics from finance

angular momentum.

This fact alone is not very surprising. But it becomes very

36
In fact, all known
interesting once we recall the roles of all charges we dis-
elementary particles
carry spin except for cussed so far.
the recently discovered
Higgs boson.
B Particles carrying electric charge change the structure of
37
Torsion was already
charge space.
mentioned in the text
B Particles carrying isospin change the structure of isospin
around Eq. 5.8. If we
calculate how a vector space.
changes if we move it
between two points via B Particles carrying color change the structure of color
two different paths, we space.
find that in general, the
answer consists of two B Particles carrying energy-momentum change the struc-
terms. The first term ture of spacetime.
contains information
about curvature, the
second term about
torsion. In other words,
What then is the role of angular momentum?
there are two possible
reasons why its final While one can argue that orbital angular momentum is
state and initial state
are different if we move somewhat different from the other charges since it involves
a vector along a closed the (arbitrary) choice of an origin in a coordinate system
path. One reason is
that the space we are about which rotational motion is measured, spin is a prop-
moving in is curved, erty which is carried by individual elementary particles. In
the second one is that other words, analogous to how some particles carry electric
the space has non-zero
torsion. charge, there are particles which carry spin.36

From the perspective described in this book it seems natural

38
"André Weil recalled that the presence of spin at some location has an effect on
that he was present at the geometry of the arena of physics, too.
a reception for Einstein
by Romain Roland; and
that at the reception he One possibility is that spin leads to a torsion of spacetime
was present during a
conversation between Eli while energy-momentum causes a curvature.37 In Ein-
Cartan and Einstein when stein’s general relativity we simply assume that the torsion
Cartan asked Einstein of spacetime is zero. The model we end up with if we re-
why he had not allowed
for torsion in his theory; lax this assumption is known as Einstein-Cartan gravity.38
and André felt even then
that Einstein did no really
understand what Cartan
was talking about!"
[Chandrasekhar, 2010].
 special relativity and general relativity 127

However, so far, this has never been confirmed experimen-

tally. In other words, a non-zero torsion of spacetime has
never been measured.39 39
This does not mean
that Einstein-Cartan
gravity is wrong.
Maybe the effect of
spin on the geometry
of spacetime is simply
too small to be measur-
able with present day
technology. It is worth
mentioning in this
context that the effect
of individual elemen-
tary particles on the
curvature of spacetime
has never been mea-
sured either because
the energy-momentum
they carry is simply
too small. But the effect
of huge macroscopic
objects on the geometry
of spacetime has, of
course, been observed
in experiments.
6

Closing Words

If at this point you have lots of questions, I think that this

book has fulfilled its purpose.

How do we use Maxwell’s equations in electrodynamics?

How can we use Schrödinger’s equation to describe quan-
tum systems? How do we calculate the probability that
certain scattering processes happen? How can we solve
Einstein’s equation and use it to describe our universe?

Of course, the answer to any of these questions is far be-

yond the scope of this book. But I think you are now opti-
mally prepared to dive deeper.

Before I recommend some of my favorite books which deal

specifically with these questions, I want to put things into
context. Hopefully, this will allow you to see clearly which
theory you should focus on if you want to answer a specific
question.
130 physics from finance

Our theories make up, in some sense, a (2 × 2) matrix.

There are theories in which we focus on fields and there are
1
Take note that it’s theories in which particles are our main focus.1 In addition,
possible to describe we can describe both particles and fields in either a classical
particles in quantum
field theory, but they or quantum framework. This yields, in total, four different
are understood as combinations:
excitations of the
underlying quantum
fields. Classical Quantum
Particles Classical Mechanics Quantum Mechanics
Fields Classical Field Theory Quantum Field Theory

It’s crucial to understand that no single framework is the

best. Each of them is useful and which one we use always
depends on the system we want to describe.

B Field theories are great for describing the fundamental

interactions themselves and systems in which the particle
number is not conserved.

B In contrast, particle theories are great for describing indi-

vidual objects.

B Classical theories are still the best tools we have to de-

scribe nature at macroscopic scales.

B But we need quantum theories to describe the behavior of

microscopic systems.

To understand this a bit better, let’s talk about concrete

2
As mentioned above examples:2
already, there are no
macroscopic models
describing the weak
field (isospin space B We describe the dynamics of the electromagnetic field
curvature) and the (charge space curvature Fµν ) and gravitational field
strong field (color space
(spacetime curvature Rαβν µ ) on macroscopic scales using
curvature) because the
associated interactions classical field theory. The resulting models are known as
are extremely short- classical electrodynamics and general relativity.
ranged.
 closing words 131

B We describe the effects of interaction fields like Fµν on

individual macroscopic objects using classical mechanics.
B We describe the effects of interaction fields like Fµν on
individual microscopic objects using quantum mechanics.
B We describe the dynamics of interaction fields on micro-
scopic scales using quantum field theory. For example, the
dynamics of the electromagnetic field Fµν on microscopic
scales is described by quantum electrodynamics and the
dynamics of the strong field Gµν is described by quantum
chromodynamics.3 3
Take note that so
far, there is no gener-
ally accepted model
which describes the
An important aspect of quantum field theory is that we can
gravitational field on
also use it to describe the effects of the interaction fields on microscopic scales,
elementary particles. Therefore, we can use it directly to i.e. gravitational in-
teractions between
describe how elementary particles interact. In contrast, we elementary particles.
need both classical field theory and classical mechanics to This is known as the
problem of quantum
describe electromagnetic interactions between macroscopic gravity and we will talk
objects for example. In particular, we need classical field a bit more about this
theory to describe the electromagnetic field and classical below.

mechanics to describe its effect on macroscopic objects.

But this does not mean that quantum field theory renders
all other frameworks obsolete. For example, we can’t use
quantum field theory to describe gravitational interactions
between planets. Therefore, we really need all frameworks
to describe nature at all scales.

Now, what are some good sources to learn more about these
theories?
132 physics from finance

6.1 Reading Recommendations

Great books to learn the basic ideas of classical mechanics

are

4
Jennifer Coopersmith. B The Lazy Universe by Jennifer Coopersmith4 .
The lazy universe.
Oxford University
Press, 2017. ISBN B Introduction to Classical Mechanics by David Morin5 .
9780198743040

5
David Morin. Introduc- Moreover, you might enjoy my book
tion to classical mechan-
ics: with problems and
solutions. Cambridge
University Press, 2008. B No-Nonsense Classical Mechanics6 .
ISBN 9780511808951

6
Jakob Schwichten- Instead of learning classical field theory in general, most
berg. No-Nonsense students learn classical electrodynamics and general rela-
Classical Mechanics. No-
Nonsense Books, 2019.
tivity directly. My favorite student-friendly electrodynamics
ISBN 978-1096195382 textbooks are

7
Richard Feynman. B Volume 2 of the Feynman Lectures7
The Feynman lectures
on physics. Addison-
Wesley, San Francisco, B A student’s guide to Maxwell’s equations by Daniel A.
Calif. Harlow, 2011.
ISBN 9780805390650 Fleisch8

8
Daniel Fleisch. A stu-
dent’s guide to Maxwell’s B Introduction to Electrodynamics by David J. Griffiths9
equations. Cambridge
University Press, 2008.
ISBN 978-0521701471 You also might enjoy my book
9
David Griffiths. Intro-
duction to electrodynam- B No-Nonsense Electrodynamics10 .
ics. Pearson Education
Limited, Harlow, 2014.
ISBN 9781292021423

10
Jakob Schwichten-
berg. No-Nonsense
Electrodynamics. No-
Nonsense Books, 2018a.
ISBN 978-1790842117
 closing words 133

Great books to start learning more about Einstein’s theory

of general relativity are

B Relativity, Gravitation and Cosmology by Ta-Pei Cheng11 . 11

Ta-Pei Cheng. Rela-
tivity, Gravitation and
Cosmology: A Basic
B Einstein Gravity in a Nutshell Anthony Zee12 . Introduction. Oxford
University Press, 2nd
edition, 1 2010. ISBN
9780199573646
Two illuminating books to learn how quantum mechanics
works in practice are 12
Anthony Zee. Einstein
Gravity in a Nutshell.
Princeton Univer-
B Lectures on Physics Volume 3 by Richard Feynman13 . sity Press, 1st edi-
tion, 5 2013. ISBN
9780691145587
B Quantum Mechanics by David J. Griffiths14 .
13
Richard Feynman.
The Feynman lectures
In addition, since most textbooks make quantum mechanics on physics. Addison-
Wesley, San Francisco,
appear much more complicated than it really is, I decided to Calif. Harlow, 2011.
write one called ISBN 9780805390650

14
David Griffiths.
B No-Nonsense Quantum Mechanics15 . Introduction to quantum
mechanics. Pearson
Prentice Hall, 2005.
ISBN 9780131118928
The best book to get started with quantum field theory is
15
Jakob Schwichten-
berg. No-Nonsense
B Student Friendly Quantum Field Theory by Robert D. Quantum Mechan-
Klauber16 . ics. No-Nonsense
Books, 2018c. ISBN
978-1719838719

16
Robert D. Klauber.
Student Friendly Quan-
tum Field Theory.
Sandtrove Press, 2nd
edition, 12 2013. ISBN
9780984513956
134 physics from finance

Moreover, the keyword you should look out for if you want
to learn more about the geometrical perspective described in
17
For a few more this book is "fiber bundle".17
details on fiber bundles,
see Appendix D.
Good starting points to learn more about fiber bundles are

B Fiber Bundles and Quantum Theory by Herbert J. Bern-

18
H. J. Bernstein and stein and Anthony V. Phillips18 .
A. V. Phillips. Fiber
Bundles and Quantum
Theory. Sci. Am., 245: B Gauge fields, knots, and gravity by John Baez and Javier
94–109, 1981
P. Muniain19 .
19
John Baez and
Javier P. Muniain.
Gauge fields, knots, and B The Road to Reality by Roger Penrose20 .
gravity. World Scien-
tific, Singapore River
Edge, N.J, 1994. ISBN A great mathematics textbook on fiber bundles and how
9789810217297 they are used in physics is
20
Roger Penrose. The
road to reality: a complete
B Topology, Geometry, and Gauge Fields: Foundations by
guide to the laws of the
universe. Vintage Books, Gregory Naber21 .
New York, 2007. ISBN
9780679776314
A particularly illuminating example of a non-trivial fiber
21
Gregory Naber. bundle is the Hopf bundle. A great discussion of the Hopf
Topology, geometry and
Gauge fields: founda- bundle and why it is important in physics can be found in
tions. Springer Sci-
ence+Business Media,
LLC, New York, 2011. B Some Elementary Gauge Theory Concepts by Hong-Mo
ISBN 9781441972538
Chan, Sheung Tsun Tsou22 .
22
Chan Hong-Mo and
Tsou Sheung Tsun.
Some elementary gauge
theory concepts. World
Scientific, Singapore
River Edge, N.J, 1993.
ISBN 9789810210809
 closing words 135

Last but not least, a fantastic book to learn more about how
fiber bundles can be used to describe the financial market is

B The Physics of Finance by Kirill Ilinski23 . 23

Kirill Ilinski. Physics
of finance: gauge mod-
elling in non-equilibrium
In contrast to the modest goal of this book, Ilinski tries to pricing. John Wiley,
Chichester England
construct a realistic model of the financial market. New York, 2001. ISBN
9780471877387
136 physics from finance

6.2 Outlook

The main idea described in this book is that we can un-

derstand all fundamental interactions beautifully from a
common perspective. Therefore, it may seem as if physics
were almost a finished project. But this couldn’t be further
from the truth.

24
Several open ques- There are still lots of open questions.24
tions have already been
mentioned in passing.
Most importantly, there is still no generally accepted way to
incorporate general relativity in the framework of quantum
field theory. While Einstein’s classical field model works
perfectly at large scales, we know that at small scales quan-
25
Take note that as tum field theory holds sway.25 But there are lots of other
long as we stick to more modest open questions, for example, about the role of
low energies, we can
incorporate General torsion in general relativity.26
Relativity as an effective
quantum field model.
But we know that this Moreover, there are lots of open questions about the par-
theory does not remain ticle zoo. Why exactly these particles? Why do they carry
valid at high-energy the charges they do? In addition, there are good reasons to
scales. In other words,
we know that this believe that there are additional elementary particles, for ex-
effective model must ample, to explain the masses of neutrinos and the observed
be replaced with a
better model at high dark matter density.
energies. (That’s the
whole point of effective I personally think there is still much to come and maybe
models. But from a
modern perspective, a completely new framework is needed to overcome the
all the models we have present obstacles. Future developments will be extremely
come up with so far are
effective models which interesting and I hope you will continue to follow the story
are only valid up to and maybe contribute something yourself.
some cutoff scale.)

26
This problem was
mentioned already
at the end of the last
chapter.
One Last Thing

It’s impossible to overstate how important reviews are

for an author. Most book sales, at least for books without
a marketing budget, come from people who find books
through the recommendations on Amazon. Your review
helps Amazon figure out what types of people would like
my book and makes sure it’s shown in the recommended
products.

I’d never ask anyone to rate my book higher than they think
it deserves, but if you like my book, please take the time
to write a short review and rate it on Amazon. This is the
biggest thing you can do to support me as a writer.

Each review has an impact on how many people will read

my book and, of course, I’m always happy to learn about
what people think about my writing.

PS: If you write a review, I would appreciate a short email

with a link to it or a screenshot to [email protected].
This helps me to take note of new reviews. And, of course,
feel free to add any comments or feedback that you don’t
want to share publicly.
138 physics from finance
 Part V
Appendices
A

Statistics

A.1 Probability Distribution

A probability distribution tells us the probability of each

possible outcome. Formulated differently, a probability
distribution eats a possible outcome o and spits out the
corresponding probability χ(o ).

For example, the probability distribution for a regular die

is1 1
An important property
is that the probabilities
χ(1) = 1/6 add to 1 = 100 %. This
is necessarily the case
χ(2) = 1/6 because one of the
outcomes must happen
χ(3) = 1/6 and a total probability
χ(4) = 1/6 of more than 100%
doesn’t make sense.
χ(5) = 1/6
χ(6) = 1/6.

But we can also imagine that we are dealing with an oddly

142 physics from finance

2
Again, take note that shaped die. Thus, our probability could look as follows2
the probabilities add to
1. χ(1) = 1/2
χ(2) = 1/6
χ(3) = 1/12
χ(4) = 1/12
χ(5) = 1/12
χ(6) = 1/12.
In words, this means that with this die, we are much more
likely to get a 1 than, say, a 6.

A.2 Mean

One of the simplest but at the same time most important

3
Alternative names for statistical tools is the so-called mean3 .
the mean are: expected
value, expectation value
or simply the average. The mean of a quantity is the average value we obtain when
we repeat a given experiment many times. We use it when-
4
Think: flipping a coin, ever we are forced to make probabilistic predictions.4
or tossing a die.

For example, imagine the following situation: A friend

offers to play a game. She flips a coin. If it lands on tails,
she pays you $1.5. But if it lands on heads, you have to pay
her $1. By calculating the expectation value for the outcome,
you can decide whether or not you should play this game.

B The probability that a coin lands on heads is p1 = 50%.

In this case, the outcome for you is: x1 = −$1
B Equally, the probability that a coin lands on tails is p2 =
50%. In this case the outcome for you is: x2 = +$1.5

The mean is defined as the sum over each outcome times

 statistics 143

the probability of the outcome

mean = ∑ xi Pi = x1 P1 + x2 P2
i
= −$1 × 50% + $1.5 × 50% = $0.25 . (A.1)

So, if you play this game many times, you will make a
profit. On average, you make $0.25 per game. For exam-
ple, let’s say you play the game only twice: you win the first
time and lose the second time. For the win in the first game,
you get $1.5. For the loss in the second game, you lose $1.
In total, you therefore win $0.5 in two games. This equals
$0.25 per game.

Of course, the system here is so simple that we could have

guessed this without calculating anything. But for more
complicated systems the situation quickly gets messy.

To summarize: the mean is the average of the outcomes if we

play the game many, many times. A common mathematical
notation for the mean of a quantity x looks like this: h x i.

A.3 Standard Deviation

There is one other statistical notion that we need whenever

we describe a system in probabilistic terms. It gives us
information on how much the system deviates from the
most probable outcome on average. In other words, this
statistical quantity gives an indication of how much our
measurements are spread out. If it is zero, the outcome is
the same for every measurement; we measure the same
value all the time. If it is non-zero, it’s possible to measure
different values.

The statistical quantity I’m talking about is called the stan-

144 physics from finance

dard deviation. It tells us how much our measurements

(on average) get spread around the average outcome ( = the
5
This strange looking expectation value):5
definition will make q
sense in a moment.
And an aside: The same
∆x = h x2 i − h x i2 . (A.2)
definition without the
square root is known as The second term under the square root is simply the ex-
variance. pectation value squared. For the first term, we calculate an
expectation value again, but this time we square each pos-
sible outcome before we weigh it with the corresponding
probability
6
For example, for h x2 i = ∑ xi2 Pi . (A.3)
x = {3, −5, 9} we have i

∑ xi2 Compare this with

i
!2
= (3)2 + (−5)2 + (9)2
= 115
2
hxi = ∑ xi Pi . (A.4)
i
whereas
 2 Yes! It makes a difference whether we square the weighted
∑ xi sum of the outcomes or we sum the weighted squares of the
i

= (3 − 5 + 9)2 outcomes!6 To see concretely the meaning of the standard

= 49 . deviation, let’s consider the example from above again. The
standard deviation reads

q
∆x = h x 2 i − h x i2
y

Eq. A.3 and Eq. A.4

v
u !2
= t∑ xi2 Pi − ∑ xi Pi
u

i i
y

Eq. A.1
r
= ∑ xi2 Pi − ($0.25) 2

i
y

q
= ((−$1)2 × 0.5 + ($1.5)2 × 0.5) − ($0.25)2
y

√
= $ 1.625 − 0.0625 ≈ $1.25.
 statistics 145

There are two important things to take note of:

B The result isn’t zero. It makes a difference whether we

square each term in the sum or the complete sum. We
can see this explicitly here. Otherwise, the standard devi-
ation would always be zero.
B If we only knew that the expectation value of some
game is $0.25 we would be missing a lot of informa-
tion. Naively, we could think that, since $0.25 is such a
small amount of money, the bets involved are not high at
all. However, imagine that we make our example more
extreme by proposing that if the coin lands on heads
you have to pay $10000, and if it lands on tails you get
$10000.5. The expectation value is again h x i = 0.25. But
the game is much more extreme now. I wouldn’t like to
play it. While the odds are still in your favor, now there
is also a high risk for you to lose a lot of money. This
kind of information is encoded in the standard devia-
tion. For this modified game the standard deviation is
∆x ≈ $10000.25. We can therefore see immediately that a
lot of money is at stake.

One final comment on the idea behind the definition of the

standard deviation: The logic behind squaring each possible
outcome in the first term on the right-hand side is to avoid
that terms cancel because they have opposite signs. This is
exactly what happens when we calculate the expectation
value. In the coin example, if the outcome was heads we
had to pay $1, which mathematically means x1 = −$1.
Thus, in the formula for the mean, the two terms almost
canceled since they have different signs. But now we want
to get information about the spread in our measurements.
Hence, we need a notion that takes the absolute size of the
outcomes into account. We accomplish this by squaring each
outcome and only then multiplying each term calculated
146 physics from finance

this way with the corresponding probability. We then also

square the expectation value such that it has the same units
as the thing we just calculated. Otherwise, we would be
7
In the example $2 with comparing apples with pears.7
$.
B

Metrics

In this appendix, we talk about Eq. 5.4 a bit more systematically.

In general, the mathematical tool which allows us to cal-

culate the distance between two points in a given space is
called the metric. In general, a metric encodes important
information about the structure of the space.

For example, the mathematical relations to calculate the

distance between two points A and B on a Euclidean plane
and between two points C and D on a sphere are differ-
ent. In the first case, the distance is given by the familiar
Pythagorean formula, irrespective of the actual distance
between the points, while in the second case, that formula
is only (approximately) valid for points which are close to
each other on the surface of the sphere. It is the metric that
encodes how distance between points can be calculated in a
given space.
148 physics from finance

The best way to understand what a metric is, is to consider

concrete examples. We can calculate the distance between
two points
   
x0 x̃0
x   x̃ 
A =  1 , B =  1 (B.1)
   
 x2   x̃2 
x3 x̃3
by calculating the length of the vector
∆x0
   
x̃0 − x0
∆x   x̃ − x 
~v ≡  1  =  1 1
(B.2)
  
∆x2   x̃2 − x2 


∆x3 x̃3 − x3
which connects them. In a four-dimensional Euclidean
1
Technically, we are space, we find1
calculating the scalar
product of the vector d( A, B) = ~v · ~v = ∆x02 + ∆x12 + ∆x22 + ∆x32 . (B.3)
with itself.
But in Minkowski space, we find

d( A, B) = ~v · ~v = ∆x02 − ∆x12 − ∆x22 − ∆x32 . (B.4)

We describe this fact by saying that a Euclidean and a

Minkowski space have a different metric g and write the
scalar product as ~v T g~v.

Now, for  
1 0 0 0
0 1 0 0
gE =  (B.5)
 

0 0 1 0
0 0 0 1
 metrics 149

we find the correct scalar product of a Euclidean space

∆x0
  
1 0 0 0
  0 1 0 0 ∆x 
  1
~v T gE~v = ∆x0 ∆x1 ∆x2 ∆x3 

0 0 1 0 ∆x2 
 

0 0 0 1 ∆x3

y
= ∆x02 + ∆x12 + ∆x22 + ∆x32 . (B.6)

And for   2
This is equivalent to
1 0 0 0 Eq. 5.4. The only dif-
ference is that now we
0 −1 0 0 
gM = (B.7) use a different vector
 
−1

0 0 0  and don’t interpret the
0 0 0 −1 zeroth component as
time.
we find the correct scalar product of Minkowski space2

∆x0
  
1 0 0 0
 ∆x1 
  0 −1 0 0   
~v T gM~v = ∆x0 ∆x1 ∆x2 ∆x3 

0 0 −1 0  ∆x2 
 

0 0 0 −1 ∆x3 y

= ∆x02 − ∆x12 − ∆x22 − ∆x32 . (B.8)

We call gE the Euclidean metric and gM the Minkowski

metric.3 3
Take note that it is
conventional to use
the symbol η for the
In index notation, we can write the scalar product in Minkowski metric.
Minkowski space as:

(∆s)2 = vµ η µν vν
∆x0
  
1 0 0 0
−1  ∆x1 
  0 0 0   
= ∆x0 ∆x1 ∆x2 ∆x3 

−1 0  ∆x2 
 
0 0
0 0 0 −1 ∆x3
= ∆x02 − ∆x12 − ∆x22 − ∆x32 (B.9)

Moreover, it is conventional to introduce superscript indices

150 physics from finance

in order to avoid writing the Minkowski metric all the time:

x µ ≡ η µν xν (B.10)

or equally

yν ≡ η νµ yµ

y
the Minkowski metric is symmetric η µν = η νµ
µν
= η yµ . (B.11)

This allows us to write the scalar product as follows:

x · y ≡ xµ η µν yν = xµ yµ = x ν yν . (B.12)

The bottom line is that when you see a superscript index in

physics, you should remember that this is usually a short-
hand notation for the Minkowski metric.
C

Group Theory

In Section 1.1, we talked about the general definition of a

symmetry. Intuitively, you can imagine that someone holds
the object in question in front of you. If you close your eyes
while the person performs a transformation and afterwards
you can’t tell if the person has changed anything at all, the
transformation is a symmetry of the object.

The branch of mathematics which deals with symmetries

is called group theory. A group is a set of transformations
which fulfill special rules plus an operation that tells us how
to combine the transformations.1 1
The rules are known
as group axioms and
they can be moti-
vated by investigating
an intuitive symme-
try like rotational
symmetry, see www.
jakobschwichtenberg.
com/group-axioms
A square consists of an infinity of points. A symmetry of
the square is a transformation which maps these points onto
themselves.
152 physics from finance

Some rotations about the origin (not all) are symmetries

of a square. This means that certain rotations map each
point that our square consists of onto another point on the
square. Therefore, we can say that the set of points our
square consists of is invariant under such a transformation.

To understand this, let’s focus on one particular corner of

the square. If we rotate the square by 50◦ , we end up with a
point outside the original set.

But if we rotate the square by 90◦ we find that the corner

point gets mapped to a location where another corner point
was previously located.

Therefore, rotations by 90◦ , 180◦ , 270◦ and of course 0◦ are

2
Take note that four symmetries of a square, while all other rotations aren’t.2 All
additional symmetries these rotations taken together plus a rule which allows us to
of a square are mirror
symmetries; namely combine rotations are the symmetry group of the square.
with respect to the
horizontal, vertical, and
(two) diagonal axes.
 group theory 153

Since the transformation parameter (rotation angle) can only

take on a discrete set of values, we say the group is discrete.

The situation is somewhat different when we investigate

transformations that leave a circle unchanged. While a circle
also consists of lots of points, there are many more transfor-
mations which map these points onto themselves.

Concretely, a circle is invariant under all rotations about the

origin. Again, our symmetry group consists of transforma-
tions and an additional rule which allows us to combine the
group elements in such a way that we get another group
element. For example, a rotation by 89.2◦ plus a rotation by
90◦ , yields a rotation by 179.2◦ which is another symmetry
of the circle.
154 physics from finance

Since the transformation parameter (rotation angle) can

take on arbitrary continuous values, we say the group is
continuous.

The main idea of group theory is to distill the defining

features of such symmetries into a precise mathematical
form. This yields the axioms of group theory. Afterwards,
all kinds of structures which fulfill these axioms can be
investigated. This is a useful approach because there exist
more abstract symmetries (e.g. of equations) and group
theory allows us to apply the basic lessons we learned by
studying intuitive examples.

C.1 Examples

In the previous chapters, we learned that we can make sense

of the internal space associated with a given theory by using
symmetries. In our financial toy model the symmetry group
is the dilation group which consists of all possible dilations

f = ee , with e ∈ R. (C.1)

of the given currency. The mathematical name for this

group is GL+ (1, R), the one-dimensional real general lin-
ear group with positive determinant.
 group theory 155

In quantum mechanics the symmetry group is the unitary

group of dimension one, denoted by U (1), and consists of
all possible phase shifts

f = eie , with e ∈ R. (C.2)

The term unitary refers to the fact that these phase shifts
fulfill the condition (eie )? eie = 1. The difference between
GL+ (1, R) and U (1) is the factor i in the exponent of the
transformation operators.

An extremely interesting aspect of groups is that we can

understand them geometrically.
156 physics from finance

C.2 Geometry of Symmetries

First of all, let’s recall two observations that we made in

previous chapters.

B In Section 1.2, we noticed that our financial market has

a symmetry because we can arbitrarily shift all the lo-
cal prices. In mathematical terms, this means that we
can multiply all prices by any number without actually
3
However, we need to changing anything to the dynamics of the system.3 The
multiply all prices with set of symmetry transformations therefore consists of all
the same number.
possible multiplications with real numbers. Geometri-
cally, these numbers yield a line.
4
To be more precise:
Lie groups are mathe-
matically differentiable B Moreover, we learned that the dynamics of quantum
manifolds. In the fol-
systems is unaffected if we multiply the wave function ψ
lowing, whenever the
word "group" is used, by a unit complex number eie . Mathematically, these lie
I mean a Lie group. on the unit circle in the complex plane.
Lie groups describe
continuous symmetries,
e.g. the symmetries of
a circle. Continuous These are examples of a beautiful and deep insight: we can
symmetries can be understand symmetry transformations geometrically.4
parameterized by a
continuous parameter
like the angle of rota- Geometrically, the dilation group is a line.
tion θ. There are also
discrete groups which
describe, for example,
the symmetries of a
square. (Only rotations
by 0◦ , 90◦ , 180◦ , 270◦
are symmetries of a
square and therefore,
these transformations
can not be parameter-
ized by a continuous
parameter.)
 group theory 157

The group U (1) is geometrically a circle.5 5

Recall that this follows
from Euler’s formula

z = eiφ
= cos(φ) + i sin(φ)
≡ Re(z) + iIm(z).

Concretely, this means

π
ei 2 =
cos(π/2) + i sin(π/2)
= 0+1·i = i
eiπ =
In the following appendix (Appendix D) we will use this cos(π ) + i sin(π )
idea to understand the arenas we use to describe our sys- = −1 + 0 · i = −1
tems in geometrical terms. 3π
ei 2 =
cos(3π/2) + i sin(3π/2)
= 0 − 1 · i = −i
C.3 Lie Algebras ei2π =
cos(2π ) + i sin(2π )
= 1+0·i = 1
There is one property that makes continuous symmetries Moreover, take note
(Lie groups) especially nice to deal with: they have elements that
π π π π π
which are arbitrarily close to the identity transformation6 . A ei 4 ei 4 = ei 4 + i 4 = ei 2
discrete group has no element that is arbitrarily close to the 6
The identity trans-
identity. formation is the trans-
formation that does
nothing at all. For ex-
Take, for example, the symmetries of a square. The set of ample, for rotations, the
transformations that leaves the square invariant consists of identity transformation
is a rotation by 0◦ .
four rotations: a rotation by 0◦ , by 90◦ , by 180◦ and one by
270◦ , plus some mirror symmetries. A rotation by 0,000001◦ ,
which is very close to the identity transformation (a rotation
by 0◦ ), is not a symmetry.
158 physics from finance

Next, take a look at the symmetries of a circle. Certainly, a

rotation by 0,000001◦ is a symmetry of the circle. Mathemat-
ically, we write an element g close to the identity I like this:

g(e) = I + eG (C.3)

where e (epsilon) is, as usual in mathematics, some really,

really small number and G is an object, called a generator,
which we will talk about in a moment.

Such small transformations barely change anything when

acting on an object. In the smallest possible case, such trans-
formations are called infinitesimal transformations. How-
ever, when we repeat such an infinitesimal transformation
many times, we end up with a finite transformation. Think
about rotations: Many small rotations in one direction are
equivalent to one large rotation in the same direction.

Mathematically, we can write the idea of repeating a small

transformation many times as follows

h(θ ) = ( I + eG )( I + eG )( I + eG )... = ( I + eG )k , (C.4)

where k denotes how often we repeat the small transforma-

tion.

If θ denotes some finite transformation parameter, e.g. 50◦

or so, and N is some huge number that makes sure we are
close to the identity, we can write Eq. (C.3) as
 
θ θ
g(e) = g = I + G. (C.5)
N N

The transformations we want to consider are the smallest

possible, which means N must be the largest possible num-
ber, i.e., N → ∞. To get a finite transformation from such
an infinitesimal transformation, one has to repeat the in-
finitesimal transformation infinitely often. Mathematically
 group theory 159

θ
h(θ ) = lim ( I + G) N , (C.6)
N →∞ N
which, in the limit N → ∞, is the exponential function7 7
This is one possible
expression of the
θ exponential function.
h(θ ) = lim ( I + G ) N = eθG . (C.7) We derive another
N →∞ N
(equivalent) expression
The bottom line is that the object G generates the finite trans- in terms of an infinite
series in Appendix E.
formation h. This is why we call such objects generators.

While a Lie group consist of group elements, we say that

the corresponding generators live in a Lie algebra which
belongs to the group. A main feature of the Lie algebra
is that we need a different rule to combine its elements.
While the group elements can be combined using the group
product (e.g. matrix multiplication), the Lie algebra product
is somewhat more complicated and involves a so-called Lie
bracket. The most famous Lie bracket is the commutator

[ A, B] = AB − BA .

So the natural product of two generators reads

A ◦ B = AB − BA ≡ [ A, B] .

We need such a weird looking product rule because neither 8

You can find a proof
the product AB nor BA are necessarily generators of the of this fact in John
Stillwell. Naive Lie
group in question. However, AB − BA is always another Theory. Springer, 1st
generator and therefore exactly what we need.8 edition, August 2008.
ISBN 978-0387782140

We can understand this by using the explicit product of two

group elements and by rewriting it using the corresponding
generators: 9
We will not talk
about the proof of
1 1 1
g ◦ h = eX +Y + 2 [ X,Y ]+ 12 [ X,{z
[ X,Y ]]− 12 [Y,[ X,Y ]]+...
(C.8) this formula here.
|{z} |{z} | You can find proofs in
}
∈G ∈G ∈G
most books about Lie
theory, for example in
To get the formula on the right-hand side, we use the fa- William Fulton and Joe
mous Baker-Campbell-Hausdorff formula.9 Harris. Representation
Theory: A First Course.
Springer, 1st corrected
edition, 8 1999. ISBN
9780387974958
160 physics from finance

The main point is that the product of two group elements

yields another group element. Therefore, the Baker-
Campbell-Hausdorff formula tells us how we have to com-
bine generators to get other generators which successfully
yield elements of the group when put into the exponential
function. And this is why we end up with a product rule
like A ◦ B = AB − BA.

Now, what do these generators explicitly look like?

Let’s consider a function f ( x, t) and assume that our goal is

to generate a spatial translation T f ( x, t) = f ( x + a, t). The
generator
10
Here ∂ x is a short- Gxtrans = ∂ x (C.9)
hand notation for the
∂
derivative ∂x .
does the job:10

a2 2
e aGxtrans f ( x, t) = (1 + aGxtrans + G + . . .) f ( x, t)
2 xtrans
y

Eq. C.9
a2
= (1 + a∂ x + ∂2x + . . .) f ( x, t)
2
y

= f ( x + a, t)

Here we used the series expansion of ex = ∑∞ x n

n=0 n! and in
the last step, we used that in the second to last line we have
11
If you’re unfamiliar exactly the Taylor expansion of f ( x + a, t).11 Alternatively
with the Taylor series consider an infinitesimal translation: a → e with e  1. We
expansion, have a look
at Appendix E.
 group theory 161

then have
e2 2
eeGxtrans f ( x, t) = (1 + eGxtrans + G + . . .) f ( x, t)
2 xtrans
e2 = 0

y
= (1 + eGxtrans ) f ( x, t)

y
Eq. C.9
= (1 + e∂ x ) f ( x, t)

y
= f ( x, t) + e∂ x f ( x, t)

y
= f ( x + e, t) .
Here ∂ x f ( x, t) is the rate of change of f ( x, t) in the x-
direction. In other words, this term tells us how much
f ( x, t) changes when we move in the x-direction. If we
multiply this rate of change with the distance that we
move in the x-direction - here e - we end up with the total
change of f ( x, t) if we move by e in the x-direction. Thus,
f ( x, t) + e∂ x f ( x, t) really is the value of f at the location
x + e.

The bottom line is: Gxtrans = ∂ x generates spatial transla-

tions.

Completely analogously Gttrans = ∂t generates temporal

translations:
f ( x, t) → e aGttrans f ( x, t) = f ( x, t + a) .
What we have learned here is that generators are the crucial
mathematical ingredient that we need to describe continu-
ous symmetries. We can describe any continuous symmetry
by acting on the function in question with the correspond-
ing generator many many times.

PS: If you’re interested in how we can understand a Lie algebra

as the tangent space to the identity element of the correspond-
ing group manifold, try http://jakobschwichtenberg.com/
lie-algebra-able-describe-group/
D

Fiber Bundles

In this appendix, as before, no attempt at mathematical rigor is

made. There are already hundreds of books and lecture notes which
rigorously present fiber bundles by stringing together definitions,
theorems and proofs. In contrast, the sole goal of this appendix
is to connect the ideas discussed in the previous chapters with
the correct mathematical notions. In other words, this appendix
is a bridge between our physical intuition and the mathematical
literature.

Fiber bundles are interesting because they are precisely the

mathematical constructions we used in the financial toy
model and in quantum mechanics.

To understand fiber bundles, we need to understand how

we can calculate the product of two spaces. Intuitively, by
calculating the product of two spaces, we add a copy of one
entire space to each point of a second space.

For example, the product of two lines yields a rectangle;

164 physics from finance

more precisely, a rectangular surface.

And the product of two circles yields a torus; more specifi-

1
Take note that it cally, the surface of a torus.1
doesn’t matter whether
we imagine that we
attach copies of the
space A to each point
of the space B or that
we attach a copy of B to
each point in A.

In the financial toy model, we multiply spacetime with

our money space. This means we attach a copy of money
space to each point of spacetime. In quantum mechanics,
2
We discuss spacetime we multiply spacetime with charge space.2 This means we
in detail in Chapter 5. attach a copy of charge space to each point of spacetime.

Since spacetime is four-dimensional and our money space

and charge space are one-dimensional, we can’t really draw
these products. But, if spacetime were two-dimensional, we
would obtain the pictures included in the previous chapters.
 fiber bundles 165

Now, the whole point of this construction is that we add

a copy of one space F to each point of another space B.
In general, we call B the base space and F the fiber. The
complete geometrical object we get by attaching a fiber F to
each point of the base space B is called a fiber bundle.

In the financial toy model and in quantum mechanics, our

fibers are one-dimensional (a line and a circle respectively).3 3
The dimension of a
space corresponds to
But we can use higher-dimensional fibers, too.4
the smallest number of
coordinates required to
localize an object in the
given space. To describe
the location of an object
on a line, we only need
one number, such as
its distance along the
line from the origin.
Similarly, to describe an
object which moves on
a circle, we only need
the angle ϕ.

4
For isospin space and
color space, our fibers
are higher-dimensional.
166 physics from finance

Another important idea is that there are different kinds of

fiber bundles. In particular, not every fiber bundle is a sim-
ple product of the form B × F. While locally, every fiber
bundle looks like a trivial product B × F, there are bundles
with a non-trivial global structure. The most famous ex-
ample of such a non-trivial bundle is a Moebius strip. The
trivial bundle we get by attaching a line to each point of
a circle looks like a cylinder. But we can also imagine that
there is a twist in the fibers as we move around the circle.
5
We will talk about The geometrical object we get if there is a twist in the fibers
why we are inter- is a Moebius strip.
ested in such strange
constructions below.
6
We will talk about cur-
vature below. Moreover,
topology is the branch
of mathematics which
deals with aspects of
geometrical objects
that remain unchanged
as we deform the ob-
ject. In other words, in The defining feature of such a non-trivial fiber bundle is
topology, we imagine that we can’t use the same coordinate system to describe
that each object is made
of some mouldable clay locations on it globally. Instead, we need at least two coordi-
which we can deform nate systems and a rule which glues these patches together.5
however we like. (How-
ever, it’s not allowed
to puncture or heal
nor to add or remove
holes). Any two geo-
metrical objects which
can be transformed into
each other this way are
topologically equiv-
alent. For example, a
doughnut and a coffee
mug are topologically
equivalent.
With this in mind, we can classify fiber bundles as follows:6

B Type-1 bundles: (topologically) trivial bundles with zero

An object is topo-
logically non-trivial curvature.
whenever there is some
knot that we can’t get
rid of by deforming the
object.
 fiber bundles 167

B Type-2 bundles: (topologically) trivial bundles with non-

zero curvature.
B Type-3 bundles: (topologically) non-trivial bundles.

Each type is important in physics and corresponds to a

different situation.

B Type-1 bundles describe systems with a vanishing elec-

tromagnetic field strength. In the financial toy model, we
have a trivial bundle with zero curvature whenever there
are no arbitrage opportunities.
B Type-2 bundles describe systems with non-zero electro-
magnetic field strength. In the financial toy model, we
have a trivial bundle with non-zero curvature whenever
there are arbitrage opportunities.
B Type-3 bundles describe systems in which topological
objects like a magnetic monopole are present.7 7
A non-trivial color
space bundle is nec-
essary to describe
instantons.
In this book, however, we will only talk about type-1 and
type-2 bundles. To properly understand non-trivial bundles,
we need quite a bit of topology and discussing this would
lead us too far astray. But there are many great resources to
learn more about non-trivial bundles and several are listed
in Chapter 6.

Nevertheless, it is important to keep in mind that fiber bun- 8

There are lots of other
dles are a generalization of the direct product of spaces. In types of bundles. For
example, the fibers of a
general, fiber bundles only locally look like a direct product. tangent bundle are the
But globally, they can be non-trivial. tangent spaces Tp M at
each point of the base
manifold M. Tangent
Before we move on, there is one final distinction we need to bundles are important
talk about. It is conventional to give special names to fiber in the Lagrangian
formulation of classical
bundles depending on what kind of space F we attach to mechanics.
each point of the base space B:8
168 physics from finance

B If F is a vector space, we call the resulting bundle a vector

bundle.
B If F is a Lie group, we call the resulting bundle a princi-
9
In the previous section pal bundle.9
we learned that we
can understand Lie
groups geometrically.
For example, the group
In physics, we need both. To understand why, we need
U (1) is geometrically to recall an extremely important fact about elementary
a circle. Therefore, we particles.
can imagine that the
U (1)-principal bundle
consists of circles which Different elementary particles are influenced differently by
we attach to each point
of the base space.
the structure of the various internal spaces.10 This is nec-
essarily the case because otherwise we couldn’t distinguish
10
For example, an them. After all, we define elementary particles depending
electron is influenced
on how they are influenced by the structure of the internal
by the structure of
charge space, while a spaces.
neutrino isn’t.

How can we describe this using fiber bundles?

First of all, recall that in Section C.2 we discussed that we

11
For example, U (1) can understand symmetries geometrically. Each symmetry
corresponds to a unit group corresponds to a particular geometrical object. There-
circle and SU (2) to a
sphere. fore, to construct the arena we use to describe elementary
particles and their interactions, we can imagine that we at-
12
The construction tach a copy of the fundamental symmetry groups to each
for the remaining point in spacetime.11
symmetry groups
works analogously. We
choose SU (2) because For concreteness, let’s focus on SU (2) for a moment.12 To
it’s sufficiently non-
trivial (unlike U (1)) construct the arena we use to describe weak interactions,
but not too complicated we therefore imagine that we attach a copy of SU (2) to each
(unlike SU (3)). point in spacetime.
 fiber bundles 169

This is a principal bundle. And this principal bundle is our

starting point.13 13
Geometrically, SU (2)
is the three-sphere
S3 . The three-sphere
Now, some particles are influenced by the structure of is the surface of a
isospin space while others aren’t. Moreover, we can imag- four-dimensional
ball. We can’t draw
ine that there are additional elementary particles which in four dimensions
would be influenced even more strongly than the particles and therefore the two-
we know so far. This would be analogous to how a particle sphere S2 (the surface
of a three-dimensional
with a large electric charge is influenced more strongly by ball) is drawn here.
the structure of charge space than a particle with a small
charge.

The branch of mathematics which allows us to describe this

is known as representation theory.

14
Once more, we
will not discuss any
D.1 Representation Theory details and focus on
the key ideas. Group
theory is discussed
in Appendix C and a
Representation theory is a part of group theory and the detailed discussion of
fundamental task is to study how a given group acts on dif- representation theory
ferent objects.14 For beginners, this is usually an extremely can be found in

confusing idea. For example: Jakob Schwichten-

berg. Physics from Sym-
metry. Springer, Cham,
Switzerland, 2018b.
ISBN 978-3319666303
170 physics from finance

B Elements of U (1) are defined as transformations of the

form eix . So, of course, U (1) acts on complex functions.

B Elements of SU (2) are (2 × 2) matrices with unit deter-

minant that fulfill the condition M† M = 1. So, of course,
SU (2) acts on complex column vectors with two entries.
15
A subtlety is that
there are right-chiral
partners of all fermions.
Chirality is a property But we can understand groups in a more abstract sense, too.
of elementary particles In fact, we need this abstract perspective to make sense of
closely connected to
their spin. For massless modern physics. We know that not all particles are affected
particles, chirality is equally by the structure of our internal spaces. For example,
equal to the projection an SU (2) transformation rotates an electron into a neutrino
of the particle’s spin
onto its momentum and vice versa. But, there are also particles which do not get
vector. In this context, affected by the structure of isospin space.15 Therefore, an
right-chiral partners are
interesting because they SU (2) transformation will not have any effect on them.
do not carry isospin.
They therefore are To make sense of this, we need to forget for a moment that
not influenced by the
structure of isospin we define groups in the first place as a concrete set of trans-
space. This is the only formations acting on specific objects. Instead, we now only
thing we will focus on
here. For example, a remember that each group is a certain geometrical object.
left-chiral electron e L Each point on this geometrical object corresponds to one
carries isospin, while specific group element.
a right-chiral electron
e R does not. For all
fermions right-chiral
partners have been
observed already except
for the neutrinos. (The
connection between
the elementary particle
property of chirality
(in its observed relation
to spacetime) and the
internal property of
isospin is one of the
unsolved mysteries in
modern physics.)
 fiber bundles 171

This is how we can abstractly view our group. The main 16

Of course, we don’t
idea of representation theory is that we can find maps from try to find just any
map, but a map which
this abstract geometrical object to the specific linear transfor- fulfills specific condi-
mations on different vector spaces.16 In practice, this means tions. These conditions
ensure that the group
that we try to find a way to represent our group elements properties are pre-
as matrices which act on the elements of different vector served.
spaces. A representation is
a map between any
group element g of a
For example, we can pick some vector space like C2 and group G and a linear
transformation T ( g) of
then search for maps which yield for each point of our ab- some vector-space V in
stract geometrical object a concrete linear transformation such a way that:
which acts on the elements of C2 . The elements of C2 are B T (e) = I (The iden-
complex column vectors with two entries. Linear transfor- tity element of the
group transforms
mations acting on them are therefore (2 × 2) matrices. This nothing at all.)
way we get the standard representation of SU (2). B T ( g −1 ) = T ( g )

−1

(Every inverse ele-

ment is mapped to
the corresponding
inverse transforma-
tion.)
B T ( g) ◦ T (h) = T ( gh)
(The combination
of transformations
corresponding to g
and h is the same as
the transformation
corresponding to the
point gh)
In mathematical terms,
we say that a map with
these properties is a
But we can also pick some other vector space like C3 and homomorphism.
then search for maps onto the linear transformation on
C3 . Elements of C3 are complex column vectors with three 17
It is crucial to un-
entries. Therefore linear transformations acting on them are derstand that this
three-dimensional rep-
(3 × 3) matrices. resentation of SU (2)
is not simply SU (3).
The three-dimensional
In other words, each element of SU (2) gets mapped to one representation of
specific (3 × 3) matrix. This yields the three-dimensional SU (3) looks very
representation of SU (2)!17 different from the
three-dimensional rep-
resentation of SU (2).
172 physics from finance

We can also pick an extremely simple vector space like C.

But the only way we can represent SU (2) as linear operators
acting on elements of C is to map them all to the identity
which changes nothing at all. This means that the one-
dimensional representation of SU (2) is trivial and consists
only of the element 1 which leaves elements of C completely
unchanged.

This is exactly what we need when we try to describe ele-

mentary particles which do not get influenced by SU (2) at
all. When SU (2) acts on the objects that we use to describe
elementary particles without isospin, we need to represent
its elements using the one-dimensional representation. In
contrast, when SU (2) acts on objects we use to describe
 fiber bundles 173

particles carrying isospin, we need to use a non-trivial rep-

resentation. It is conventional to say that particles without
isospin live in the one-dimensional representation of SU (2),
while particles like the (left-chiral) electron and (left-chiral)
electron-neutrino live in the two-dimensional representa-
tion.18 18
Take note that for a
particle with electric
charge q, elements of
Statements like this often sound extremely abstract. It is the corresponding U (1)
therefore important to keep in mind that the main idea is representation look
like eiqe . Therefore, the
really that different elementary particles experience the U (1) representation
structure of our internal spaces differently. Representation for particles without
theory is the mathematical toolkit we can use to take this electric charge q = 0
is simply ei0e = 1 .
fact into account. In other words, the
q = 0 representation
of U (1) is trivial,
With this in mind, let’s get back to our fiber bundles. analogous to the trivial
representation of
SU (2) and is therefore
exactly what we need
to describe particles
that are not affected by
the structure of charge
space.

D.2 Associated Bundles

Before we started talking about representation theory, we

noted that the object we start with in physics is called a
principal bundle. A principal bundle consists of a copy of a
specific group G attached to each point in spacetime.19 19
Since in fundamental
physics there are
multiple internal
Now, we somehow need to take the lessons learned in the symmetries, we attach
previous section into account. all these groups to each
point in spacetime.

We can do this by using so-called associated vector bun-

dles. The idea here is that a principal bundle has an associ-
ated vector bundle for each of its representations.
174 physics from finance

For the one-dimensional representation, we add a copy

of the corresponding one-dimensional vector space that
the group elements act on to each point of the base space.
Similarly, for the two-dimensional representation, we add
a copy of the corresponding two-dimensional vector space
and for the three-dimensional representation, we add a copy
of the three-dimensional vector space.

Intuitively, these associated vector spaces represent how

the internal space (e.g. isospin space) looks like for a par-
ticular particle. A particle without isospin, only sees the
20
An example of a one-dimensional version20
particle without isospin
is the right-chiral
electron.

In contrast, the left-chiral electron/electron-neutrino state

sees the two-dimensional version:
 fiber bundles 175

Now that we have a rough understanding of the arena we

use in physics, it’s time to talk about the objects which glue
our fibers together. These objects tell us how a given particle
moves from fiber to fiber as it moves through spacetime. In
other words, they tell us which path in a fiber a given parti-
cle follows as it follows a specific trajectory in spacetime.
21
In mathematical
terms, the connections
live in the adjoint
representation of
the corresponding
symmetry group.
This is an important
representation for each
Lie group because it
tells us how the group
acts on its own Lie
algebra. While a Lie
group is, in general, a
complicated manifold,
the corresponding Lie
algebra is always a
vector space. Therefore,
we can investigate how
Mathematically, we call connections Aµ , the objects which any given group acts
on this particular vector
tell us how particles move from fiber to fiber.21 And tech- space.
176 physics from finance

nically, a connection allows us to parallel transport a given

object. Therefore, before we can talk about connections, we
need to talk about parallel transport.

D.3 Parallel Transport

A crucial task whenever we are dealing with a non-trivial

geometrical arena is to understand how we can move a
given object around consistently. We need a method that
22
Reminder: We com- allows us to compare two objects at different locations.22
pare objects at different This is not an easy task because any apparent difference
locations whenever we
calculate the derivative. between two objects could simply be the result of different
local conventions or a result of the non-trivial structure
of our geometrical arena. This means that although two
objects may look very different at first glance, we may find
that they are equivalent when we bring them to the same
location.

For example, 100 units of the local currency in some coun-

try A and 1000 units of the local currency in another may
look very different at first glance. But if the exchange rate
is actually 10 : 1 these two amounts of money are actually
equivalent. The exchange rate is an example of a connec-
tion which is the topic of the next section. In this section,
however, we want to understand a little better the process of
moving objects around consistently.

Imagine that you’re walking on a curved surface while hold-

ing a stick in your hand. Your task is to bring the stick to
another location while keeping its orientation unchanged (at
every step) relative to the local surface. When you success-
fully do this, you parallel transport the stick.
 fiber bundles 177

In contrast, a non-parallel transport of the stick looks as

follows

Moreover, take note that to understand parallel transport

you really need to imagine that you’re really walking along
the surface. The transport shown in the following picture is
not parallel even though it looks parallel from the outside:
178 physics from finance

With this intuitive idea in mind, let’s talk about connections.

D.4 Connections

In the financial toy model, after the introduction of the

23
To unclutter the bookkeepers Aµ , we can perform local dilations23
notation, we restrict
ourselves to one spatial
dimension here.
f ( x ) = ee( x ) (D.1)

since our bookkeepers adjust accordingly (Eq. 2.21). Anal-

ogously, in quantum mechanics we can then perform local
phase shifts
f ( x ) = eie( x) (D.2)

as soon as we introduce the bookkeepers Aµ .

The crucial point is that now our transformation parameters

e are functions of the location x.

In other words, we can shift the prices or analogously the

phase at each point x in space by a different amount. With-
out the bookkeepers, only global shifts are permitted which
 fiber bundles 179

shift the prices, or analogously the phases, everywhere by

exactly the same amount.

We then exploited this knowledge to understand the inter-

nal space of the toy model and of quantum mechanics. In
the toy model, our internal space at each location is the real
line R+ . This line represents all possible prices.

In quantum mechanics, the internal space at each location is

a unit circle in the complex plane. This circle represents all
possible phase factors. The total internal space is all these
individual internal spaces taken together.

A crucial ingredient to make this possible are the bookkeep-

ers Aµ which glue the individual spaces together.

Mathematically, the bookkeepers Aµ are called connections.

This name makes sense because the connections are what
we need to consistently describe how goods (like copper)
and elementary particles move around.

This is non-trivial, as we have discussed extensively in the

180 physics from finance

context of the financial toy model, because there can be local

currencies. This means that it is not immediately clear what
a given amount of money is worth in terms of the currency
of a neighboring country. In financial terms, a connection
tells us the exchange rates between currencies. Formulated
differently, we need the connections to keep track of local
shifts. Nothing physical in the system can be influenced by
local phase or price shifts because these quantities do not
have any meaning in an absolute sense. Every time such a
shift happens, the bookkeepers adjust accordingly such that
the dynamics stay the same.

But connections are not only important to allow arbitrary

local transformations. Connections are absolutely essential
whenever there is curvature. In the context of the financial
toy model, we called curvature a currency arbitrage oppor-
tunity Fµν (t, ~x ). In electrodynamics, we call curvature the
electromagnetic field Fµν (t, ~x ). We will talk about how the
name curvature comes about below, but first let’s clarify this
important point.

In the financial toy model, an arbitrage opportunity has,

so to speak, an absolute reality, it is something physical
which directly influences the dynamics; it is not something
which relies on local conventions. Hence, it cannot go away
through adjustments of the local currencies.

Formulated differently, when there is an arbitrage opportu-

nity (Fµν 6= 0), it is impossible to choose a unique (global)
currency. If it were possible to choose a global currency,
all exchange rates would be trivial (Aµ = 0 everywhere)
and there would therefore be no arbitrage opportunities.
We can see this immediately by recalling the relationship
between the exchange rates and the gain factors associated
 fiber bundles 181

with currency arbitrage opportunities (Eq. 2.22):

∂Aν (t, ~x ) ∂Aµ (t, ~x )

Fµν (t, ~x ) ≡ − . (D.3)
∂x µ ∂x ν
The existence of an arbitrage opportunity means Fµν 6= 0
somewhere, and therefore Aµ has to be non-vanishing.

In other words, since arbitrage opportunities cannot go

away through changes of local conventions, we can conclude
that whenever there is one, there will be local currencies
and non-trivial exchange rates. A non-trivial exchange rate
means that our description of the dynamics must include the
bookkeepers Aµ .

We can also imagine that countries use local currencies but

there is no currency arbitrage opportunity. In this case, the
bookkeepers are not essential parts of the model because we
are free to use the symmetry to introduce a global currency.
In other words, by making use of local gauge transforma-
tions, we can make the exchange rates trivial everywhere
and hence there is no need for us to use the bookkeepers Aµ
at all. Moreover, we learned that as soon as there is copper
located somewhere in the system, there will be an arbitrage
opportunity and we therefore need the bookkeepers Aµ .24 24
This is what Eq. 2.34
tells us.

We have exactly the same situation in quantum mechanics.

On the one hand, we need the bookkeepers to allow arbi-
trary local phase shifts. But these local phase shifts do not
influence the dynamics of the system and we could easily
live without them.25 25
Formulated differ-
ently, the freedom to
perform local phase
But on the other hand, as soon as there is a charged parti- shifts and to intro-
cle, there is a non-zero electromagnetic field Fµν 6= 0 and duce local currencies
is nice to have but not
the bookkeepers Aµ are essential.26 Formulated more tech- essential.
nically, if there is no charged object present in the system,
we can use local gauge transformations to get rid of the
26
Again, this is what
Eq. 2.34 and Eq. 2.22
tell us.
182 physics from finance

bookkeepers Aµ everywhere. But the presence of just one

charged object in the system is enough to make bookkeepers
essential parts of the system and we can’t get rid of them
everywhere at the same time.

Using the terminology introduced above, we can say that

connections are only essential for type-2 bundles (and type-
3 bundles). For type-1 bundles, there is no curvature and
therefore we can use our local symmetry to set the connec-
tions to zero everywhere.
 fiber bundles 183

D.5 Curvature

Now, why exactly do we call the quantity

∂Aν (t, ~x ) ∂Aµ (t, ~x )
Fµν (t, ~x ) ≡ − . (D.4)
∂x µ ∂x ν
the curvature?

To understand this, we need to discuss how we can mea-

sure the curvature of a given space in general. While the
details always depend on the space at hand, the main idea
is always that we move an object along a closed loop and
then compare the state of the object after the journey with
its state at the beginning.

The prototypical example of a curved space is a sphere. To

understand how we can measure the curvature of a sphere
imagine that you are walking on the sphere while holding
a stick in your hand. While you walk, you always do your
best to keep the stick straight.27 An important observation 27
If you do this you’re
is that if the space you are moving in is curved, it’s possible parallel transporting
the stick. In our frame-
that the stick, even though it returns to its starting location, work, the connections
does not end in its starting orientation if you move along a Aµ make sure that we
are transporting our
closed loop. objects correctly. This is
crucial because we want
to measure a real fea-
ture of the system (the
curvature) and want
to avoid that any local
convention distorts our
result.

Hence, the difference between the original vector and the

184 physics from finance

vector which was parallel transported along a closed loop

encodes information about the curvature. In contrast, if we
parallel transport a vector in a flat space, its initial and final
state will be equivalent:

Completely analogously, if there is an arbitrage opportunity

in the financial toy model, we can move a specific amount
of money along a closed loop and end up with more money
than we started with.
 fiber bundles 185

This indicates that we are dealing with a curved internal

space. This is precisely how we defined the gain factor in
Eq. 2.9.28 28
Reminder: Eq. 2.9
reads
Fij (~n) =
A j (~n +~ei ) − A j (~n)
− [ Ai (~n +~e j ) − Ai (~n)] .

In this sense, this gain factor is a measure of the curvature

of the internal money space. And, in exactly the same sense,
we can say that the electromagnetic field strength tensor
Fµν (t, ~x ) encodes how much charge space is curved.

This is one of the main observations at the heart of modern

physics.
E

Taylor Expansion

The Taylor expansion is one of the most useful mathemat-

ical tools and we need it in physics all the time to simplify
complicated systems and equations.

We can understand the basic idea as follows:

Imagine you sit in your car and wonder what your exact
location l (t) will be in 10 minutes: l (t0 + 10 minutes) =?

B A first estimate is that your location will be exactly your

current location

l (t0 + 10 minutes) ≈ l (t0 ) .

1
Here ∂t is a shorthand
Given how large the universe is and thus how many notation for ∂t∂ and
possible locations there are, this is certainly not too bad. ∂t l (t) yields the velocity
(rate of change). After
B If you want to do a bit better than that, you can also taking the derivative,
we evaluate the velocity
include your current velocity l˙(t0 ) ≡ ∂t l (t) t .1 The total function l˙(t) ≡ ∂t l (t) at
0
distance you will travel in 10 minutes if you continue to t0 : l˙(t0 ) = ∂t l (t) t .
0
188 physics from finance

move at your current velocity is this velocity times 10

minutes: l˙(t0 ) × 10 minutes . Therefore, your second
estimate is your current location plus the velocity you are
traveling times 10 minutes

l (t0 + 10 minutes) ≈ l (t0 ) + l˙(t0 ) × 10 minutes . (E.1)

B If you want to get an even better estimate you need to

take into account that your velocity can possibly change.
The rate of change of the velocity l¨(t0 ) = ∂2t l (t) t is what
0
we call acceleration. So in this third step you additionally
2
The factor 21 and that take your current acceleration into account2
we need to square the
10 minutes follows
since, to get from l (t0 + 10 minutes) ≈ l (t0 ) + l˙(t0 ) × 10 minutes
an acceleration to a
1
location, we have to + l¨(t0 ) × (10 minutes)2 .
integrate twice: 2
Z Z
dt dt ẍ (t0 ) =
Z
dt ẍ (t0 )t = B Our estimate will still not yield the perfect final location
1 since, additionally, we need to take into account that
ẍ (t0 )t2
2 our acceleration could change during the 10 minutes.
where ẍ (t0 ) is the value We could therefore additionally take the current rate of
of the acceleration at
change of our acceleration into account.
t = t0 (= a constant).

This game never ends and the only limiting factor is how
precisely we want to estimate our future location. For many
real-world purposes, our first order approximation (Eq. E.1)
would already be perfectly sufficient.

The procedure described above is exactly the motivation be-

hind the Taylor expansion. In general, we want to estimate
the value of some function f ( x ) at some value of x by using
our knowledge of the function’s value at some fixed point a.
3
Here the superscript The Taylor series then reads3
(n), within brack-
ets, denotes the n-th
derivative. For example
f (0) = f and f (1) is ∂ x f .
 taylor expansion 189

∞
f (n) ( a)( x − a)n
f (x) = ∑ n!
n =0
f (0) ( a)( x − a )0 f (1) ( a)( x − a)1 f (2) ( a)( x − a)2
= + +
0! 1! 2!
f (3) ( a)( x − a)3
+ +..., (E.2)
3!
where f ( a) is the value of the function at the point a we are
expanding around. Moreover, x − a is analogous to the 10
minute timespan we considered above. If we want to know
the location at x = 5:10 pm by using our knowledge at a =
5:00 pm, we get x − a = 5:10 pm − 5:00 pm = 10 minutes.
Therefore, this equation is completely analogous to our
estimate of the future location we considered previously.

To understand the Taylor expansion a bit better, it is helpful

to look at concrete examples.

We start with one of the simplest but most important ex-

amples: the exponential function. Putting f ( x ) = e x into
Eq. E.2 yields4

∞ ( e x )(n) ( x − 0) n
ex = ∑ x =0
n!
n =0

4
The notation
( e x )(n) means
The crucial puzzle pieces that we need are = and (e x ) 0 ex x =0
that we take the n-th
e0 = 1. Using these relations and substituting into the derivative of the func-
previous equation yields tion e x and then plug in
x = 0 into the result.
∞ ∞
e0 ( x − 0)n xn
ex = ∑ n!
= ∑
n!
(E.3)
n =0 n =0

This result can be used as a definition of ex .

Next, let’s assume that the function we want to approximate

is sin( x ) and we want to expand it around x = 0.
190 physics from finance

5
As before, the notation Putting f ( x ) = sin( x ) into Eq. E.2 yields5
 (n)
sin( x ) means
x =0
that we take the n-  (n)
th derivative of the ∞ sin( x ) ( x − 0) n
function sin( x ) and
then plug in x = 0
sin( x ) = ∑ n!
x =0
n =0
into the result. So for
example, since
  (1) The crucial information we therefore need is (sin( x ))0 =
sin( x ) = cos( x ),
cos( x ), (cos( x ))0 = − sin( x ), cos(0) = 1 and sin(0) = 0.
we have, Because sin(0) = 0, every term with even n vanishes, which
  (1) we can use if we split the sum. Observe that
sin( x ) = cos(0)
x =0
= 1.
∞ ∞ ∞
∑ n = ∑ (2n + 1) + ∑ (2n)
n =0 n =0 n =0
1+2+3+4+5+6... = 1+3+5+... +2+4+6+...

Therefore, splitting the sum into even and odd terms yields

 (2n+1)
∞ sin( x ) ( x − 0)2n+1
sin( x ) = ∑ x =0
(2n + 1)!
n =0
 (2n)
∞ sin( x ) ( x − 0)2n
+ ∑ (2n)!
x =0
n =0
y

sin(0) = 0
 (2n+1)
∞ sin( x ) (x − 0)2n+1
= ∑ (2n + 1)!
x =0
. (E.4)
n =0

Moreover, every even derivative of sin( x ) (i.e., sin(2n) ) is

again sin( x ) or − sin( x ). Therefore the second term van-
ishes since sin(0) = 0. The remaining terms are odd deriva-
tives of sin( x ), which are all proportional to cos( x ). We now
 taylor expansion 191

use

sin( x )(1) = cos( x )

sin( x )(2) = cos0 ( x ) = − sin( x )
sin( x )(3) = − sin0 ( x ) = − cos( x )
sin( x )(4) = − cos0 ( x ) = sin( x )
sin( x )(5) = sin0 ( x ) = cos( x )

The general pattern is sin(2n+1) ( x ) = (−1)n cos( x ), as you

can check by putting some integer values for n into the
formula6 . 6
sin(1) ( x ) =
sin(2·0+1) ( x ) =
(−1)·0 cos( x ) = cos( x ),
Thus, we can rewrite Eq. E.4 as sin(3) ( x ) =
sin(2·1+1) ( x ) =
 (2n+1) (−1)1 cos( x ) = − cos( x )
∞ sin( x ) ( x − 0)2n+1
sin( x ) = ∑ x =0
(2n + 1)!
n =0
∞
(−1)n cos(0)( x − 0)2n+1
= ∑ (2n + 1)!
n =0
y

cos(0) = 1
∞
(−1)n ( x )2n+1
= ∑ (2n + 1)! (E.5)
n =0

This is the Taylor expansion of sin( x ), which we can also

use as a definition of the sine function.
Bibliography

John Baez and Javier P. Muniain. Gauge fields, knots, and

gravity. World Scientific, Singapore River Edge, N.J, 1994.
ISBN 9789810217297.

H. J. Bernstein and A. V. Phillips. Fiber Bundles and Quan-

tum Theory. Sci. Am., 245:94–109, 1981.

Subrahmanyan Chandrasekhar. A scientific autobiography.

World Scientific, Singapore Hackensack, NJ, 2010. ISBN
9789814299572.

Ta-Pei Cheng. Relativity, Gravitation and Cosmology: A Basic

Introduction. Oxford University Press, 2nd edition, 1 2010.
ISBN 9780199573646.

Jennifer Coopersmith. The lazy universe. Oxford University

Press, 2017. ISBN 9780198743040.

Richard Feynman. The Feynman lectures on physics.

Addison-Wesley, San Francisco, Calif. Harlow, 2011. ISBN
9780805390650.

Daniel Fleisch. A student’s guide to Maxwell’s equations.

Cambridge University Press, 2008. ISBN 978-0521701471.
194 physics from finance

William Fulton and Joe Harris. Representation Theory: A

First Course. Springer, 1st corrected edition, 8 1999. ISBN
9780387974958.

David Griffiths. Introduction to quantum mechanics. Pearson

Prentice Hall, 2005. ISBN 9780131118928.

David Griffiths. Introduction to electrodynamics. Pearson

Education Limited, Harlow, 2014. ISBN 9781292021423.

Chan Hong-Mo and Tsou Sheung Tsun. Some elementary

gauge theory concepts. World Scientific, Singapore River
Edge, N.J, 1993. ISBN 9789810210809.

Kirill Ilinski. Physics of Finance. 1997.

Kirill Ilinski. Physics of finance: gauge modelling in non-

equilibrium pricing. John Wiley, Chichester England New
York, 2001. ISBN 9780471877387.

Robert D. Klauber. Student Friendly Quantum Field Theory.

Sandtrove Press, 2nd edition, 12 2013. ISBN 9780984513956.

Pia Malaney. The index number problem: A differential

geometric approach. 1996.

Juan Maldacena. The symmetry and simplicity of the laws

of physics and the Higgs boson. Eur. J. Phys., 37(1):015802,
2016. doi: 10.1088/0143-0807/37/1/015802.

David Morin. Introduction to classical mechanics: with prob-

lems and solutions. Cambridge University Press, 2008. ISBN
9780511808951.

Gregory Naber. Topology, geometry and Gauge fields: founda-

tions. Springer Science+Business Media, LLC, New York,
2011. ISBN 9781441972538.

Roger Penrose. The road to reality: a complete guide to the

laws of the universe. Vintage Books, New York, 2007. ISBN
9780679776314.
 bibliography 195

Jakob Schwichtenberg. No-Nonsense Electrodynamics. No-

Nonsense Books, 2018a. ISBN 978-1790842117.

Jakob Schwichtenberg. Physics from Symmetry. Springer,

Cham, Switzerland, 2018b. ISBN 978-3319666303.

Jakob Schwichtenberg. No-Nonsense Quantum Mechanics.

No-Nonsense Books, 2018c. ISBN 978-1719838719.

Jakob Schwichtenberg. No-Nonsense Classical Mechanics.

No-Nonsense Books, 2019. ISBN 978-1096195382.

John Stillwell. Naive Lie Theory. Springer, 1st edition,

August 2008. ISBN 978-0387782140.

Samuel E. Vazquez and Simone Farinelli. Gauge invari-

ance, geometry and arbitrage, 2009.

K. Young. Foreign exchange market as a lattice gauge

theory. American Journal of Physics, 67(10):862–868, 1999.
doi: 10.1119/1.19139. URL https://doi.org/10.1119/1.
19139.

Anthony Zee. Einstein Gravity in a Nutshell. Princeton

University Press, 1st edition, 5 2013. ISBN 9780691145587.
Index

Ampere-Maxwell Law, 54 Current

Antisymmetry, 42, 53 Copper, 48, 49
Arbitrage, 28, 181, 184 Curvature, 31, 119, 183
Dynamics, 51 Dynamical, 32
Arena of Physics, 70 Finance, 184

Base Space, 165 Diffusion Equation, 57, 77

Bookkeepers, 97, 180 Derivation, 59
Electromagnetic, 100 With Covariant Derivative, 64
Gravitational, 109 Dilation Group, 154
Indispensable, 101 Geometrically, 157
Isospin, 100 Down Quark, 72
Spacetime, 117
Boost, 115 Ehrenfest Theorem, 81
Einstein Equation, 123
Chirality, 74, 170 Einstein-Cartan Gravity, 127
Color, 74, 101, 124 Electric Charge, 74, 91, 101, 124
Color Space, 96 Electric Field, 87
Symmetry, 97 Electrodynamics, 87
Commutator, 11 Electromagnetic Field Tensor, 87
Connection, 25, 178 Electromagnetic Interactions, 69,
Conserved Charge, 124 104
Continuity Equation, 51 Electromagnetic Potential, 87
Energy-Momentum, 122 Elementary Particle, 71
Quantum Mechanics, 79 Energy-Momentum, 74, 124
Derivation, 80 Euclidean Space, 110
Continuum Limit, 43, 44 Exchange Rate, 37
Covariant Derivative, 63, 90 Exchange Rates
198 physics from finance

Temporal, 42 Light Cone, 115

Lorentz Group, 115
Fermion, 75
Fiber, 165 Magnetic Field, 87
Fiber Bundle, 163 Maximum Velocity, 114
Associated Bundles, 173 Maxwell Equation
Finance, 164 Derivation, 53
Principal Bundle, 168 Homogeneous, 54
Types, 167 Inhomogeneous, 53
Vector Bundle, 168 Mean, 60, 142
Field Theory, 130 Metric, 123, 147
Minkowski Metric, 149
Gain Factor, 43, 44, 46 Minkowski Space, 112
Gauge Boson, 100 Moebius strip, 166
Gauge Bosons, 73
Gauge Symmetry, 21 Noether’s Theorem, 124
Gauge Invariant, 26
Gauge Transformation, 38 Open Questions, 136
Continuous, 44
Gauss’s Law, 54 Parallel Transport, 176
General Relativity, 117 Particle Theory, 130
Generator, 161 Phase Factor, 83
Gravitational Interactions, 108 Poincaré Group, 115
Gravity, 69, 108 Probability Density, 58, 79
Group Theory, 151 Probability Distribution, 59, 141
Abstract, 170
Geometrically, 156 Quantum Field Theory, 130
Quantum Mechanics, 77
Homomorphism, 170 Quantum Theory, 130
Quarks, 72
Internal Space, 23
Charge Space, 70 Random Walk, 56
Color Space, 70 Mathematically, 59
Isospin Space, 70 Pollen, 56
Quantum Mechanics, 82, 85 Traders, 57
Isospin, 74, 101, 124 Representation Theory, 169
Isospin Space, 94 Adjoint, 176
Symmetry, 97
Schrödinger Equation, 78
Lattice Spacing, 43 with Covariant Derivative, 91
Lie Algebra, 38, 157 Spacetime, 107
Lie Group, 38 Curvature, 120
 index 199

Lattice, 35, 43 Rescaling, 23

Torsion, 126
Special Relativity, 109 Taylor Expansion, 187
Speed of Light, 114, 123 Torsion, 121, 126
Spin, 74, 126 Translation Group, 115
Standard Deviation, 61, 143
Statistics, 142 U(1), 154
Strong Interactions, 69, 100 Geometrically, 157
SU(2), 97, 168 Up Quark, 72
Three-dimensional Represen-
tation, 172 Vacuum Fluctuations, 78
Trivial Representation, 172 Variance, 143
Two-dimensional Representa- Virtual Particles, 78
tion, 171
SU(3), 97 Wave Function, 83
Symmetry, 20 Weak Interactions, 69, 100