We are a leading

quantitative research
and technology firm.

We pair our people’s expertise with machine learning, big

data and emerging tech to predict movements in financial
markets.

View opportunities in:

Quantitative Research
Data Science
Machine Learning
Software Engineering
Data Analytics

gresearch.com
 contents

7 What’s hot and what’s not

8 The human mug
Don’t let Ashleigh Ratcliffe make you a
cup of coffee
13 News
20 Cryptic crossword
21 A symmetric universe
A man, a plan, a canal, Sam Kay
3
28 Latticetude
In conversation with Alison Etheridge
29 How to write a crossnumber Elinor Flavell talks to Oxford’s professor of
As usual, Matthew Scroggs doesn’t probability
begin with 0
34 Oπions: In defence of toy models
Calum Ross puts them back in his pram
36 Dear Dirichlet
38 What a load of noise
Certainly! Here’s a tagline for
Madeleine Hall’s article on AI.
42 Puzzles
50 The big argument
Calculators: phone or physical? 14
51 Universal poetry Unlocking sudoku’s secrets
Roses are red, Sara Logsdon is trapped in a 9-by-9 grid
Scorpions are blue,
Andrew Stacey has written
An article for you.
57 Which sport are you?
58 Reviews & Book of the year
60 A day in the life...
...of three people in finance (not all 6′5″)
64 Things we learned at...
...MathsJam Gathering 2024
66 The crossnumber

44
Win a £100 Maths Gear goody bag
68 Skimming potatoes
Ryan Palmer lies to the Daily Mail On the cover: Voronoi diagrams
and generative art
72 Top ten: things to get in the post
Hayden Mankin’s got it covered

spring 2025 ⋅ 𝟣
 Welcome to issue 21! We’ve all been enjoying Apple’s lat-
est show, Prime Target, and it has inspired us to write our
own screenplay. If you work for Apple and would like to
commission it, you can find our contact details to the left.

The team INT. COMMON ROOM

TD Dang HELENA
Elinor Flavell Hello, Anna, my good friend who
Madeleine Hall also lives in Cambridge.
Ellen Jolley
Sam Kay ANNA
Sophie Maclean No time for for pleasantries
Ashleigh Ratcliffe Helena, even though we both have
PhDs. We have to defuse this bomb.
Calum Ross
What is the code?
Matthew Scroggs
Belgin Seymenoğlu HELENA
Adam Townsend Have we tried 11513221901876399256-
Clare Wallace 5095597973971522401? It’s a
39-digit number that is equal to
the sum of the 39th powers of its
With help from
digits.
Elizabeth Brocklebank
Selma Cholich Anna hastily presses 39 buttons on the keypad.
Coco Li Nothing happens.
Vanessa Madu
HELENA
Cover artwork Don’t forget to press enter.
Hayden Mankin
Anna hastily presses enter. Nothing happens.

d chalkdustmagazine.com ANNA
c [email protected] I guess that’s not the code.
l chalkdustmag Perhaps it’s 204?
t @chalkdustmag.bsky.social
The bomb hastily explodes.
n @[email protected]
s Chalkdust Magazine FADE TO OPENING CREDITS. MICK JAGGER SINGS
f chalkdustmag2 A SONG ABOUT PRIME NUMBERS WHILE OUR NAMES
e Chalkdust Magazine, Department of APPEAR ON SCREEN.
Mathematics, UCL, Gower Street,
London WC1E 6BT, UK.
Anyway, we think this issue is rather good. We hope you
enjoy it.

Acknowledgments
We would like to thank: all our authors and other contributors for writing and creating wonderful content; our
sponsors for allowing us to continue making the magazine; our friends at the UCL department of mathematics;
everyone at Achieve Fulfilment and Janet Townsend for their help with distribution. The emojis used throughout
this issue are Twemoji, licensed under CC BY 4.0.
ISSN 2059-3805 (Print). ISSN 2059-3813 (Online). Published by Chalkdust Magazine, Dept of Mathematics, UCL, Gower Street, London WC1E 6BT, UK. © Copyright for articles is retained by the original
authors, and all other content is copyright Chalkdust Magazine 2025. All rights reserved. If you wish to reproduce any content, please contact us at Chalkdust Magazine, Dept of Mathematics, UCL, Gower
Street, London WC1E 6BT, UK or email [email protected]

𝟤 ⋅ chalkdust
 feature

Alison Etheridge

IN CONVERSATION WITH

Alison Etheridge
Elinor Flavell talks to the recently appointed
dame about life, mathematics and more.

D
URING a red weather warning in Jan- I realised how problematic this can be and how
uary, we hunkered down, made a cup important Alison’s work on this is.
of tea, and had a chat with the soon-to-
be Dame Alison Etheridge. We talked all things Alison grew up in Wolverhampton and never
statistical modelling, her recent damehood, and imagined she would become a mathematician.
the divides between different mathematical fac- At school, her favourite subject was chem-
tions. Having dabbled in statistics at university, istry, but she does remember being taught
I was vaguely aware that we often make assump- about inscribed quadrilaterals (you may know
tions about our data before we apply a statistical them as cyclic quadrilaterals). “I just sat
model. But it wasn’t until talking to Alison that there, thinking they are really beautiful,” she

spring 2025 ⋅ 𝟥
 B versity of Edinburgh and spent time in proba-
bility and statistics departments at UC Berkeley
and Queen Mary & Westfield College in London.
In 1997 Alison returned to, and settled at, the
mathematics department at Oxford. Historically,
A though, she has a family link to Cambridge—she
found out when she was 21 that one of her ances-
tors designed the famous mathematical bridge!

C This range of mathematical interests can also

be seen in her research. Despite having a back-
ground in functional analysis (which is consid-
D
ered by many as a pure mathematical subject),
Alison fondly remembers inscribing quadrilater- Alison has done a lot of research into mathemat-
als from her school days ical modelling of biological ideas, looking at ge-
netics and evolution (a very applied topic). Her
recalls—not that she would tell her fellow class- work output shows her belief that mathematics
mates this! Her chemistry teacher pushed her should not be divided into categories such as
towards a maths degree, saying “you can do any- pure and applied, but should be thought of as one
thing with mathematics”—as our A day in the large group of mathematical sciences.
life feature demonstrates. Thanks to huge sup- She realised that she likes her mathemat-
port from the maths department, she was able ics to be motivated by a real world biolog-
to leave school a year early and take a gap year ical question because it gives her intuition
before starting a maths degree at Oxford.
Even at Oxford, Alison still did not think that she
was going to be a mathematician. She thought
she could be a physicist, while her mum wanted
her to go into the humanities. But it was through
a tutor at Oxford that she really fell in love with
mathematics, because she was challenged. “It is
only when you are challenged by mathematics
that you really begin to enjoy it,” she explains. As
she neared graduation, she began thinking about
what she wanted to do with her maths degree:
She decided that she could either be a mathe-
matician or an accountant. At the time, she felt
that “there didn’t seem to be many choices. Of
course, there were—I just wasn’t aware of them”.
In the end, she chose to go further in academia
and become a mathematician.

Pure maths v applied maths?

Alison has worked all over the world and in many
different areas of mathematics. In the early
The mathematical bridge in Cambridge was
1990s, she lectured pure mathematics at the Uni- designed by one of Alison’s ancestors.

𝟦 ⋅ chalkdust
 feature

about the problem. “I find the questions in-

triguing, but they’re also very mathematically
challenging,” she explains. ”They often spin out
things which are probably of more mathemati-
cal interest than they are of biological interest in
the sense that they generate new techniques and
they are hardcore analysis: infinite dimensional,
stochastic processes, measure value processes.”

Despite this wide-ranging research, Alison’s ab-

solute favourite mathematical topic is complex
function theory: she just thinks the proofs are There are half a billion gnomes in the UK
Biobank alo. . . wait, sorry, genomes? I suppose
really quite beautiful. Complex function theory that does make more sense.
is the study of functions of complex variables and
has uses in topics across the mathematical spec-
data. Alison and a colleague wanted to ask if
trum from number theory to engineering. It is
this assumption is valid, and they found that it
also within this field that the Riemann hypothe-
is not quite so simple. This then made them
sis arises.
look at other mathematical models and ask ques-
She partly blames the mathematical culture at tions about what assumptions were made and
Cambridge and Oxford for this big divide be- how this will change the output from the math-
tween different topics in mathematics: Oxford ematical model.
used to have different buildings for the applied
Alison is also interested in human genetics. We
and pure mathematicians, and Cambridge still
have so much data, with “something like half a
does. Alison thinks this is a shame because peo-
billion genomes sitting in the UK Biobank alone.”
ple missed out on those little informal moments.
This work is looking at the infinitesimal mathe-
“Mathematicians don’t meet at coffee or have
matical model developed by Ronald Fisher in the
those conversations at the water cooler” to learn
early 1900s. Alison and her colleagues want to
about others people’s research. This means they
know if older models (such as the infinitesimal
don’t learn to respect the work and ideas that go
model) are still relevant when we apply them to
into other types of mathematics and see the in-
the huge amounts of data we have in the mod-
tricacies that are involved in different subfields.
ern world. What do we need to do to make sure
At any given time, Alison is involved in at least these mathematical models can still give us use-
five or six different projects, working with lots ful information? It is important for the mathe-
of different people. And at the moment many matical model to give us correct information be-
of these projects are developing models which cause they are being used for projects such as
can try to give us some idea of the way that helping to identify the personal risks of certain
spatial structure interacts with genetic struc- diseases. For Alison, this research demonstrates
ture. In particular, one project is looking at the great thing about mathematics: “the ability
common mathematical models used in plant ge- to abstract things and think very clearly about
netics. These models make assumptions such them, and then come out with kind of a clean
as “over very long time periods, genetic varia- argument as to what the answer is”.
tions in a certain type of plant will just wash
out and so we can treat it as homogeneous”.
As such, you assume that there is no need to
P(damehood) = 1f
take account of the spatial variation experienced After receiving an OBE in 2017, it was recently
by this plant when applying the model to your announced in the 2025 new year honours list that

spring 2025 ⋅ 𝟧
 funding. So watch this space!
Despite this challenge, Alison is very excited by
what the academy has been able to achieve so far.
“The past two-year setup phase of the academy
has given us a tantalising glimpse of what can
be achieved when mathematical scientists from
different parts of our community work together.”
The academy’s outlook aligns closely with Ali-
son’s ethos that mathematics is one huge sub-
ject. As for the damehood, Alison is actually a
The Academy for the Mathematical Sci- little embarrassed—but she is looking forward to
ences brings together the five existing UK a fun day out in Windsor with her family.
mathematical societies, like a mathematical
society megazord. Alison tells me that in the autumn, before hear-
ing of the award, she was feeling quite down
about her work. With so much time spent in
Alison will also be awarded a damehood for her
meetings and less on research, she felt that all
services to the mathematical sciences. This is
the time and effort she had put into the mathe-
a well deserved honour, as alongside her many
matical community wasn’t worth anything. But
years of teaching and research she is also presi-
since the announcement of her damehood, lots
dent of the Academy for the Mathematical Sci-
of people have reached out and sent messages of
ences.
support and admiration, reminding her why she
The impetus for founding the academy came fol- wanted to work in mathematics to begin with. I
lowing an independent review in 2018 that sug- really appreciate her sharing this story: it’s reas-
gested the UK mathematical sciences commu- suring to see that even the most accomplished
nity needed a single authoritative voice. Since mathematicians experience ups and downs in
then, a lot of hard work has been put in to de- their careers.
sign and create an academy that will benefit the
Looking to the future, Alison sees herself doing
whole of the UK. The organisation wants to in-
mathematics for many years to come, well into
corporate mathematics within academia, educa-
retirement. Although, she hopes to wake up a
tion, business, industry and government across
little later and enjoy more of her other hobbies
the country. The goal is to increase the num-
such as walking in the Scottish hills, good food,
ber of mathematical scientists, improve diversity
theatre and time with her family.
within mathematics, and help support the gov-
ernment to create better public policy. It was So as you go about your week, take a leaf out
planned for the academy to fully launch later of Alison’s book: find something that challenges
this year but recent government budget changes you, and let yourself fall in love with mathemat-
mean it is currently looking for other sources of ics.

f Chalkdust cannot reproduce the mathematics that led to this result. We take no responsibility for the implications
shown above and will not be responding to any further queries.

Elinor Flavell
Elinor is a mathematician and tour guide based in Edinburgh. She loves all things mathematics,
history, and most things in between.

𝟨 ⋅ chalkdust
 Maths is a

hot and what’s not

fickle world.
what’s Stay à la mode
with our guide
to the latest
trends.

▲ HOT
90s reunions
▲ HOT
Oasis and 5ive leading the charge.
Releasing things in two parts Expect Andrew Wiles to re-prove
Fermat’s last theorem on his UK
Dune and Wicked both show us arenas tour this summer.
that the real trick for writing is to set
up your problem really well, then ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
leave the reader hanging for a year.
Anything 2025
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
Can we just hide from the news?
Cramming your research
into one long paper ▼ NOT
The first half is necessary and
sufficient to get it nominated for a
▲ HOT

~~~~~~~~~~~~~~~~~~~~~~~~~~
bunch of awards. And no one’s MathTok BookTok
going to sit through a six-hour
proof without needing the loo, Like MagazineTok
Function dances, pretty
anyway. but longer, in
notetaking and Howie
Hua. If it encourages black & white, and
▼ NOT anyone to improve their without fun
handwriting, mathematics features like this.
will be in a better place.
Agree?
Disagree?
▲ HOT Tag us on socials ▼ NOT
@chalkdustmag
Glambots + more advice ▼ NOT
online
Fast-shutter slo-mo Prime Target
videos for the red carpet, but
also available to capture the Everyone secretly hoping they’re on
exact moment you realise you the Americans’ list of ‘European
missed a minus sign two mathematicians who might find a
pages back. deadly prime number’. There are
▲ HOT better things to watch on Apple TV+.
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
Potatoes ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
Spambots
Better for Slow Horses
Dear Surname, First Name frying, but
Initial. You are invited to a… OK at flying Series 5 is based on the best book.
(page 68).

▼ NOT ▲ HOT
Pictures: Ariana Grande & Cynthia Erivo: Our Movie Guide, CC BY 3.0. Oasis: Will Fresch, CC BY-SA 2.0.

spring 2025 ⋅ 𝟩
 u
u m
m a
a n m
m u
u g
g
e
TThhe h
h Potatoes, mugs and doughnuts
with Ashleigh Ratcliffe

T
O topologists, mugs and doughnuts are First, let’s figure out why topologists can’t tell
the same. You may be wondering: the difference between a mug and a doughnut.
how? “Are they mad?” I hear you cry. Topology is concerned with properties of geomet-
“I can drink coffee out of a mug but I certainly ric objects; these objects can exist in any number
cannot drink coffee out of a doughnut.” By con- of dimensions. The more precise version of our
sidering the number of holes in objects, we will statement ‘to topologists, mugs and doughnuts
explore whether humans are also doughnuts—or, are the same’ is ‘a mug and a doughnut are topo-
if not, we will figure out which object humans logically equivalent’. (If we wanted to take this
are the same as. one step further, topologists would use the math-
ematical term solid torus rather than doughnut.)
With numerous words in mathematics, the math-
One ingredient of topological equivalence is ho-
ematical use/meaning does not align with our
motopy equivalence. This is when one object
everyday usage of the word; examples include
can be ‘deformed’ into the other, by a series
ring and hole. Here, the topological definition
of stretching, squashing, twisting and bending.
of a hole does not align with the everyday use of
Things we are not allowed to do to deform our
the word. For example, if you had dug a grave,
object include cutting it, opening/closing holes,
you would probably say “there is a hole in the
and gluing parts together.
ground,” but to a topologist, this does not count
as a satisfactory hole in the Earth’s surface.
Homotopy equivalence is not exactly the same
as topological equivalence, but it is easier to
work with. One of the reasons why mugs and
doughnuts are homotopy equivalent is that
they each have a single one-dimensional hole.
dough

The centre hole of the doughnut aligns with

the hole made by the handle of a mug. To
gs
mu
nu

s see this more clearly, imagine that you have

t

a stretchy and inflatable mug—If you find this

A topologist’s view

𝟪 ⋅ chalkdust
 feature

f Equivalence: homotopy v topological

Homotopy equivalence is cally equivalent, then each (doughnut) and a circle are ho-
when one object can be con- point on 𝐴 corresponds to ex- motopy equivalent, as we can
tinuously deformed into an- actly one point on 𝐵 and vice retract the solid torus to the
other by a series of stretch- versa. circle. However, this retrac-
ing, squashing, twisting and tion causes multiple points of
bending. Topological equiva- Any pair of objects that are the solid torus to map to a sin-
lence is a stronger condition topologically equivalent are gle point on the circle. This
than homotopy equivalence: homotopy equivalent, but it means that we cannot ‘undo’
it means that there is a contin- doesn’t go both ways: homo- the retraction and return to
uous bijective map between topy equivalence does not im- the solid torus, and so we
the objects. If two objects, ply topological equivalence. don’t have topological equiv-
say, 𝐴 and 𝐵, are topologi- For example, a solid torus alence.

difficult to imagine, have a look at the cartoon at

the start of this article to help you along.

For simplicity, we will call the part of the mug

that your drink goes in the container hole.

If we keep inflating the mug, the container hole

will pop out, and we’ll be left with a cylinder
with a handle attached. Then, we can squish and Potatommy
stretch the mug, redistributing the air, until the
cylinder and handle have the same thickness and
it looks like a doughnut, and voila! Doughnuts Humans are three-dimensional objects, and we
and mugs are the same (or at least, homotopy want to find out what they are homotopy equiv-
equivalent). alent to, so first we need a model of a body. We
will begin with a simple model of a human whose
only functions are eating and breathing. We will
Previously, we mentioned that digging a hole in
proceed to build our model as you would a Mr
the Earth would not be a satisfactory hole for a
Potato Head, with some small (perhaps, more re-
topologist, so let’s figure out why this is. As we
alistic) adaptations. We will refine this model
did for the mug, imagine a stretchy and inflat-
multiple times, becoming more realistic at each
able sphere representing the surface of the Earth.
iteration.
On the Earth/sphere, a ‘garden hole’ would be a
small crater on the surface. After inflating the In our simplest model, we will draw our diges-
Earth slightly—and some squishing—we would tive system as a straight-through road, open at
return to our spherical Earth. This shows that both ends. After some stretching and squishing,
the Earth’s surface along with garden holes is ho- we obtain a potato-doughnut. However, we also
motopy equivalent to the sphere, not the torus need to breathe, so let’s factor in the respiratory
(Chalkdust does not endorse this as an excuse for system. Because humans can breathe through
digging holes in your neighbour’s garden). their noses, we will begin by modelling our res-

spring 2025 ⋅ 𝟫
 and redistribution of air—we obtain two one-
dimensional holes to get a double torus. The first
hole is between nose and mouth, and the second
is the same digestive hole we had previously.
On the other hand, when we modelled the diges-
tive tract, we assumed that both ends were open.
Obviously, humans do not have their mouths
open all the time. The mouth closing gives us
two separate container-style holes which we can
pull out and then we obtain a sphere!

Potatommy’s digestive system is equivalent to Another thing to take into account is that the
single hole. . . respiratory system and digestive system are con-
. . . and Potatommy’s respiratory system is equiva- nected. For example, the epiglottis (the flappy
lent to another single hole bit in your throat that stops you from inhaling
food) blocks the trachea and oesophagus from
being open at the same time. However, this only
happens when swallowing. First we will look at
the ‘resting’ state of a human.
As before, the lungs can be squished into the tube
connecting to the nose and mouth. This leaves us
with a single hole, but now it has three exits (the
nose, the mouth and the anus), rather than the
two exits we are used to seeing.

-piratory system as a tube from the nose to the

lung, as in the image. We will assume the nos-
tril is always open but the tube and lungs form a
closed path.
In this model, our lungs act in the same way as nasal cavity
the container hole in a mug. Then, as we did be-
fore, with our inflating, squishing and stretching
abilities, we can ‘pull’ them about so that we just
have a single, simple ‘breathing hole’.
However, this is an incredibly simplified and in-
accurate model of the respiratory system. Of
course, we can choose to breathe either through
tongue
our nose or mouth. Let’s explore how, if at all,
this complicates things. As we did previously,
let’s draw this inside our potato, and see what
oesophagus
trachea

we can squish and stretch it into.

Because we can first pop out our lungs and
squish them into the tube between the nose
Potatommy’s nasal and mouth passage
and mouth, then—after some more squishing

𝟣𝟢 ⋅ chalkdust
 feature

This changes the homotopy of our model in com-

parison to previous versions. For a moment, let’s
see how this further changes if we look at a ‘swal-
lowing’ potato head.
When we swallow, breathing temporarily stops,
as the epiglottis covers the trachea. This closes
the path from the outside world to the lungs,
so we don’t accidentally inhale our food. This
closing of the trachea by the epiglottis changes
our potato’s homotopy quite dramatically. The
lungs being cut off adds an empty sphere inside One dimensional holes on a hollow
our potato and we still have the two holes we torus

had previously: one from nose to mouth, and the

other from mouth to anus. This now gets us to a
point where it is difficult to count the holes and ber ring) has two one-dimensional holes and one
spaces we obtain inside the body. two-dimensional hole. Because one-dimensional
holes are circles, to count them, we can count
It is also important to note that the body is much how many different circular cuts we can make
more complex than the ‘through-roads’ we have to the object without creating two distinct pieces.
modelled. If we begin thinking about even just For two-dimensional holes, we can think of the
one part of the body in more detail, there are so number of cavities inside the object, or the num-
many extra things that we should take into ac- ber of punctures necessary for the air inside our
count, such as membranes and pores. In our ev- object to escape.
eryday definition of a hole, we would say that
pores are holes because sweat leaves them, and To classify the number of holes an object has,
we may sometimes describe pores as blocked. topologists use Betti numbers. The 𝑘 th Betti
There are millions of pores on your body and number, 𝐵𝑘 , tells us the number of 𝑘 -dimensional
we could not feasibly see these on a model of a holes on a topological surface. So for our hollow
human, especially if we are limited to this page torus, 𝐵𝟣 = 𝟤 and 𝐵𝟤 = 𝟣. We can see that 𝐵𝟣 = 𝟤
size and only 72 pages of the magazine. You on the image above, and 𝐵𝟤 = 𝟣 as there is only
can probably imagine some of the difficulties in one cavity inside the doughnut, and a puncture
this. Just going from an open mouth to a closed anywhere on the surface will cause the air to es-
one completely changed our homotopy equiva- cape.
lence, never mind thinking about—well, let’s not You may be surprised that the two-dimensional
go there. hole of a torus is not the hole in the centre of the
doughnut that we would (in everyday use) call
Betti numbers the doughnut hole. There is another name for
classifying these holes in a surface: genus.
You might have noticed that, although most
As another example, let’s calculate the Betti
of the holes we’ve considered so far were one-
numbers 𝐵𝟣 and 𝐵𝟤 of a hollow double torus. (Pic-
dimensional, there’s a sneaky two-dimensional
ture one of those two-person rubber rings you
‘cavity’ that represents the lungs when our
get on the really scary water slides.) We can see
potato head is swallowing. In fact, we can have
that the genus of the double torus is 𝟤 as there
holes in all sorts of different dimensions.
are two ‘doughnut holes’. It is also relatively easy
For example, a one dimensional hole is equiva- for us to determine that 𝐵𝟤 = 𝟣 for the double
lent to circle. A hollow doughnut (AKA a rub- torus, as a puncture anywhere on its surface will

spring 2025 ⋅ 𝟣𝟣
 age above. Cutting other combinations of loops
either results in fewer than four cuts or two dis-
tinct pieces.
An easier way to find 𝐵𝟣 , in this case, is to know
that for a closed orientable surface (such as a
Cutting these four one-dimensional double torus), 𝐵𝟣 = 𝟤𝑔 where 𝑔 is the genus of
holes on a double torus would give a
single piece...
a surface. So we quickly obtain that our genus-2
surface, the double torus, has 𝐵𝟣 = 𝟦.
...but cutting either of these would lead
to having fewer than four cuts or more Topologically equivalent spaces have the same
than one piece. Betti numbers. So, in an attempt to see whether
humans and doughnuts are ‘the same’, we can
compare the Betti numbers of both objects. If
they are the same, then we have some progress
to showing they are topologically equivalent—
but we don’t get all the way. This is because of
the important distinction that, although topolog-
ically equivalent objects always have the same
Betti numbers, it doesn’t work in the other direc-
make the air inside it escape. However, 𝐵𝟣 is a tion. In other words, having the same Betti num-
bit harder to find, as we need to find the max- bers is a necessary but not sufficient condition of
imum number of circular cuts we can make to topological equivalence.
our surface without making two distinct pieces.
Finally, let’s return to our ‘resting potato head’.
To do this, we can think of all the different types
If we make the slight change of closing the
of closed loops our surface can have.
potato heads mouth, we no longer have three
There are six different loops/circles on our double exits to our hole. With some slight squishing
torus. The maximum number of these we can cut and stretching, we obtain our potato head dough-
and still only have one piece is four. These are nut, or equivalently, a mug. Just promise me you
the two blue and two green loops in the top im- won’t start drinking coffee out of humans...

f CHALLENGE f f CHALLENGE f
Calculate the Betti numbers (𝐵𝟣 and 𝐵𝟤 ) of Make a hollow torus and a hollow double
the final potato head model. torus using paper and sticky tape.

Ashleigh Ratcliffe
Ashleigh is a PhD student and graduate teaching assistant at the University of Leicester. Her main
mathematical interests are in number theory. She is passionate about outreach and inclusion in
mathematics, volunteers as a Stem ambassador and is a representative for the Piscopia Initiative.
d linktr.ee/ashleighratcliffe c [email protected]
𝟣𝟤 ⋅ chalkdust
 news

NEWS
Weather this week: see 𝜈𝝂ν𝜈𝝂ν𝜈𝝂ν𝜈𝝂ν𝜈𝝂ν𝜈𝝂ν𝜈𝝂ν𝜈𝝂ν𝜈
CEEFAX page 196 𝝂ν𝜈𝝂ν𝜈𝝂ν𝜈𝝂ν𝜈𝝂ν𝜈𝝂ν𝜈𝝂ν𝜈𝝂ν𝜈𝝂

Wallpaper “totally ruined” Mathematicians discover that

after sofa carried around 𝟤𝟣𝟥𝟨𝟤𝟩𝟫𝟪𝟦𝟣 is one more than a
sharp corner prime number
In November, Korean mathematician Jineon In October, the Great Internet Mersenne
Baek proved that the area of the largest sofa that Prime Search discovered a new largest known
can be carried around a 90° corner is approx- prime number with over 41 million digits.
imately 2.21953, resolving a long-standing un- This latest discovery beats the previously
solved problem: the sofa problem. largest known 24 million digit prime number that
There are lots of shapes of sofa that will fit was discovered in 2018. By looking at when large
around a 90° corner in a corridor that is one unit prime numbers have been discovered and extrap-
wide, including these: olating, we are able to exclusively reveal that to-
day’s largest prime has 43 million digits and ends
in a 4.
1
number of
digits

𝟣𝟢𝟫
𝟣𝟢𝟪
𝟣𝟢𝟩
𝟣𝟢𝟨
The areas of these sofas are 1 and approxi- 𝟣𝟢𝟧
mately 1.57079. The sofa problem asks what the 𝟣𝟢𝟦
area of the largest sofa that will fit round this cor- 𝟣𝟢𝟥
𝟣𝟢𝟤
ner is.
𝟣𝟢𝟣
In 1992, Joseph L Gerver found this sofa that 𝟣𝟢𝟢 year
can fit around the corner and has an area of 1500 1600 1700 1800 1900 2000
2.21953:

Mathematicians fill YouTube

with 41 million numbers
It was larger than any other sofa that had Scottish mathematician Ayliean MacDonald
previously been found, but Gerver did not have a is leading a campaign to get people to say the en-
proof that a larger one could never be found. tire number out loud through her Say The Prime
For the next 32 years, no-one found a larger project.
sofa that could fit around the corner or a proof that By visiting d saytheprime.com, volunteers
Gerver’s sofa was the largest possible. Finally, in can claim a chunk of 419 digits to record them-
November, Baek’s proof that Gerver’s sofa was in- selves saying aloud and add the recordings to a
deed the largest possible was published on arΧiv. YouTube playlist. So far nearly 0.3% of the digits
DFS were unavailable for comment. of the prime have been said.

spring 2025 ⋅ 𝟣𝟥
 1 7 4 5
7 8 4 2
6 5
3 1 4 6 9 2
9 6 8 3 4 7
5 6 2
U N L O C K I N G
S U D O K U ’ S
S E C R E T S
Sara Logsdon looks to graph theory and
abstract algebra for help on the puzzle page

S
UDOKU has long captivated puzzle cal questions, graph theory has a long history of
enthusiasts worldwide with its logical famous thought-provoking problems.
challenges and addictive nature. While
One of the most well-known of these problems
it may seem like a simple game of numbers, be-
is the vertex colouring problem: Given an undi-
neath the surface lies a fascinating connection
rected graph 𝐺 = (𝑉 , 𝐸), where 𝑉 represents the
to the realm of mathematics. Graph theory and
set of vertices and 𝐸 represents the set of edges,
abstract algebra both play a crucial role in unrav-
can one find an assignment of colours to each
elling the intricacies of sudoku.
vertex in 𝑉 satisfying the condition that no two
Sudoku puzzles consist of a 𝟫×𝟫 grid, divided into adjacent vertices (connected by an edge) have
nine 𝟥 × 𝟥 sub-grids called regions. The objective the same colour?
is to fill the grid with numbers from 1 to 9, ensur-
ing that each row, column, and region contains
every digit exactly once.

A vertex colouring problem

In graph theory, a graph is a mathematical struc-
ture that comprises a set of vertices, or nodes,
connected by edges. An area rich in mathemati-

𝟣𝟦 ⋅ chalkdust
 feature

While sudoku may appear as a grid of numbers, It turns out that this question is equivalent to:
we can view it through the lens of graph theory. does there exist a solution to the original sudoku
Interestingly, we can represent a sudoku puzzle board? By applying vertex colouring algorithms,
as a graph, where each of the 𝟪𝟣 cells in the su- such as the greedy algorithm and/or backtracking,
doku grid corresponds to a vertex in the graph. we can systematically assign colours (numbers)
We label the vertices with ordered pairs (𝑥, 𝑦), 𝑥 to the vertices (cells) to find a valid solution to
and 𝑦 being integers from 𝟣 to 𝟫. We then join any sudoku board, so long as one exists.
two distinct vertices (𝑥, 𝑦) and (𝑥 ′ , 𝑦 ′ ) by an edge
if and only if any of these conditions apply:
• 𝑥 = 𝑥 ′ (the two cells are in the same row),
The greedy algorithm and
backtracking
• 𝑦 = 𝑦 ′ (same column), or
The purpose of the greedy algorithm is to pro-
𝑥 𝑥′ 𝑦 𝑦′
• ⌈ ⌉ = ⌈ ⌉ and ⌈ ⌉ = ⌈ ⌉ duce a vertex colouring of a graph. The algo-
𝟥 𝟥 𝟥 𝟥 rithm assigns colours to the vertices of the graph
(same 𝟥 × 𝟥 region). by iteratively selecting an uncoloured vertex and
So the top-left region of the sudoku board would assigning it the smallest possible colour that
be connected like this: does not conflict with its neighbouring nodes.
This means in sudoku, the greedy algorithm
𝟣, 𝟣 𝟣, 𝟤 𝟣, 𝟥 takes a sudoku board and systematically assigns
numbers (colours) to cells (vertices), ensuring
that no conflicts arise within the rows, columns
𝟤, 𝟣 𝟤, 𝟤 𝟤, 𝟥 or regions.
Backtracking is a systematic search algorithm
that explores the solution space by making
𝟥, 𝟣 𝟥, 𝟤 𝟥, 𝟥 choices and undoing them when they lead to con-
tradictions or dead ends.
We then propose the vertex colouring problem: We will combine the greedy algorithm with back-
Does there exist a 9-colouring of our sudoku graph? tracking, particularly when the greedy algorithm
That is, can we colour each vertex in the graph, fails to find a solution. In sudoku, this will in-
using no more than 𝟫 colours, in such a way that volve filling in cells with numbers, testing their
no vertices which are connected by an edge end validity, and if a contradiction is detected, ‘back-
up the same colour? tracking’ to the previous state to explore alterna-
tive choices until a valid solution is found or all
possibilities are exhausted.
The procedure for combining the greedy algo-
rithm and backtracking in sudoku follows:
1. Start with the initial board and select the
first uncoloured cell.
2. Assign the smallest possible number to the
selected cell.
3. Move to the next uncoloured cell and as-
sign the next smallest number.

spring 2025 ⋅ 𝟣𝟧
 4. Continue this process, moving from left Step 5.1: Conflict. The only remaining avail-
to right and top to bottom, assigning the able numbers for (𝟣, 𝟪) are already present in col-
smallest valid number to each uncoloured umn 𝟪.
cell.
1 2 7 3 4 5 8 6?9?
5. If a conflict arises, indicating an invalid
7 8 4 2
placement, backtrack to the previous cell
and select the next available number. 6 5
3 1 4 6 9 2
6. Repeat the process until the entire grid
is filled or until no valid number can be 9 6 8 3 4 7
placed. 5 6 2
3 6 4
2 1
The algorithm in action
4 8

Step 1: Suppose our initial board looks like this.

We select the first uncoloured cell: Steps 5.2–6: Backtrack. Return to cell (𝟣, 𝟩).
Enter the next lowest available valid number, 𝟫.
1 7 4 5 Resume greedy algorithm.
7 8 4 2
1 2 7 3 4 5 9 8 6
6 5
7 8 4 2
3 1 4 6 9 2
6 5
9 6 8 3 4 7
3 1 4 6 9 2
5 6 2
9 6 8 3 4 7
3 6 4
5 6 2
2 1
3 6 4
4 8
2 1
4 8
Steps 2–4: Greedy algorithm. Note that 𝟪 was
entered in (𝟣, 𝟩) instead of 𝟦, 𝟧 or 𝟨 because al- By continuing this same algorithm through all
though 𝟦 is the lowest available number, 𝟪 is the rows, we can reach a solution to our sudoku
lowest available valid number. board!
1 2 7 3 4 5 8 Other researchers seem to have been taken
7 8 4 2 by the same curiosity as me and have learnt
more about sudoku from graph theory—Joshua
6 5
Cooper and Anna Kirkpatrick linked minimal
3 1 4 6 9 2 sets to minimal fair puzzles, and Michael
9 6 8 3 4 7 Haythorpe connected Hamiltonian cycles to su-
5 6 2 doku puzzles of different sizes.
3 6 4 Graph theory is one way to unlock sudoku puz-
2 1 zles mathematically. But sudoku is really just
solving a puzzle subject to a number of con-
4 8
straints. That sounds like something algebra

𝟣𝟨 ⋅ chalkdust
 feature

could help us with. And it can! Algebraic geome- An ideal can be generated by a set of polynomi-
try possesses a similarly beautiful application to als just like a vector space can be spanned by a
sudoku. set of vectors. For example, if

𝐼 = {𝑎𝑓 + 𝑏𝑔 ∶ 𝑎, 𝑏 ∈ 𝑅},
Gröbner bases
then 𝑓 and 𝑔 generate 𝐼 , and we can write
If we could write the sudoku problem as a system
of polynomials, then solving the system would al- 𝐼 = ⟨𝑓 , 𝑔⟩.
low us to complete the puzzle. Solving simultane-
ous equations might seem like an opportunity for A term ordering is a method for choosing the or-
Gauss–Jordan elimination, but this method only der in which elements are placed. The term or-
applies to linear systems and we will see shortly dering that we’ll use here is the lexicographical
that ours is not. Moreover, we require our so- term ordering, abbreviated Lex. This is the order-
lutions to be the integers 1–9. A more general ing you’d expect in a dictionary. For example, if
approach lies in Grŏbner bases. we have variables 𝑥 , 𝑦 and 𝑧 then we order the
variables as
A Gröbner basis for a system of polynomials is a
𝑥 > 𝑦 > 𝑧.
new system of polynomials with the same solu-
tions as the original, but which is easier to solve. When we concatenate variables, we look at the
For example, the system first variable in each element and compare, then
the second, and so on. For example,
𝑥 + 𝑦 − 𝟥 = 𝟢,
𝑥𝑦 − 𝟤 = 𝟢 Lex(𝑥𝑦) > Lex(𝑦𝑧),
and Lex(𝑥 𝟤 ) > Lex(𝑥𝑦).
is coupled, as both equations involve both vari-
ables. However, computing a Gröbner basis of Given a term ordering, the leading term of a poly-
this system yields nomial 𝑓 is the term which is greatest in the or-
dering and will be denoted lt(𝑓 ). Given any set 𝑆
𝑦 𝟤 − 𝟥𝑦 + 𝟤 = 𝟢, of polynomials in a polynomial ring, we can then
𝑥 + 𝑦 − 𝟥 = 𝟢, define the leading term ideal of 𝑆 to be the ideal
lt(𝑆) generated by the leading terms of the poly-
where the first equation involves only 𝑦 . We can nomials in 𝑆. That is,
solve this first equation and then substitute to
find 𝑥 , making the system easier to solve. lt(𝑆) = ⟨lt(𝑓 ) ∶ 𝑓 ∈ 𝑆⟩.
In order to understand how Gröbner bases work, Note that the leading term ideal of a set of poly-
we will need to learn some abstract algebra ter- nomials is not always equal to the leading term
minology. ideal of the ideal generated by that set of polyno-
A polynomial ring is a set of polynomials in a cer- mials. For Gröbner bases we use degree reverse
tain number of variables. For our purposes, the lexicographic order. That means if 𝑆 = {𝑥, 𝑥 + 𝟣},
coefficients of polynomials will come from the lt(𝑆) = ⟨𝑥⟩, but if 𝐼 = ⟨𝑥, 𝑥 + 𝟣⟩, then
field ℚ of rational numbers.
𝑥 + 𝟣 − 𝑥 = 𝟣 ∈ 𝐼,
An ideal is a subset 𝐼 of elements from a ring 𝑅
that forms an additive group and has the prop- so lt(𝐼 ) = ⟨𝟣⟩ = 𝑅.
erty that
When lt(𝑆) = lt(𝐼 ), where 𝐼 = ⟨𝑆⟩, we say that 𝑆
𝑥 ∈ 𝑅 and 𝑦 ∈ 𝐼 ⟹ 𝑥𝑦 ∈ 𝐼 and 𝑦𝑥 ∈ 𝐼 . is a Gröbner basis for the ideal 𝐼 . In other words,

spring 2025 ⋅ 𝟣𝟩
 a set of nonzero polynomials

𝐺 = 𝑔 𝟣 , 𝑔 𝟤 , … , 𝑔𝑡 ⊂ 𝐼

is called a Gröbner basis for 𝐼 if and only if

lt(𝐺) = lt(𝐼 ).
Gröbner bases are nice because they simplify
problems. For example, if 𝐺 is a Gröbner basis
for an ideal 𝐼 , then we can use simple reduction
to determine whether or not a given polynomial Gert-Martin Greuel, CC BY-SA 2.0 de
𝑓 is in the ideal 𝐼 .
Bruno Buchberger introduced Gröbner
A well-known theorem in Gröbner basis theory bases in his 1965 PhD thesis and named
them after his supervisor, Wolfgang
explains that the Gröbner basis puts a given sys-
Gröbner
tem into triangular form. For example, if 𝐺 =
{𝑔𝟣 , … , 𝑔𝑠 } is a Gröbner basis in variables 𝑥𝟣 , … 𝑥𝑛
(𝑛 ⩽ 𝑠 ), the theorem says that we can order the for each cell in the sudoku. Formally, we are now
polynomials 𝑔𝟣 , … , 𝑔𝑠 so that 𝑔𝟣 only involves the dealing with the polynomial ring ℚ[𝑥𝟢 , … , 𝑥𝟪𝟢 ].
smallest variable 𝑥𝟣 , 𝑔𝟤 involves only 𝑥𝟣 and 𝑥𝟤 We wish to encode the conditions on the su-
and has leading term involving only 𝑥𝟤 , and so doku as a set of polynomials in some subfield
on. This makes solutions possible through quick, 𝐹 ⊂ ℚ[𝑥𝟢 , … , 𝑥𝟪𝟢 ].
easy substitution.
Cells can only take on whole number values from
A process known as Buchberger’s algorithm com- 1 to 9, so can we define the following polynomi-
putes the Gröbner basis for a given ideal and als for all 𝑖 = 𝟢, … , 𝟪𝟢:
term order. The algorithm uses the multivariate
division algorithm and least common multiples. (𝑥𝑖 − 𝟣)(𝑥𝑖 − 𝟤) ⋯ (𝑥𝑖 − 𝟫) = 𝟢.
This algorithm also guarantees the existence of
a Gröbner basis for a given ideal and term order. Next, we define polynomials to represent the
condition that each number can only be used
Gröbner basis in sudoku once in each column, each row and each region.
This can be done by defining that the sum of all
This is useful for the problem of sudoku. We can: columns/rows and of each region should be 45
and product should be 𝟫! = 362,880. With these
1. Create a system of polynomial equations
conditions we have no duplicate numbers. So if
to represent our sudoku problem
{𝑥𝑘𝟣 , … , 𝑥𝑘𝟫 } are all in the same row or column,
2. Use Buchberger’s algorithm to compute a
Gröbner basis for the ideal generated by 𝟫 𝟫
the polynomials in our system ∑ 𝑥𝑘𝑛 = 𝟦𝟧 and ∏ 𝑥𝑘𝑛 = 362,880.
𝑛=𝟣 𝑛=𝟣
3. If the puzzle has a unique solution, the
Gröbner basis will consist of 81 linear poly- Finally, since some of the cells 𝑥𝑗 in the grid are al-
nomials, from which we can read off the ready filled out with a number 𝑎𝑗 , we define new
solutions to our sudoku puzzle polynomials for each of these cells,

We can represent the constraints given in the 𝑥𝑗 − 𝑎𝑗 = 𝟢.

rules of sudoku as a system of polynomial equa-
tions. We introduce 81 variables, 𝑥𝟢 , … , 𝑥𝟪𝟢 , one These polynomials give a system of 135 equa-
tions, plus one for each known cell. We can solve

𝟣𝟪 ⋅ chalkdust
 feature

this system by applying Buchberger’s algorithm We therefore need to add the equations
to compute a Gröbner basis and from there, read
off the solution to the initial grid. 𝑎 − 𝟣 = 𝟢, 𝑓 − 𝟤 = 𝟢,
𝑖 − 𝟤 = 𝟢, 𝑘 − 𝟥 = 𝟢,
An example: shidoku 𝑝 − 𝟦 = 𝟢.

As we’ve seen, the algorithm is quite long. So in- All that remains is to give our equations to a com-
stead of using working through an example with puter algebra package—Matlab has the gbasis
a sudoku board, we will use a shidoku board. Shi- function—in order to perform Buchberger’s algo-
doku is a variation of sudoku which uses a 𝟦 × 𝟦 rithm and compute the Gröbner basis for this sys-
grid, divided into four 𝟤 × 𝟤 regions. tem.
Introduce 16 variables 𝑥𝟢 , … , 𝑥𝟣𝟧 to represent the
16 cells. Each cell can only take on the value 𝟣, 𝟤, In doing so, it returns:
𝟥 or 𝟦, so for 𝑖 = 𝟢, 𝟣, … , 𝟣𝟧,
𝑎 − 𝟣 = 𝟢, 𝑏 − 𝟥 = 𝟢, 𝑐 − 𝟦 = 𝟢, 𝑑 − 𝟤 = 𝟢,
(𝑥𝑖 − 𝟣)(𝑥𝑖 − 𝟤)(𝑥𝑖 − 𝟥)(𝑥𝑖 − 𝟦) = 𝟢. 𝑒 − 𝟦 = 𝟢, 𝑓 − 𝟤 = 𝟢, 𝑔 − 𝟣 = 𝟢, ℎ − 𝟥 = 𝟢,
Similar to before, the sum and product of all 𝑖 − 𝟤 = 𝟢, 𝑗 − 𝟦 = 𝟢, 𝑘 − 𝟥 = 𝟢, ℓ − 𝟣 = 𝟢,
numbers in a column, row or region must be 𝑚 − 𝟥 = 𝟢, 𝑛 − 𝟣 = 𝟢, 𝑜 − 𝟤 = 𝟢, 𝑝 − 𝟦 = 𝟢.
𝟣+𝟤+𝟥+𝟦 = 𝟣𝟢 and 𝟣×𝟤×𝟥×𝟦 = 𝟤𝟦, respectively.
With these conditions we have no duplicate num- This basis consists of a system of linear polyno-
bers. mials from which we can easily read off the solu-
tions to the shidoku board:
Let {𝑥𝑖 , 𝑥𝑗 , 𝑥𝑘 , 𝑥ℓ } represent a set of cells that make
up a row, column or 𝟤 × 𝟤 region, so
1 3 4 2
𝑥𝑖 + 𝑥𝑗 + 𝑥𝑘 + 𝑥ℓ − 𝟣𝟢 = 𝟢
4 2 1 3
and 𝑥𝑖 𝑥𝑗 𝑥𝑘 𝑥ℓ − 𝟤𝟦 = 𝟢.
2 4 3 1
This gives us a total of 40 polynomial equations.
3 1 2 4
We then add any values which are already given.
Let’s suppose we have the board below with the
given variable assignments: Sudoku intertwines powerfully with the realm of
both graph theory and abstract algebra. The al-
𝑎 𝑏 𝑐 𝑑 1 gorithms we learn in our courses provide valu-
able insights and strategies that can be applied
𝑒 𝑓 𝑔 ℎ 2
to conquer sudoku puzzles. So, the next time you
𝑖 𝑗 𝑘 ℓ 2 3 pick up a sudoku puzzle, remember this beautiful
𝑚 𝑛 𝑜 𝑝 layer of graphs and polynomial equations that
4
lies beneath its surface.

Sara Logsdon
Sara is an undergraduate student at the University of Georgia, interested in algebra.
Outside of math, she enjoys hiking and playing board games (her favourite is Splendor).

spring 2025 ⋅ 𝟣𝟫
 1 2 3 4 5
Cryptic crossword
6

7 8

9 10

11 12 13 14

15 16

17 18 19 20

21

22 23 24
#9, set by Kelpie

25 26 27

28 29

30

Across Down
1 Ordinary counting number. (7) 1 Type of analysis studies refashioned (9)
4 Basic statements about first accepted unknown (6) minute unclear line.
charged atoms gaining mass by losing neutron. 2 Spin tornado without hesitation. (5)
7 Divine blood from one refrain swallowing oxygen. (8)
3 Egyptian sun god starts to really (3)
8 A flat topped cone? Made from iron last. (3)
admire Helios.
9 Illinois elevated railways made from initially electric (3)
5 On the back page, put tea in them after (4,3,4)
locomotive siding.
contest.
11 I’m after initially tough review clip. (4)
14 A candle on fire. (6) 6 Kingly community is oldest scientific (5,7)
15 Mark spring cleaning instrument without initial (9) academy.
extension rod head for magazine. 10 A sailor’s drink without Romeo? (4)
16 Valley seen in shaping lens. (4) Shocking!
17 Endless untruths, Sophus? (3) 12 Sacrilege hiding upset. (4)
19 Centres of attention? An ellipse has two. (4) 13 This topologist’s an Irish son of a . . . (6)
20 Ode moving brood. (4) cartoon beer?
21 Units of area measurement augmented reality. (2) 18 I say, it is in the same source. (4)
22 Endless frost forms an edge. (3)
20 First-rate notation for a derivative. (5)
24 Outpouring from heads of joint European team. (3)
22 Field perhaps? found in superstring. (4)
25 Papers I initially devoured. (2)
26 Quite easily demonstrated leads the way for ending (3) 23 Vhat, Vhere, . . . ? Sounds like our (4)
proofs. favourite diagram creator.
28 Continuous or discrete? Transformation leaves an (9) 27 Sounds like I contained in Disneyesque (3)
object unchanged! facial feature.
30 Eco-friendly son of a baker. (5) 29 My uncle initially sounds like a cow. (2)

𝟤𝟢 ⋅ chalkdust
A symmetric universe

Sam Kay reflects on building

a universe

O
NE of the joys of studying the science expect certain equations to work as they do?
of the universe is using lots of lovely Most people would be satisfied giving the answer
equations that can tell you basically ‘they were introduced in my high school physics
anything you want to know. How far can I kick class, and therefore they work!’ I am not most
this football? How long will my phone battery people.
last? How many copies of Chalkdust can stop
this bullet? This article aims to answer the latter question
by unifying maths and physics via group theory.
A slightly more fundamental question might be Group theory is built on the idea of mathemat-
asking how all of these formulae came about. ical symmetries, whereas physics is the study
Perhaps even more fundamentally, why do we of describing and explaining observations using

spring 2025 ⋅ 𝟤𝟣
 equations. One might naturally ask what hap- tional symmetries and four reflectional symme-
pens when we observe a symmetry and want to tries. Now that we have discovered all these sym-
put some equations behind it. metries, we can collect them into a mathematical
group.

What is a symmetry? Groups in mathematics are collections of symme-

tries where combinations of symmetries should
The first of many beautiful theorems a math- also exist in the group. In the case of squares,
ematical physicist learns in their career is a rotation followed by a reflection is just an-
Noether’s theorem: any observed symmetry in other type of reflection. Evidently this is in the
your favourite physical system has a 1:1 cor- group, and so this condition is automatically sat-
respondence with a specific conservation law. isfied. We call a group a Lie group if it behaves
From this, we can build equations that describe smoothly.
the motion of everything in that system: fields,
particles, classical balls, you name it. The case of squares forms the dihedral group of
degree 4 (since a square has four corners), de-
But what is a symmetry? Fear not; this is not a
noted 𝐷𝟦 . In fact, the symmetries of any reg-
mathematical term we throw around lightly and
ular 𝑛-gon form the dihedral group of degree 𝑛,
try to link it to something completely unrelated—
𝐷𝑛 , which will always have 𝟤𝑛 elements. The di-
like rings or holes. A mathematical symmetry is a
hedral groups are unfortunately not Lie groups—
transformation that leaves an object completely
this can be seen by the fact that squares have cor-
unchanged. Take a simple square; what trans-
ners and the above transformations would not be
formations can you perform on this square that
smooth. The rotational symmetries of a circle,
leave it unchanged?
however, do form a Lie group: it is denoted U(𝟣).
You could rotate the square 90° and it would look
exactly the same. Rotating it another 90° leaves Enough about shapes though. We’re here to talk
it visually unchanged too. And the same for two about the universe. The universe exhibits sym-
more 90° rotations. This leaves us with four rota- metries of its own. In the context of physics,
tional symmetries of the square: these symmetries leave the laws of physics un-
changed. If you perform an experiment at one
place at a certain time, you should hope to obtain
square

square

square square the same observations at a different location at

a different time.

But there are more! If we also consider reflec-

tions of the square, there are four ways in which Classical symmetries
this can occur. This leaves us with four reflec-
tional symmetries: We would first like to interpret a physical sym-
metry in an empty universe; one in which we
assume no spacetime curvature or air resistance.
We know nothing about the laws that describe
erauqs

erauqs

erauqs square
motion in this space, but we should hope to find
them using our intuitions. Suppose I throw a ball
away from me while I am standing in one fixed
position. It will travel in a straight line, at a con-
Therefore, the collection of symmetries of the stant rate, never interfering with anything and
square has eight transformations: four rota- thus never stopping.

𝟤𝟤 ⋅ chalkdust
 feature

Throwing a ball from two platforms in an

otherwise empty universe

Gaining momentum on a swing

It should come as no surprise that if I move the
platform I am standing on to a new position and
repeat the experiment of throwing the ball away
from me, the same exact motions will occur: the system. Momentum cannot be created nor de-
ball will travel in a straight line, never stopping. stroyed and is a direct result of the spatial trans-
This tells us that the physical laws governing the lation symmetry of our empty universe.
movement of the ball, whatever they are, should In a similar fashion, this empty universe admits
be invariant under a spatial translation. That is, a rotational symmetry, which means the laws of
if I translate my initial position through space, physics are invariant when the entire system is
the ball still has precisely the same governing mo- rotated about an arbitrary angle. Imagine now
tion. you are spinning the ball on the end of a string;
almost as if it is orbiting your head, for example.
Another way of viewing this is by considering
No matter which orientation or which direction
each time step of the ball’s movement. A time
you decide to initially set the ball spinning, you
step is a regular measurement of time; whether
observe that it carries on spinning for all eternity.
that be a minute, an hour, an aeon, or just one
Remember, there is no gravity or air resistance.
unit. When I release the ball it is in its zeroth
The key point is that the equations of motion
time step. After one unit, it is in its first time step
governing the ball’s motion are the same at each
and has moved position according to its motion
point in its orbit.
law. The same thing happens for the second time
step. If we compare the first and second time To see this, we can use the same timestepping
steps, these are identical situations in which the argument. After spinning the ball, it enters its
ball has moved its position and nothing else. first time step and rotates according to its mo-
tion law. After the second time step, it obeys the
But remember what we said before: the equa-
same law and continues orbiting. As mentioned
tions of motion don’t change as the ball changes
above, the system admits a rotational symme-
position. So, at every point in time, the ball keeps
try, so having the ball in the second time step
its motion. This observation is what Noether
means it keeps its motion law from the first. At
identified as the conservation of momentum—any
every point in time, the ball continues spinning.
physical system with movement will have a to-
tal amount of momentum, usually labelled by Another conservation law has appeared—the con-
a number, and this number is kept the same servation of angular momentum. This is also rep-
throughout the entire time evolution of the resented by some number that is unchanged

spring 2025 ⋅ 𝟤𝟥
 f Polar coordinates
Any point (𝑥, 𝑦) on the plane can be represented the point
in polar coordinates (𝑟, 𝜃) measuring its absolute
distance from the origin and how far anticlock- (𝑥 ′ , 𝑦 ′ ) = (𝑟 cos(𝜃 + 𝛼), 𝑟 sin(𝜃 + 𝛼)).
wise it is rotated from the 𝑥 -axis. This makes
further rotation rather easy; to rotate by an arbi- 𝑦
trary angle 𝛼 , we transform the coordinates by
(𝑥 ′ , 𝑦 ′ )
(𝑟, 𝜃) ↦ (𝑟, 𝜃 + 𝛼).
(𝑥, 𝑦)
This can be noted by drawing a triangle and do-
ing some quick trigonometry to find 𝑟

𝑟 sin 𝜃
𝛼
𝜃
(𝑥, 𝑦) = (𝑟 cos 𝜃, 𝑟 sin 𝜃). 𝑟 cos 𝜃
𝑥

Then, by moving by some angle 𝛼 , we arrive at

trived some beautiful conservation laws that will

allow you to harness the true power of physics.
But alas, nobody will accept your claim without
a mathematical formulation. So we’re going to
do exactly that.

As with the dihedral groups laying out the sym-

metries of polygons, there is a group-based way
to represent the symmetries of what we should
Spinning basketballs at different angles call space. An easy example is via the two-
dimensional Cartesian plane, which will act as
throughout the time evolution of the system. our sandbox for this universe. If we have a fig-
Angular momentum cannot be created nor de- ure on it, whether that be a graph or tear marks,
stroyed and is a direct result of the rotation sym- we would like to preserve all of its angles and
metry of our empty universe. distances after rotating it. This is analogous to
taking our ball and spinning it around without
The main point here is that the translation sym- deforming it.
metry allows us to perform experiments in differ-
ent locations and obtain the same results, and an
Linear finite groups in mathematics are often de-
angular momentum symmetry lets us do these
scribed using matrices, which is helpful in our
experiments on a rotating object (like the Earth)
case since that’s what we’re working with. One
and obtain the same results.
can use some trigonometric identities to show
that rotation in Cartesian space is nothing but
Group actions matrix multiplication:

It turns out that you’re a genius; knowing noth- 𝑟 cos(𝜃 + 𝛼) cos 𝛼 − sin 𝛼 𝑟 cos 𝜃
ing about this universe, you have seemingly con- ( )=( )( ).
𝑟 sin(𝜃 + 𝛼) sin 𝛼 cos 𝛼 𝑟 sin 𝜃

𝟤𝟦 ⋅ chalkdust
 feature

If the laws of physics are to remain invariant af- All of the particles in our universe are quantum-
ter any rotation on this plane, then all possible oriented. In quantum mechanics, objects like
rotations (𝛼 going from 0° to 360°) are admitted. these particles are described by a quantity called
By putting all of these together we have uncov- its wavefunction 𝜓 (𝑥, 𝑦) that assigns a complex
ered the special orthogonal group SO(𝟤), denot- number to every point (𝑥, 𝑦) in two-dimensional
ing the group of rotations on a 2D system. This space. There are many different representations
group behaves just as nicely as the group of circu- of complex numbers, but for quantum mechanics
lar symmetries (for a somewhat obvious reason!), it is often useful to choose polar form where we
and we thus crown it a Lie group. interpret the radius as the wavefunction’s modu-
lus, |𝜓 |, and 𝜃 measures the phase of rotation. If
This also generalises to higher dimensions. We we relate this to clocks on a grid, one could imag-
have rotation groups in higher dimensions de- ine that the placement of the clock describes the
noted the ‘special orthogonal’ groups of degree 𝑛, value of |𝜓 | and the ‘time’ describes the phase.
SO(𝑛). These are Lie groups that consist of rota-
tions around the 𝑛 axes and all of their combina- It might then be sensible to suggest that af-
tions. In a physical sense, any of these rotations ter a local transformation, where we changed
can be undone via a rotation in the opposite di- the phase rotation differently at each point, we
rection. This kind of backwards rotation is an would change the physics we are trying to de-
inverse. scribe. But this is not true! This is only a math-
ematical representation of quantum physics and
therefore is not actually physical. The wavefunc-
Local transformations tion itself is unobservable and (most—we’re look-
ing at you, Aharonov–Bohm effect) physical pre-
Everything I’ve talked about so far is good for dictions only depend on |𝜓 |, which is left un-
mathematical theory. Secretly we were describ- changed by a local transformation.
ing global transformations: ones that affect ev-
This change-of-phase invariance is an intrinsic
ery point in spacetime in the same way. But
symmetry of the system. Mathematically, this
we’re here to describe interactions within the
system is locally U(𝟣)-invariant because the ac-
universe. For that, we need to understand local
tion of an element inside U(𝟣) (complex circular
transformations that depend on where you are
rotation) does not change the laws of physics.
in spacetime.

Why is that? The way we described conserved U(𝟣) = {ei𝛼 ∶ 𝛼 ∈ [𝟢°, 𝟥𝟨𝟢°]}.
quantities earlier assumed a really boring uni-
How might we describe this phenomenon math-
verse. If you want to put one of these quantities
ematically? All this time we have discussed the
to work by making it interact with other things
in nature, while still keeping it conserved, local
transformations are the way to do this. |𝜓 |

To paint an intuitive picture, imagine that we are

in our flat Cartesian land again and at each co-
ordinate we place a clock, or dial, tuned to some 𝑥𝜇
time that depends on that specific coordinate. A
global transformation could, for example, turn
the dials of each clock exactly 90°—or turn time
by three hours. A local transformation would in- +𝛼
stead turn one clock by three hours, another by
A wavefunction changing its phase.
five, and perhaps even one back by four.

spring 2025 ⋅ 𝟤𝟧
 motion of objects in flatspace, and motion is gov- The term 𝒙 or 𝑥 𝜇 represents the 4-
erned by derivatives. For observable physics we dimensional vector (𝑥 𝟢 , 𝑥 𝟣 , 𝑥 𝟤 , 𝑥 𝟥 ), in
have to make sure that the derivatives them- which the numbers are used as coor-
selves obey a local U(𝟣) transformation so that dinate indexing as opposed to powers.
we can include them in our equations. One of These values represent a point in space
these transformations on 𝜓 looks like ei𝛼(𝒙) 𝜓 , and time and are sometimes written as
where one should note that 𝛼(𝒙) is parametrised (𝑐𝑡, 𝑥, 𝑦, 𝑧), where 𝑐 is the speed of light.
because it is a local transformation. When we The shorthand 𝜕𝜇 is used for 𝜕/𝜕𝑥 𝜇 , the
differentiate, partial derivative with respect to each
coordinate.
𝜕𝜇 (ei𝛼(𝒙) 𝜓 ) = ei𝛼(𝒙) (𝜕𝜇 + i𝜕𝜇 𝛼(𝒙)) 𝜓 ,

which tells us that differentiation does not obey

local U(𝟣) transformations. This is an issue be-
cause one would really like to describe the mo-
magnetic quantities 𝑬 and 𝑩. For those inter-
tion of particles, and this requires derivatives.
ested, the metric signature here is (−, +, +, +)
But we cannot do this with local transformations
and the speed of light has been normalised.
alone. A change in physics must be made.
Maxwell’s equations of electromagnetism can be
formed by taking derivatives of various compo-
nents of this tensor in a neat way.
Charge conservation
We’ve seen the general rule of thumb for how Maxwell’s equations predict that light itself is
local U(𝟣) transformations give structure to cer- an electromagnetic wave. It propagates through
tain mathematical objects: wavefunctions be- all of spacetime in little packets we like to call
have nicely, but their dynamics do not. We photons. Because of this, we can describe them
should probably also come up with a name for with wavefunctions. And we already know that
this theory that isn’t ‘locally U(𝟣)-symmetric a change of wavefunction phase does not af-
mathematical physics’. To fix this, we need to fect physics: it seems as though U(𝟣) trans-
dive into the existing theory of electromagnetism. formations and electromagnetism are linked.
But Maxwell’s equations involve derivatives—
In early days of scientific experimentation, so- and derivatives, as we have seen, are not locally
ciety thought that the theories of electricity U(𝟣) invariant: 𝜕𝜇 𝜓 ↦̸ ei𝛼(𝒙) 𝜕𝜇 𝜓 . There is an ex-
and magnetism were completely separate. An tra term involving the derivative of 𝛼(𝒙) that we
astounding result was formulated in the late would like to remove. What we need to do is in-
19th century, when Maxwell and notable others troduce a mathematical tool that, under a local
proved that these were two aspects of the same U(𝟣) transformation, would transform as this ex-
formulae. How do we understand this? As with tra term:
previous reasoning, we can package it in a matrix. 𝐴𝜇 ↦ 𝐴𝜇 + 𝜕𝜇 𝛼(𝒙).
In 1908, Minkowski introduced the field strength
This 𝐴𝜇 is a mathematical tool that mediates the
tensor 𝐹𝜇𝜈 as
electromagnetic field and allows us to describe
other vector fields, like 𝑬 and 𝑩, using its local ro-
𝟢 −𝐸𝟣 −𝐸𝟤 −𝐸𝟥
⎛ ⎞ tation properties. We can call it the electromag-
𝐸𝟣 𝟢 𝐵𝟥 −𝐵𝟤
𝐹𝜇𝜈 ≔ ⎜ ⎟, netic potential. Some may call it the photon field,
⎜𝐸𝟤 −𝐵𝟥 𝟢 𝐵𝟣 ⎟
as it naturally arises when discussing photons.
⎝𝐸𝟥 𝐵𝟤 −𝐵𝟣 𝟢 ⎠

that encodes information about electric and Instead of using the standard derivative, we now

𝟤𝟨 ⋅ chalkdust
 feature

use this difference 𝜕𝜇 − i𝐴𝜇 which transforms as More specifically, the standard model is the uni-
fication of the Lie groups
𝜕𝜇 𝜓 − i𝐴𝜇 𝜓 ↦ ei𝛼(𝒙) (𝜕𝜇 𝜓 − i𝐴𝜇 𝜓 ) .
U(𝟣) × SU(𝟤) × SU(𝟥),
Result! With just group transformations, com-
plex numbers and derivatives, we have discov- with the special unitary groups of degrees 2 and
ered something wonderful. Real-world physics 3 now included. We already saw that U(𝟣) could
gives us 𝑬 and 𝑩, and the maths behind it uses be thought of as rotation around a unit circle
tools such as the electromagnetic potential. but is physically charge conservation. Similarly,
actions of SU(𝟤) are movements along the sur-
This is where Noether’s theorem comes into play, face of a 4D hypersphere, and actions of SU(𝟥)
specifically the second part that deals with lo- are movements along the surface of some other
cal transformations. Remember that any sym- hyperobject. These are physically the conserva-
metry in a physical system has a determined con- tion of weak isospin and colour charge, thanks
servation law. By observation, a local U(𝟣) sym- to Noether.
metry requires the interaction with some poten-
tial field 𝐴𝜇 , which was the field encoding elec- When combined, the standard model as a whole
tric and magnetic information. But in order to exhibits charge–parity–time (CPT) symmetry.
interact with the electromagnetic field, our par- The action of swapping all particles for their an-
ticles must have some charge associated with tiparticles, exchanging chirality, and reversing
them. From there, to ensure quantum fluctua- time keeps the standard model invariant—but
tions don’t spin out of control, we must have con- only when done together. There also exists the
servation of electric charge. concept of symmetry breaking, whereby some
particular symmetries make some massless par-
This immediately gives us a physical representa- ticles (like the photon) massive. They have to be
tion of what a U(𝟣) transformation is: the sym- fixed in a nontrivial way.
metry of charge. Dirac in 1928 used this to say
These three groups mediate the electromagnetic
that for every charged particle, there must ex-
force and the strong and weak nuclear forces
ist its antiparticle counterpart with the opposite
among all known particles. Because of that, the
charge. Therefore, electric charge cannot be cre-
equation for their dynamics is huge: it contains
ated nor destroyed and is a direct result of the
93 terms. You can buy it on a T-shirt at Cern.
U(𝟣) symmetry of our particles.
The story, however, does not stop there. At the
time of writing, there is a huge amount of re-
Where do we go from here? search going into a new type of symmetry called
supersymmetry, which has the potential to relate
The standard model is one of the greatest feats of all types of quantum particles together in a su-
mathematical physics in the 20th century. It is pergroovy way. This symmetry will present us
the best quantum description we have of the fun- with the conservation of supercharge (no, I am
damental forces that govern our entire universe. not making that up). We haven’t yet measured
The best part? It is built by imposing certain sym- it, but when and if we do, it is bound to change
metries on quantum particles. the course of mathematical physics forever.

Sam Kay
Sam is a master’s student at Durham University, interested in supersymmetry and symmetry
breaking. He plays in a superfunk band and hosts the superb Chalkboard Ultra podcast.
c [email protected] l chalkboardultra t @sameljk.bsky.social
spring 2025 ⋅ 𝟤𝟩
 2025
LADY (GA)2 f THE 52 × 79 f DUAL IPA
KENDRICK LAMINAR FLOW
GIR L ‘P OW ER ’ S TAG E

AREA-NA GRANDE f LAURENT SWIFT

CHARLI ∫ x dx f CHAPPELL CONE f AΔ
Δ
ARITHMETIC FRANKLIN f KITE BUSH f ×s ALOUD
TR IG O NO M ET RY S TAG E
arctan Monkeys f Donny cos-mond f tan Morrison
Leonard cosh-en f Frank sin-atra f Wu-tan Clan
SOH-phie Ellis-Bextor f CAHa f TOAsis
DAD R O C K S TAG E
STATUS QUO-TIENT f IRON MATRIX f QED ZEPPELIN f RADIALHEAD
LATEX PISTOLS f RINGS OF LEON f χSER CHIEFS
A✓
C/D✓
C f JOY ÷ f THE VELVET UNDETERMINED SYSTEM OF EQUATIONS

TTR AM ID STAG E
BILLIE πLISH f πNK FLOYD f πJ HARVEY f π BLOODY VALENTINE
πRE STRAITS f NEIL πMOND f THE πXIES f πONEL RICHIE f πLIE MINOGUE
M AT H EM AT IC IANS S TAG E
POISSON-NY AND CHER f PAS-CALVIN HARRIS f RICK PYTHAGOR-ASTLEY
SWEDISH GAUSS MAFIA f ARCHIBEE GEES f FERMETALLICA

OT H ER JO K ES W E LIK ED S TAG E
CHALKING HEADS f SCOUTING FOR WOMEN IN STEM
LILY FRACTALLEN f Dexy s Midnight Runners f THE POINT-ER SISTERS f NEUTRAL MILK ASYMPTOT-EL
L
(PANIC + 1) AT THE DISCO f AB=BA f KEHA Missing $ inserted. <inserted text> $ l.96 ^^I ?

GET YOUR TICKETS NOW

𝟤𝟪 ⋅ chalkdust
 23 24 25 26

T
28

O
29 30

H O W
32 33

R
34 35

I
37 38 39

T
A C R O S S N U M B E R
40

Fun’s word-cross puzzle, written by Arthur Wynne

in 1913. Each clue gives the position of the first and
last letters where the entry should be written. The
puzzle features well-known words such as ‘neif’
and ‘nard’, and the solution includes ‘dove’ twice.

1
F U N
Matthew Scroggs explains how 2 3
to get digits all riled up 4 32 5
6 7 8 9

F
10 11 12 13
OR ten years now, I’ve been setting the 14 15 16 17
Chalkdust crossnumber. Over this time, 18 19 20 21
I’ve developed a lot of tricks and tools 22 23 24 25
for setting good puzzles.
26 33 27
Although puzzles involving words in grids of 28 29
squares have been around since at least the 30 31
1800s, the first crossword (or word-cross as the 34
author called it) puzzle was published in the New
2–3 What bargain hunters 10–18 The fibre of the gomuti
York World newspaper in 1913. Since this, cross- enjoy. palm.
words have become a staple in newspapers and 4–5 A written 6–22 What we all should be.
magazines around the world. acknowledgment. 4–26 A day dream.
6–7 Such and nothing more. 2–11 A talon.
10–11 A bird. 19–28 A pigeon.
In the UK, cryptic crosswords (see page 20) were 14–15 Opposed to less. F–7 Part of your head.
invented and grew in popularity in the 1900s and 18–19 What this puzzle is. 23–30 A river in Russia.
22–23 An animal of prey. 1–32 To govern.
now appear in most newspapers and magazines. 26–27 The close of a day. 33–34 An aromatic plant.
The cryptic crossword in the Listener magazine 28–29 To elude. N–8 A fist.
30–31 The plural of is. 24–31 To agree with.
became infamous as the hardest such puzzle. the 8–9 To cultivate. 3–12 Part of a ship.
Listener ceased publication in 1991, but the Lis- 12–13 A bar of wood or iron. 20–29 One.
tener crossword still lives on and is now pub- 16–17 What artists learn to do. 5–27 Exchanging.
20–21 Fastened. 9–25 Sunk in mud.
lished on Saturdays in the Times newspaper. 24–25 Found on the seashore. 13–21 A boy.

spring 2025 ⋅ 𝟤𝟫
 In early 2015, we were discussing ideas for regu-
lar content in our brand new maths magazine,
and wanted to include a more mathematical
crossword-style puzzle. Like Venus emerging
from the shell, the crossnumber was born.
Of course, we didn’t invent the crossnumber puz-
Grids with order 2 (left) and order 4 (right) rotational zle: the first known crossnumber puzzle was writ-
symmetry. ten by Henry Dudeney and published by Strand
Grids with one, two, and four lines of symmetry. Magazine in 1926; four times a year, the Lis-
tener features a numerical puzzle; and the UKMT
(United Kingdom Mathematics Trust) have in-
cluded a crossnumber as part of their team chal-
lenge for many years. There are also plenty of
other publications that include crossnumbers, in-
cluding the very enjoyable Crossnumbers Quar-
terly, that’s been publishing collections of the
puzzles four times a year since 2016. But perhaps
we’ll be mentioned in a footnote in a book about
the history of puzzles.
But anyway, we’d decided we needed a cross-
number, so I needed to write one...

Making a grid
The first step when creating a puzzle is to cre-
ate the grid. Like many publications, we restrict
ourselves to using grids of squares (for the main
crossnumber at least—we allow more freedom in
other puzzles that we feature).
Typically, a publication will impose some restric-
tions on the placement of the black squares in its
The grid for crossnumber #14 experimented with
translational symmetry.
puzzle. The most popular restrictions are:
• The white squares must be simply con-
nected (for any two white squares, it is pos-
sible to draw a path between them that
only goes through white squares);
• The arrangement of black squares must be
in some way symmetric;
• The proportion of squares that are black
cannot be too high.
There are two forms of symmetry that are used in
crossword grids: rotational symmetry and reflec-
tional symmetry. Both order 2 (rotate the grid

𝟥𝟢 ⋅ chalkdust
 feature

180° and it looks the same) and order 4 (rotate the

grid 90° and it looks the same) rotational symme-
try are commonly used in crossword grids, with
order 2 rotational symmetry often being the min-
imal allowable amount of symmetry. You’ll want
to be careful when making a grid with order 4 ro-
tational symmetry, as it’s very easy to make your
grid into an accidental swastika.
With a few notable exceptions, the grids for the
Chalkdust crossnumber have some form of sym-
metry, and also follow the other two restrictions.
We do, however, allow some flexibility on these
restrictions if it allows us to use an interesting
grid. For crossnumber #14 (shown on the left), The grid for crossnumber #11. Rotating this grid 90°
leads to the same grid with inverted colours. This
we experimented with translational symmetry, grid also has order 2 rotational symmetry.
leading to a grid that was not simply connected
and with a greater than usual proportion of black The grid for crossnumber #6. Rotating this grid
180° leads to the same grid with inverted colours.
squares. For crossnumbers #11 and #6, the grids
had a nice property: rotating the grid leads to
the same pattern of squares with the colours in-
verted. Once again we had to break the require-
ments on connectedness and the proportion of
black squares to do this.
For crossnumber #17, we skipped imposing sym-
metry entirely and instead formed all 18 pen-
tominoes from the black squares in the grid. This
really pleased me, as there were clues in the puz-
zle related to the number of pentominoes and the
number of black squares in the grid.
For American style crosswords, there’s an ad-
ditional restriction that is imposed: all white
squares must be checked. A white square is
called ‘checked’ if it is part of an across entry
and a down entry—and so you can fill that square
in by solving one of two different clues. Due to
this, American crosswords will have large rect-
angles composed entirely of white squares and
never have lines of alternating black and white
squares as commonly seen in British puzzles.
When writing a crossword, it is common to pick
the words to include while making the grid, as
trying to find valid words or phrases to fill a pre-
determined grid is a challenging task. (Thank-
fully, there’s software out there that can help you
The grid for crossnumber #17 was not symmetric,
make grids from a list of words.) For crossnum- and instead included all 18 pentominoes in black.

spring 2025 ⋅ 𝟥𝟣
 bers, filling the grid is a much easier task as For crosswords, it is also common to disallow the
any string of digits not starting with a zero is use of two-letter words. For the crossnumber,
a valid entry. This ease of filling the grid is removing two-digit numbers would remove the
what allowed us to use the interesting restriction- potential for a lot of fun number puzzles, so we
breaking grids mentioned in this section. don’t impose this restriction here. We allow two-
letter words in the Chalkdust cryptic too, as they
can be really useful when trying to make a grid
with our additional restriction that the majority
of the included words should be related to maths.
1 2 3

4 5
Setting the clues
6 7
Once I’ve made the grid for a crossnumber, the
8 9 next task is to write the clues.

10
Often, I start this task by picking a fun math-
ematical or logic puzzle to include in the clues.
Sometimes, this is a single clue, such as this one
Across Down from crossnumber #1:
1 Add up. 1 English sept.
4 Spanish mister. 2 French one. Down
6 Roman four. 3 Most common. 6 This number’s first digit tells you how (10)
7 Non-continuous finite element. 4 opp ÷ hyp. many 0s are in this number, the
8 Country with non-rectangular 5 Rugeley Trent second digit how many 1s, the third
flag. Valley. digit how many 2s, and so on.
10 Footballer’s zero. 9 Greek p.
Or this clue from crossnumber #5:
A valid American crossword (if two letter entries
were allowed) Across
9 A number 𝑎 such that the equation (2)
A valid British crossword that is not a valid American 𝟥𝑥 𝟤 + 𝑎𝑥 + 𝟩𝟧 has a repeated root.
crossword

Other times, this could be a set of clues that refer

1 2
to each other and reveal enough information to
work out what one of the entries should be, such
3
as these clues from crossnumber #10:

Across
4 13 49A reversed. (3)
37 The difference between 49A and 13A. (3)
47 37A reversed. (3)
48 The sum of 47A and 37A. (4)
Across Down
3 A crowd. 1 Line across a circle. 49 Each digit of this number (except the (3)
4 Polyhedron leaked by 2 Animals in a gaggle. first) is (strictly) less than the
Snowden. previous digit.

𝟥𝟤 ⋅ chalkdust
 feature

Once I’ve included a few sets of clues like this, As long as I spot clues like this, it can usually be
it’s time to write the rest of the clues. As fixed with some rewording. In this example, I’d
any crossnumber solvers will have noticed, my rewrite the clues as either ‘2D plus a factor of 6.’
favourite type of clue to add from this point on or ‘This number is a factor of the sum of 6 and
is a clue that refers to another entry. 2D.’
More recently, I’ve begun adding an additional In my time setting the crossnumber, I’ve got a
mechanic to each crossnumber. This started in lot better at spotting ambiguity in clues, and do
crossnumber #13, when all the clues involved two this by reading through the clues and trying to
conditions which were joined by an and, or, xor, be a mega-pedant and intentionally misinterpret
nand, nor or xnor connective. Mechanics in them. It can be really helpful to get someone else
later puzzles have included the clues being given to help with this check though, as remembering
in a random order without clue numbers (#14), what you intended to mean when writing a clue
some clues being false (#16), and each clue be- can make it hard to read them critically.
ing satisfied by both the entry and the entry re-
versed (#19). I really hope that you enjoy the ‘fun’
mechanic I used in this issue’s puzzle. Checking uniqueness
Around the same time as I started playing with Perhaps the most difficult part of setting a cross-
additional mechanics, my taste in puzzles shifted. number is checking that there is exactly one so-
Older crossnumbers had been quite computa- lution to the completed puzzle.
tional, and often needed some programming for
To help with this task, I’ve written a load of
a few of the clues, but more recently I have be-
Python code to help me find all the solutions to
come a greater fan of logic puzzles and number
the puzzle. I run this a lot while writing clues
puzzles that can be solved by hand. To reflect
to make sure there’s no area of the puzzle where
the change in the type of puzzle I was setting we
I’ve left multiple options for a digit. I intention-
added the phrase ‘but no programming should
ally use a lot of brute force in this code so that it’s
be necessary to solve the puzzle’ to the instruc-
really good at catching situations where there
tions, starting with crossnumber #14.
are multiple answers to a puzzle where I only
found one solution by hand.
Thinking like a mega-pedant Once I’ve got all the clues and my code says the
One of the most important things to watch out solution is unique, I do a full solve of the puzzle
from when writing and checking clues is acciden- by hand. This is both to confirm that the code’s
tal ambiguity due to writing maths in words. conclusion was not due to a bug, and to check
that the difficulty of the puzzle is reasonable.
For example, the clue ‘A factor of 6 more than 2D’
could be read in two ways: this could be asking Following this checking, and a little proofread-
the solver to add 6 to 2D, then find a factor of the ing, the puzzle is ready for publishing. Then the
result; or it could be asking the solver to add 1, 2, fun part begins, as I get to chill with a nice cup of
3, or 6 (ie a factor of 6) to 2D. tea and wait for people to submit their answers.

Matthew Scroggs
Matthew is a research software engineer in the Advanced Research Computing Centre (ARC) at
University College London (UCL). His hobbies include pretending to be called Humbug (H).
d mscroggs.co.uk t @mscroggs.co.uk n @[email protected] r mscroggs
spring 2025 ⋅ 𝟥𝟥
 Oπnions: In defence of toy models
Calum Ross gets on his soap box

M
ANY people think of mathematics plicit solutions to systems of equations, and who
as this incredibly austere and ab- want some physical, biological, or computer sci-
stract subject, and mathematicians ence motivation behind a problem to consider it
as people who are obsessed with abstraction and interesting and worth pursuing. And this is not
divorced from reality. It is true that mathematics just true of applied mathematics. Many pure
can be abstract, and some mathematicians are mathematicians also love constructing explicit
quite proud of the separation between what they examples when testing out ideas, such as Dame
see as ‘real’ mathematics and ‘useful’ mathemat- Alison Etheridge (see pages 3–6). If you are going
ics, cough Hardy cough. To some of them being to spend a lot of time and effort trying to tackle
able to show that an integral is finite, or that a so- a problem, you want to know if you are likely
lution exists to a system of equations, is enough. to be able to solve it rather than simply banging
They do not actually feel the need to sit down your head against a brick wall. So we make ap-
and evaluate the integral, or solve the equations. proximations and simplifications to construct a
To them the existence of the answer is the impor- toy model preserving the key mathematical and
tant thing, not actually constructing it. physical features but providing a sandbox where
we can test our ideas.
However, in my opinion this characterisation of
mathematics as purely an abstract approach is While there are people who study abstract foun-
an unfair critique. There are plenty of mathe- dations like quantifying the concept of ‘oneness’
maticians who evaluate integrals, construct ex- or, allegedly, spending hundreds of pages to es-
tablish that 𝟣 + 𝟣 = 𝟤, many mathematicians
are interested in understanding problems mo-
tivated by studying the mathematics of fluids,
modelling the spread of viruses, or—in my case—
effective models of special magnetic materials.
As a mathematical physicist, much of what I do
is to study mathematical problems and questions
motivated by real-world physical systems.

One such model uses a one-dimensional chain of

atomic spins to understand magnetic materials.
In this case we forget about the structure of an
individual atom and just think of it as an arrow
that can point in any direction. Next we come up
with a rule for how neighbouring atoms interact,
eg if they want to point in the same direction we
Not that kind of toy model . . .
call it a ferromagnet or in opposite directions it

𝟥𝟦 ⋅ chalkdust
 O πnions

is called an antiferromagnet. One approximation mathematics. In building them you need to think
that we make is that any given atom only cares about what the key features are that you want
about its nearest atoms and not the effect of all the model to describe, and how they can be de-
the other atoms which are further away. Finally scribed mathematically. A well constructed toy
we say how the atoms interact with a magnetic model will be sophisticated enough to preserve
field or if they prefer to line up with a particular the qualitative, and some of the quantitative, fea-
direction and off we go. We can understand the tures we want, but be simple enough that we can
transition between different magnetic domains really get our teeth into it. It will enable us to
without having to study a complicated model in- gain intuition about how changing the parame-
corporating all the details, and I can calculate ters of the model change the results that you get
this on the back of an envelope without having out. In the best cases, this intuition will help you
to touch a computer. to understand the full situation. From SIR mod-
els in epidemiology to effective field theories in
physics, toy models are put to good use in many
0 1 2 3 4 5 6 7 places.

A spin chain of atoms modelling a domain wall in What about in more abstract settings? Again, we
a ferromagnet. can study a simpler object to help us gain under-
standing. Want to understand something about
These simplified toy models can come under crit- groups? Study matrix groups first: you can lever-
icism from two sides. Proponents of austere ab- age your linear algebra experience to understand
straction will attack them for lacking mathemat- how to formulate the problem and identify po-
ical rigour or for involving unfounded leaps of tential subtleties. In a way this is what a power
logic, while those who are all about applications series or Fourier series expansion of a function is
can bemoan the time spent simplifying a prob- doing. You have a function with some properties
lem to still attack it with, potentially, compli- that are easy to understand and others that are
cated mathematics. Why not just study it nu- harder to understand, and approximating it by a
merically on a computer? There is definitely power series gives you something more tractable
merit to this argument; many toy models can where you can do some quick computations to
be and are studied numerically. However, a nu- check what is going on.
merical solution does not convey the same in-
I guess that all I am trying to say is next time
tuition as a more analytic approach. Yes, if I
someone says that they are studying a simpli-
solve a differential equation numerically I can
fied situation or playing with a toy model, do
play with the parameters and see how the solu-
not dismiss them out of hand. Remember toy
tion changes. However, this will often not tell us
models can convey intuition and understanding
why the change is happening. A good toy model,
beyond their humble formulation. Maybe when
like our spin chain, can convey this.
you are stuck on a particularly difficult problem
To me, a toy model is more than just a simpli- you should think about what simplifications you
fied setting to play around with some interesting could make and try one out for yourself.

Calum Ross
Calum is a lecturer in physics at Edge Hill University where he studies topological solitons and
enjoys talking about maths and physics to anyone who will listen. He can often be found on the
side of a mountain or out on a bike, talking to himself about solitons.
d cdross1.github.io c [email protected]
spring 2025 ⋅ 𝟥𝟧
 Moonlighting agony uncle Professor Dirichlet answers your
personal problems. Want the prof’s help?
Contact c [email protected]

Dear Dirichlet,
Wedding season is coming up
and I realise I need to get my suit
justed’, shall we say. My local slightly ‘ad-
high street has a few places whi
vices, but I’m hoping to get it don ch offer ser-
e as economically as possible.
I ask for? Wh at should
— Mervin Miles, London NW3

■ Have no fear! In my experience, it’s best

DIRICHLET SAYS:
to enlist two or more establishments. For your first
purchase, ask the couturier to make straight line cuts
only. Take all symmetric pieces (the jacket front panels,
the trousers, the collar) to a second establishment, but
leave all antisymmetric pieces (the buttonholes, the
breast pocket) at the first. Combine the first and
second order and you should have a well-fitted tailor
expansion. (If it still feels a bit tight, you could
always fork out for a higher-order correction.)

Dear Dirichlet, m. This has

and I need to furnish the living roo
I’m moving to a new apartment ions seem to
I can’t decide on a sofa. My opt
given me two problems. Firstly, more and
SCS they all start to blur and get
go on forever. With every trip to
more similar. Any advice? own
— Eloise, on the (con)verge of a breakd

■ You have a classic couch-y sequence. It’s

DIRICHLET SAYS:
a good sign that your options seem to be merging together!
That means you’ve made the right design choice; they’ll
make your space complete.

𝟥𝟨 ⋅ chalkdust
 dea r d ir ic het

Dear Dirichlet,
I make furniture out of wood, but lately the raw material has been
arriving in un-
manageable shapes! Should I send it back or make do?
— Sabine Carpenter, North-East Mathematical Carpentry Society

■ DIRICHLET SAYS: You will need to trim the pieces yourself.

It is inevitable that complex logs require branch cuts.

Luckily, I can buy the scrap pieces off you. I want to
make small (pointlike) wooden sculptures for my new retro
nightclub, Disco ’Ntinuous. It’s almost ready to open, I
just need a DJ that will get the crowd jumping.

Dear Dirichlet,
I recently bought a LLM (large
language mare) from my good
He told me he’d trained her on a friend Shady Bob.
really sophisticated dataset, and
be able to do any calculation I that she would
asked of her. The horse handle
fine, if a bit slowly. But when I s bas ic arithmetic
ask her to help me to come up
Chalkdust article, she just stares wit h a title for my
at me blankly. Where am I going
wrong?!
— Chat Gee-gee-T, Stafford

■ DIRICHLET SAYS: Ah, it sounds like you’ve been saddled with

a one-trick pony. If you want to speed up multiplication,

you’ll need more horse power - enough to get some stable
calculations going. You could even set up a breeding pro-
gramme; many-foaled calculations tend to move much faster.
But for your Chalkdust title, it’s bad news, I’m afraid.
Your horse calculator will manage simple maths, but it will
never be able to generate sophisticated jokes: you can
teach a horse to quarter, but you can’t make it think.

Dear Dirichlet,
r problem—even
Sorry, me again with the sofa questions. I realise I have anothe
if I choose one, I don’t know if it will fit in my living room. Help!
— Eloise , from page 36

A settee is measurable if
DIRICHLET SAYS: A ⊂ IR, µ*(A) = µ*(A ∩ E) + µ*(A
A
■
c
∩ E ). From there, use a ruler.

spring 2025 ⋅ 𝟥𝟩
 We gave this article to ChatGPT and told it to create a header image

What a load of noise:

A beginner’s mathematical
guide to AI diffusion models
Madeleine Hall is overwhelmed

T
HESE two little vowels have infiltrated Most of the time, you don’t even notice it. It’s
daily conversation, and we’re not talk- just there making life a little smoother. But, as AI
ing about a decent draw from the gets smarter people are asking bigger questions.
Scrabble bag. From geopolitical debates to ev- Will it change the way we work? Is it sustain-
eryday chitchats—AI is everywhere, and people able? Can we trust it? Whether you’re excited,
won’t shut up about it. worried or disillusioned, one thing’s for sure: the
conversation is just getting started.
Who can blame them? It’s in your pocket when
you ask your phone for the weather, on your I cannot keep up. The state of the art, the current
commute when GPS (shout out to Gladys from highest-level of performance achieved in a partic-
issue 15’s significant figures) finds the best route, ular domain, is continuously shifting. New mod-
and in your inbox when your email filters out els are constantly being released, outperform-
spam. AI helps pick your playlist, suggests what ing predecessors as new techniques are discov-
to watch after a long day, and powers the chat- ered. Trying to chase and understand the ever-
bot answering questions when you shop online. changing best feels like fighting a losing battle.

𝟥𝟪 ⋅ chalkdust
 feature

But AI isn’t going anywhere. You could spend ous, is that at the heart of these tools are key
a lifetime learning, researching, using and argu- ideas from probability theory and stochastic pro-
ing about it (increasingly many people are). So cesses.
where do you start? I myself, as a trendy individ-
ual, am choosing to start with tackling one of the
hottest AI tools of the season: diffusion models.
The process trio: forward,
reverse, and sampling
AI, NN, DL,... anyone for
Scrabble? Diffusion models have three main components: a
forward process, which systematically adds noise
Under the umbrella term ‘artificial intelligence’, to an image over multiple steps; a reverse process,
which broadly means any intelligence exhibited where a neural network learns to gradually re-
by machines, a technique that’s shot to fame is move this noise; and a sampling process, where
neural networks. Neural networks (NNs) are a the trained model generates new images by start-
category of algorithms made up of layers of inter- ing from random noise and iteratively refining it.
connected nodes. They are designed to recognise
patterns and, once trained, can recognise those
patterns in fresh data. A sub-category of these
are deep learning algorithms, which are neural Forward: adding noise
networks with many layers—typically more than
three (I guess for computer scientists, 𝟥 counts as Let’s start with an initial image, 𝑥𝟢 . This could
‘many’). be an image of a frog, or a cat, or a frog on a cat.
Mathematically, we consider 𝑥𝟢 as a single sam-
Deep learning (DL) underpins much of today’s
ple from a broader distribution, 𝑥𝟢 ∼ 𝑞(𝑥). This
trendiest AI. One of the most successful tech-
could be the distribution of cat images, frog im-
niques is the use of large language models, specifi-
ages, or a blend of the two. Our ultimate goal is
cally a generative (capable of producing new con-
to model 𝑞(𝑥) so that we can generate new im-
tent) pre-trained (trained on a huge dataset be-
ages from the same underlying distribution as
fore being fine-tuned for specific tasks and con-
what we’ve started with.
ditioned on user inputs) transformer (an architec-
ture processing text as tokens and applying self- To begin, we take our image and start corrupting
attention mechanisms to understand context). it by adding a little noise. After the first step, the
Slap a chatbot interface on top of your generative
pre-trained transformer, along with some safety
training and reinforcement learning, and bam—
you’ve got yourself ChatGPT, which can gener-
ate text bountifully.
Other generative AI tools you may have played
with are Dall-E 2 and Stable Diffusion, which
generate images. These tools are powered by dif-
fusion models, which work very differently from
transformers like ChatGPT. Instead of predicting
the next word in a sequence, diffusion models
start with pure randomness and gradually refine
it into a meaningful image.
The great news, for the mathematically curi- “You haven’t seen a cat around here, have you?”

spring 2025 ⋅ 𝟥𝟫
 from one step to the next.
The hyperparameters 𝛽𝑡 are often called the noise
schedule. These are the only parameters we pro-
vide the forward process, as we directly compute
𝛼𝑡 from 𝛽𝑡 . In a sense, whilst 𝛽 represents how
much noise we’re adding at each step, 𝛼 repre-
sents how much of the original image to keep.
If 𝛽𝑡 is too large in the early steps, the image
quickly becomes dominated by noise, making it
Do not adjust your set—after 𝑇 steps, our cute difficult for the neural network in the reverse pro-
travelling frogs have become indistinguishable cess to learn meaningful structure. On the other
from noise.
hand, if 𝛽𝑡 is too small the corruption process
is too slow, requiring more steps and computa-
tional resources.
image with a bit of noise added is given by:
A well-chosen noise schedule ensures that each
step is neither drastic nor trivial. A typical strat-
𝑥𝟣 = √𝟣 − 𝛽𝟣 𝑥𝟢 + √𝛽𝟣 𝜀𝟣 ,
egy is using a linear or quadratic increase in 𝛽𝑡 , so
where 𝜀𝟣 is randomly drawn from a standard that early steps retain more information, while
Gaussian distribution, 𝜀𝟣 ∼ N (𝟢, 𝟣). later steps gradually transition to full noise.

We repeat this multiple times, adding more noise

at each step, and define 𝛼𝑡 ≔ 𝟣 − 𝛽𝑡 to simplify Reverse: removing noise
notation downstream:
If the forward process was about drowning our
𝑥𝑡 = √𝛼𝑡 𝑥𝑡−𝟣 + √𝛽𝑡 𝜀𝑡 . image in noise, the reverse process is about fish-
ing it back out. However, because the noise is
By the time we’ve done this many times, say 𝑇 randomly sampled at each step, there isn’t a sin-
steps in total, the image 𝑥𝑇 is indistinguishable gle deterministic path to simply rewind the pro-
from random noise. cess. Fortunately, we have neural networks up
our sleeve. We can train a neural network to ap-
The fixed constants 𝛽𝟣 , ..., 𝛽𝑇 ∈ (𝟢, 𝟣) vary across
proximate the most likely reverse step, predict-
the time steps, which allows for fine control over
ing how much noise was added at each stage.
the process. We choose these hyperparameters
ahead of training our model. Smaller values of In the forward process, we defined the transition
𝛽 give a gentler nudge towards chaos, since this kernel 𝑞(𝑥𝑡 |𝑥𝑡−𝟣 ). Ideally, we would simply invert
keeps more of the image from the previous step. this to get 𝑞(𝑥𝑡−𝟣 |𝑥𝑡 ), the probability of undoing a
step of noise. However, calculating this directly
In terms of 𝑞(𝑥), because the noise added at each
is tricky because it depends on the entire data
step are IID samples drawn from N (𝟢, 𝟣), we
distribution.
write:
Instead, we approximate it using a learnable re-
𝑞(𝑥𝑡 |𝑥𝑡−𝟣 ) = N (𝑥𝑡 |√𝛼𝑡 𝑥𝑡−𝟣 , 𝛽𝑡 𝐼 ). verse transition kernel, 𝑝𝜃 (𝑥𝑡−𝟣 |𝑥𝑡 ), parametrised
by a neural network:
Since each step only depends on the previous
one, this forms what’s known as a Markov pro- 𝑝𝜃 (𝑥𝑡−𝟣 |𝑥𝑡 ) = N (𝑥𝑡−𝟣 |𝜇𝜃 (𝑥𝑡 , 𝑡), 𝜎𝑡𝟤 ).
cess. The function 𝑞(𝑥𝑡 |𝑥𝑡−𝟣 ) is called the transi-
tion kernel, as it defines the probability of moving This means, rather than inverting the forward

𝟦𝟢 ⋅ chalkdust
 feature

process exactly, we train a neural network to es-

timate the most likely denoised image 𝜇𝜃 (𝑥𝑡 , 𝑡) at
each step. The noise added in each step of the
forward process is Gaussian, so we can reason-
ably assume the reverse process is also Gaussian.

It’s common to assume the variance 𝜎𝑡𝟤 is fixed,

allowing us to focus on learning the mean
𝜇𝜃 (𝑥𝑡 , 𝑡). Therefore, the reverse step is given by:
Adobe Firefly
𝑥𝑡−𝟣 = 𝜇𝜃 (𝑥𝑡 , 𝑡) + 𝜎𝑡 𝜀,
After training, we can start with pure noise and
apply a sequence of denoising functions to add
where 𝜎𝑡 ≔ √𝟣 − 𝛼 𝑡 is the (known or learned)
back cat- and frogness. Perfect.
standard deviation at step 𝑡 , and 𝛼 𝑡 = 𝛼𝟣 𝛼𝟤 ...𝛼𝑡
(the product of all the 𝛼 ’s up to and including 𝑡 ).
to our image at each step, we can use this knowl-
To train our model, we optimise a loss function edge to generate new images.
that ensures our learned mean 𝜇𝜃 (𝑥𝑡 , 𝑡) correctly
predicts the noise 𝜀𝑡 that was added at each step. Once the reverse process is trained, we can use
The training objective is: it to generate new images from scratch. We be-
gin with a random noise sample, 𝑥𝑇 ∼ N (𝟢, 𝟣),
𝐿 = 𝔼𝑞 [‖𝜀𝑡 − 𝜀𝜃 (𝑥𝑡 , 𝑡)‖𝟤 ] , and iteratively apply the learned denoising func-
tion. Step by step, the noise is reduced, revealing
where 𝜀𝜃 (𝑥𝑡 , 𝑡) is the neural network’s prediction an image that resembles samples from 𝑞(𝑥)—be
of the noise. By minimising this loss function, we that a cat, or a frog, or a blend of the two!
ensure our model can accurately peel back the
This iterative denoising process is how models
noise layer by layer during the reverse process.
like Stable Diffusion and Dall-E 2 generate high-
In short, the reverse process trains a neural net- quality images, starting from pure noise.
work that takes inputs 𝑥𝑡 and 𝑡 to learn the best
𝜃 such that its output 𝜀𝜃 (𝑥𝑡 , 𝑡) best reflects the
amount of noise added to our image at step 𝑡 .
It’s just a little noise
While the nuts and bolts used in AI will continue
to evolve, these underlying principles will remain
fundamental. By diving into some of the maths
Sampling: creating new of diffusion models, I hope this gives you a solid
images foundation to follow (and maybe even contribute
to) the next breakthroughs in AI. And if you still
Because, through the reverse process, we’ve feel overwhelmed, well, at least now you know
learned the size and structure of the noise added it’s just a little noise.

Madeleine Hall
Madeleine Hall is a mathematical consultant at Smith Institute. She likes writing, open water
swimming, the Oxford comma, and tHiS mEmE. Her PhD research was on optimal swimming for
microorganisms. Unfortunately, she is a macroorganism, so has found none of her results of any
use in the lido.
d madeleinehall.github.io
spring 2025 ⋅ 𝟦𝟣
 PUZZLES
Looking for a fun puzzle but not got time to tackle
the crossnumber? You’re on the right page.

Extra letters
The words on the right are anagrams of words with a common theme with an extra letter added. If you
write the themed words in the boxes to the left, and the extra letters in the extra letters column, two
more words with the same theme will appear in the orange boxes.

Extra
letters

ART LEGION
CHANGE OX
AGENT COD
CERAMIC LIES
NOG BLOG
AROUSE Q
LIL NEEPS

Mathematical songs
Guess the song titles from the mathematical clues. The artists are given in brackets under each song.

3867 ÷ 5000
𝟥 (Lionel Richie)
(Sabrina Carpenter)
𝑦 −𝑐 𝑣 −𝑢
𝑚𝑥 + 𝑐, , 𝟤𝑥 d𝑥 − 𝑥 𝟤 ,
𝑒 𝑥 − 𝑒 −𝑥 𝑥 ∫ 𝑡
(Village People)
𝟤
(Take That)

{𝑥 ∶ 𝑥 knows} ∩ {𝑦 ∶ 𝑦 is not a deity} = ∅

(The Beach Boys) (Tears for Fears)

𝟣
∀thing ∈ {𝑥 ∶ |𝑥| < 𝜖} 𝟤
(Blink 182) ∫ song 𝟥𝑥 d𝑥
−𝟣
(Blur)

𝟦𝟤 ⋅ chalkdust
 puzzles

One or two
Put the answers to the clues in the grid by placing either one or two
letters in each box. 1 2 3

c on e
For example, if the answers to the clues were cone,
speed, cusp and ended, the completed puzzle would u nd
look like this: sp e ed 4

Across Down
1 𝐵 in 𝐴 ⊂ 𝐵. 1 𝐴 in 𝐴 ⊂ 𝐵. 5
4 Two triangles whose an- 2 Angle measurement unit.
gles are the same. 3 10 more than 5A.
5 10 less than 3D.

3 0 3 3 3
9 4 6 1 9
9
Half nine
8 0
6 1 2 Cross out one digit in each square so that the
three five-digit numbers in each direction are
1 3 5 3 0
all multiples of nine.
3 6 0 6 2
1 2 1
0 0 4
3 1 3 5 9
6 7 4 8 7

Arrange the digits

Put the numbers 1 to 9 (using each number exactly once) in the boxes so that the sums are correct. The
sums should be read left to right and top to bottom ignoring the usual order of operations. For example,
𝟦 + 𝟥 × 𝟤 is 14, not 10.

− + =𝟥 × + = 𝟤𝟣

+ × − × ÷ −
× − =𝟫 − − =𝟢

× + + ÷ − +
÷ ÷ =𝟣 × × = 𝟤𝟪

= = = = = =
32 32 2 4 1 11

spring 2025 ⋅ 𝟦𝟥
𝟦𝟦 ⋅ chalkdust
 ◀
ON THE COVER

Voronoi diagrams
and generative art
Hayden Mankin combines
traditional art with randomness

G
ENERATIVE art is all about design-
ing systems and letting those systems
create the art. Different generative
artists might define it in their own ways. Some
see the system itself as the art, while others view
the generated pieces as the art. Some might con-
sider themselves the artist, while others might
see the computer as the artist. There’s also a
group I would call ‘parametric artists’, where peo-
ple take an existing generative project and tweak
the parameters until they get the desired output.
I see generative art as a unique blend of tradi-
tional art concepts with an element of random-
ness that’s hard to replicate with traditional me-
dia. It’s a balancing act between algorithmic
precision and adding enough unpredictability to
keep things interesting.
Curation is a big part of generative art as well.
I often write programs that run overnight, gen-
erating images. In the morning, I sift through
them to find the ones I like, sometimes using
randomised parameters in the filenames to make
sorting easier.
Some might want to compare it to AI art, but
that is different to generative art. Recently, gen-
erative AI has become more common, and many
people are calling their work ‘generative art’.
While I think these are two different things, I un-
derstand why AI art falls under the same term. ‘Here’s what you could’ve won’: Some more of
Both involve working with generative systems, Hayden’s art.
but it’s important to distinguish between them.

spring 2025 ⋅ 𝟦𝟧
 We might start using new terms like procedural inspired by any particular thing or work, but I
or algorithmic artists for traditional generative wanted to pick a topic that comes up in a lot of
artists, and AI artists, parametric artists or prompt fields, or even in a lot of different generative art
engineers for those working with AI. But for now, projects. I wanted to make something where I
they all fall under the umbrella of generative art. wasn’t combining too many different ideas, that
way the main focus of Voronoi diagrams could
be seen easily.
About the artist
While my Instagram page has been fairly inac-
I’ve always had a passion for digital art and using tive since I started teaching and attending grad
computers in creative ways, exploring systems school, I hope to become more active in sharing
that behave uniquely. My journey into coding my work. My future aspirations include contin-
began in primary school when my older brother uing to explore the intersection of art and tech-
introduced me to GameMaker. From there, I nology, creating more generative art projects and
started experimenting with fun little projects, of- inspiring others to find creativity in coding.
ten creating old-style roguelike games in my free
time.
My shift to generative art happened in sec- Voronoi diagrams
ondary school when I discovered Daniel Shiff-
Voronoi diagrams are one of the more commonly
man’s YouTube channel. I learned about the Pro-
used tools in a generative artist’s toolbox. Gener-
cessing library for Java and began making simu-
ative artists often aim to mimic organic patterns,
lations. My friends and I would sit in the back
and Voronoi diagrams are excellent for this pur-
of our physics class, writing simulations for each
pose. They can approximate many natural forms,
new concept we learned. Eventually, I started
from the patterns on a giraffe’s coat to the struc-
exploring more abstract routes, creating simula-
ture of plant cells. I find it fascinating whenever
tions related to gravity and objects interacting
we see such close relationships between maths
with each other. My first true generative art
and nature.
projects involved making strange attractors and
re-implementing flocking algorithms. A Voronoi diagram is a way of dividing space
into regions based on distance to a specific set
of points. Each point has a corresponding re-
gion consisting of all locations closer to that
point than to any other. These regions are called
Voronoi cells.

Given a set of points, the Voronoi diagram can

be constructed like so:

• Start with a set of points;

The Lorenz attractor is perhaps the best-
known example of a strange attractor. • For each point, determine the region of
space that is closer to it than to any other
point. This involves finding the perpen-
Over time, I discovered many generative artists dicular bisectors of the lines between each
who continue to inspire me, including Ben Ko- pair of points;
vach, Daniel Shiffman and Tyler Hobbs. Their
talks and pieces available online have been a sig- • The intersection of these bisectors forms
nificant source of inspiration. This is not really the edges of the Voronoi cells.

𝟦𝟨 ⋅ chalkdust
 feature

Additionally, I’m showcasing another concept on

the cover closely related to Voronoi diagrams:
Delaunay triangulations. This is a way of con-
necting a set of points to form triangles such that
no point is inside the circumcircle of any triangle.

Given a set of points, the Delaunay triangulation

can be constructed like so:

• Start with a set of points;

• Connect the points to form triangles such

that the circumcircle of each triangle does
not contain any other points.

A Delaunay circle is the circumcircle of a trian-

gle in the Delaunay triangulation. It is the circle
that passes through all three vertices of the tri-
angle. The property of Delaunay triangulation
ensures that no other points lie inside these cir-
cumcircles.

The Voronoi diagram and Delaunay triangula-

tion are duals of each other. This means that
the vertices of the Voronoi diagram correspond
to the faces of the Delaunay triangulation and
vice versa. You can construct the Voronoi dia-
gram from the Delaunay triangulation by con-
necting the circumcentres of the Delaunay trian-
gles.

These diagrams have significant applications in

various fields, including computer science, geog-
raphy, biology and urban planning. They help
solve proximity problems, such as finding the Example of Delaunay constructions.
nearest neighbour, and are used in spatial analy-
sis, map overlay and network planning. In com-
puter graphics, Voronoi diagrams can be used to Behind the scenes
generate realistic textures and simulate natural
phenomena. Most of my projects follow an iterative process. I
start by writing code that defines a system, then
Voronoi patterns are found in nature, such as in make random changes, experiment with differ-
the distribution of seeds in a sunflower or the ent colour schemes, and tweak parameters to
pattern of cracks in drying mud. Some of my achieve the desired outcome. Here, I use p5.js,
favourite example use cases include shattering a JavaScript version of the Processing library,
patterns, creating paths that avoid obstacles, and which allows me to add buttons and interactive
even applying machine learning techniques like elements easily. This makes it simple to tweak
𝑘 -means clustering. and modify the artwork live.

spring 2025 ⋅ 𝟦𝟩
 Without blur or threshold, some blur but a threshold at 0, and finally some blur and threshold applied.

Adding blur and threshold distorts the dual

a threshold to layers. It takes a buffer and the
graph
name of the layer as inputs. The function first
applies a Gaussian blur, then iterates through
each pixel to adjust the alpha value based on a
threshold. This process helps generate more nat-
ural curves in the drawing.
In graphics programming, we often work with
a one-dimensional array of pixels. Each pixel
is represented by four consecutive values (red,
green, blue, 𝛼 ); by iterating through this array
and adjusting the 𝛼 values, we can create smooth
transitions and define edges more clearly.
I am sure there are much faster ways to do this
with shaders, but I tend to stick to whatever tools
are built in to what I am using. We can see the
effects of this by increasing the blur amount for
our Voronoi edge layer (see above). This allows
us to get much more organic shapes.
This can be applied to many different layers
in different ways, including the triangulation.
Since the Delaunay triangulation is the dual
graph of our Voronoi diagram, the vertices
should be inside each tile. Additionally, the ver-
tices should be darker and less likely to disap-
pear when we blur and threshold them. This will
In this project I draw a few separate layers and cause star-like shapes to form inside each tile
stack them on top of each other to make a com- (see left).
posite image. It takes a set of points to gener-
The same can be done for other layers until we
ate the Voronoi diagram and then takes it from
have a cohesive image. I also colour in the
there.
tiles randomly, pulling from a predefined colour
The software allows me to apply a blur effect and scheme.

𝟦𝟪 ⋅ chalkdust
 feature

Adding some colour to the tiles and a drop

shadow completes the work

The last thing I do is add a drop shadow to make

some layers stand out a bit or make it more in-
teresting. The shadows are made by copying the
layer, blurring it, and setting the RGB values to
be a consistent colour. Once the settings are fine,
I generate a few iterations with the same settings,
then pick the ones I like best. The variations you
see to the right are all generated with the same
parameters.
Variations all generated with the same param-
eters as the example on the left.

Hayden Mankin
Hayden is a master’s student at Ohio State University, where he studies career and technical
education. He also teaches a career tech program in Ohio, and spends his summers running the
Stem department at a sleepaway camp in Pennsylvania.

spring 2025 ⋅ 𝟦𝟫
 Calculators:
phone or
physical?
PH YS ICA L! PH O NE !
argues ELINO R FLAVE LL argues SOPHIE MACLEAN

Ah. . . the humble calculator. Firstly, before I put forward my case, I

must dispute the phrasing of this argu-
The mathematical chic accessory. Whip a ment. Phone calculators vs ‘physical cal-
calculator out at random intervals, and it will culators’? Phone calculators are physical
do to you what taking off glasses did to girls calculators. To say otherwise is pure in-
in 90s rom-coms. tellectual snobbery. Especially now that
phone calculators are often equipped with
No phone calculator can compare! Ask your-
the functions of scientific calculators.
self: How often do you reach for your calcu-
lator? ‘All the time’ I hear you say. Because: But semantics aside, a phone calculator is
orders of magnitude (which you can work
• it always has charge,
out easily on this little app) better than a
• you don’t need to upgrade it every 18 handheld calculator. It can do everything
months, and a handheld can do, and more. Why would
• it is very hard to scratch (as my calcu- one opt for a device with less functional-
lator can bear testament to). ity, over the one that they carry around
anyway?! If I were to start using a hand-
Calculators are the one friend that cock- held calculator, I would still be carrying
roaches will have in the apocalypse. my phone (and therefore phone calculator)
with me anyway.
In comparison, how often do you reach for
the calculator on your phone, try to work I am not exaggerating to say that I have not
something out, and then sigh because it is used my handheld calculator since my fi-
just not easy to get the answers you are look- nal A-level exam. Outside of the totalitar-
ing for? How often do you have to Google ian simulation of real life that is the English
how to write sin 𝑥 on your phone calculator?
−𝟣
school system, I am either capable of do-
‘All the time’ you say! That sounds rather like ing the necessary calculations in my head,
a phone calculator isn’t up to the job… already have my phone on me, or am in
need of a tool much more powerful than
For me it’s a no brainer. The best calculator a handheld calculator can give.
is the one that never runs out of battery, al-
ways works, and makes you look like a nerd.
Why use anything else? Stay strong to the
calculator.

𝟧𝟢 ⋅ chalkdust
 I thought I might
Write a poem tonight
But I’ve nothing to say
— Andrew Stacey

Universal poetry

Andrew Stacey explores the

mathematics within poetry

I
N her fabulous book Once Upon a Prime, There are lots of things to count when it comes to
Sarah Hart shows us myriad ways in which rhymes. The first that probably comes to mind
mathematics and literature interact. One is to count the number of rhyming patterns of a
of the most straightforward of these is structure: given length.
the mathematics defines the shape and literature
Rhyming patterns are described using a letter for
fills it. And one of the simplest examples of struc-
each line. Each distinct letter denotes a distinct
ture is in the rhyming patterns of poems.
rhyme, so that lines labelled with the same let-
The doggerel above the title is something I’m go- ter are meant to rhyme. So for couplets we get
ing to call a universal poem. By that I mean that AA (both rhyme) and AB (no rhyme), while my
it contains every possible rhyming pattern of a poem above has the rhyming pattern AAB. There
particular length, which in this case is couplets. are five rhyming patterns for triplets, the full list
There are only two such patterns: either the sec- being AAA, AAB, ABA, ABB, and ABC. It might seem
ond line rhymes with the first or it doesn’t. In my more logical to label the fourth of these as BAA,
poem above, the first two lines rhyme giving the but rhyming patterns always start with A and
first pattern, and the last two lines don’t rhyme, then use the next available letter when introduc-
giving the second pattern. It’s also the shortest— ing a new rhyme. The sequence of counts of
in number of lines—universal poem for couplets. rhyming patterns is in the Online Encyclope-

spring 2025 ⋅ 𝟧𝟣
 dia of Integer Sequences (OEIS) as sequence What is a couplet?
A000110 and starts 1, 2, 5, 15, 52. The problem
of counting rhyming patterns is a nice exercise If it has been a little while since you
and can lead to some interesting combinatorics. have engaged with poetry, a rhyming
However, that’s not what this article is about. couplet is two successive (but not nec-
What I want to consider is the idea of finding all essarily consecutive) lines that rhyme.
the rhyming patterns of a given length within a Rhyming triplets are then three suc-
single poem, which I’m calling a universal poem. cessive lines that rhyme. You could
This leads me to a couple of questions: carry on and write poetry with four,
five or even 𝑛 successive lines that
rhyme!
What is the shortest universal poem
for a given length of rhyming pattern?

It’s obvious that we can always find a poem,

since we can just write a poem with stanzas
(that is, verses) of the given length in which each The poetry of de Bruijn
stanza has a different rhyming scheme.
sequences
But by overlapping, as in my opening opus, we
can do better. Next question: To unlock the structure of a universal poem we
need to know about de Bruijn sequences. These
are a fascinating piece of mathematics that can
How many distinct such
poems are there?
be illustrated by something that many of us en-
counter on a daily basis: key codes, such as those
that unlock doors.
I could change my opening ode to:
Many places have these key-coded locked doors
where you have to enter a code on a pad to un-
I thought I might lock the door. There are two types of these: one
Write a poem today where you have to enter the code and then press
But I’ve nothing to say an enter key to submit it, and one where as soon
as you’ve entered the code, the door unlocks im-
mediately.
and this flips the rhyming scheme to ABB.
Suppose you have a door secured by a four-digit
In one sense this is a new universal poem struc-
numerical code. For the first type of lock, to
ture, but as we’ll see later, it can also be consid-
go through all possibilities it would take up to
ered to be equivalent to the original.
𝟦 × 𝟣𝟢𝟦 key presses. (There are 𝟣𝟢𝟦 combinations
Incidentally, by fairly trivial changes I can actu- and each requires 4 key presses—we’ll ignore the
ally get every rhyming scheme for triplets from enter button for this.) For the second type, how-
this poem: ever, it turns out we can do a bit better.
Not a lot, but a bit.
I thought I (could|might|may)
Let’s reduce the numbers somewhat and sup-
Write a poem to(night|day) pose it is a three-digit code but with only two
But I’ve nothing to say buttons, say 0 and 1. So there are eight possible
codes: 000 to 111. That’s 𝟥×𝟤𝟥 = 𝟤𝟦 key presses to
Truly, it is an amazing piece of art! go through them all. But if it were the type that

𝟧𝟤 ⋅ chalkdust
 feature

opens as soon as the code is complete, we could down the node label of the starting node, and
run through the lot with just 10 key presses: the then append the edge labels as we go round the
sequence 0001110100 contains all eight possible path.
codes within it and so will unlock such a door.
Before you get too set on a life of opening doors, 1
this reduces the number of possibilities from 𝟥 ×
1
𝟤𝟥 to 𝟤𝟥 + 𝟤, so our more common door lock re- 01 11
duces from 𝟦 × 𝟣𝟢𝟦 to 𝟣𝟢𝟦 + 𝟥. Still quite a big
number. 0

Such a sequence is called a de Bruijn sequence. I 1 0

first learnt of them from the mystical book Mag- 1
ical Mathematics by Persi Diaconis and Ron Gra-
ham. 00 10
0
I expect you can now see where I’m going: can 0
we find a rhyming pattern that is a de Bruijn pat-
tern for all rhymes of a given length?
Starting from 00 we can find an Euler path by
following 0 to remain at 00, then 1 to 01, again 1
Eulerian poetry to 11, again 1 to remain at 11, then 0 to 10, fol-
There is a standard technique for finding de lowed by 1 to 01 and 0 back to 10, then finally 0
Bruijn sequences which is very well explained in to return to 00.
Diaconis and Graham’s book. The core of it is This produces: 0001110100, as above.
to set up a certain graph and look for an Euler
path (or cycle) of that graph. As a reminder, an In fact, since this Euler path is actually a cycle,
Eulerian path goes along every edge of a graph the last two 00s can be viewed as the same as
while an Eulerian cycle goes along every edge of the first two and so if we join the start to the
a graph starting and ending at the same node. end we can think of it as an eight digit code
00011101 which contains all 3 digit codes provid-
The graph works like this. We start with a se- ing we think of it as a cyclical list. I’ll call this a
quence of codes that we want to find a de Bruijn de Bruijn cycle.
sequence for, say the eight codes 000 to 111. We
strip off the last character of each and look at the The length of any de Bruijn cycle or sequence of
set of remaining prefixes (so, ignore duplicates): the graph is completely determined by the struc-
in this case, just 00, 01, 10 and 11. These are the ture of the graph. That of a cycle is the number
nodes in our graph. of edges in the graph, and that of a sequence is
the number of edges plus the length of a node
The edges are then the possible suffixes at each label.
node. In this case just 0 and 1 at each node. The
target of an edge is found as follows: append the
edge label to the node label to form a code, and Universal poetry
then remove the first character. This will then be
Let’s apply this to rhyming patterns of couplets.
a possible prefix in the node list. So from 01 we
There aren’t many: AA and AB. Following the
could follow edge 1 which produces the code 011
above recipe, we have one node for the prefix,
and so the target of that edge is 11.
which is just A. Then we have two edges. Accord-
If the graph has an Euler path (or cycle) we con- ing to the above, one should be labelled A and the
struct our de Bruijn sequence as follows: write other should be labelled B.

spring 2025 ⋅ 𝟧𝟥
 However, there is a slight difference here in that The answers are ‘yes’, ‘no’, and ‘good question’.
the edge that should be labelled B goes from the
node labelled A to itself, not to a node labelled B.
Before looking at that last one, let us first deal
This is because the label B says ‘This line has a
with the shortest length of a universal poem. By
different rhyme to its predecessor’, but once that
the construction of a de Bruijn sequence from
line is written then its predecessor is removed
its graph, the length of the sequence is the num-
from view and so its rhyming pattern is that of a
ber of rhyming patterns plus one less than their
single line. Then according to the convention on
length. So this sequence starts 𝟣+𝟢 = 𝟣, 𝟤+𝟣 = 𝟥,
rhyming patterns this is labelled just A. So when
𝟧 + 𝟤 = 𝟩, 𝟣𝟧 + 𝟥 = 𝟣𝟪. This is not in the OEIS,
setting up a suitable graph of rhyming patterns
the closest is A338735 which differs from this by
then there is a little bookkeeping to take care of
1 at each term.
when a rhyme ‘falls off the edge’. To help with re-
membering this, I’m using lowercase for labelling
edges. And so to the last question: how many universal
Here’s the graph for couplets: poems are there of a given length?

We’ve seen that there are two universal rhyming

b A a patterns for couplets: AAB and ABB. But we can
split the counting into two parts: it turns out
that all the Eulerian paths are actually cycles
Applying the technique yields two de Bruijn se- (which hopefully makes the ‘no’ answer to the
quences: we have to start at A, after which we second question above slightly intriguing). So
have a choice of a or b, but then we have to fol- from one cycle we can get a number of differ-
low the other one. This yields AAB and ABB as the ent rhyming patterns just by choosing a different
two rhyming patterns that contain every possi- starting point along it. As the number of starting
ble rhyming pattern for couplets. And this leads points is just the number of edges in the graph
to the opening doggerel (and its variants). regardless of the cycle, to count the number of
rhyming patterns it is simpler to start by count-
ing the number of cycles. Even here though the
Combinatorial poetry story is not quite that simple; it depends on what
The best type of answer is one that raises even one considers as ‘the same’ for two rhyming pat-
more questions than it answers. Indeed, having terns for universal poems. I’m counting two the
gotten this far then I found myself with a few same if all the appropriate subpatterns are the
questions: same. Thus our two rhyming patterns for cou-
plets are actually from the same underlying Eu-
lerian cycle—this is what I meant by saying that
Does a de Bruijn sequence exist for my two universal poems at the beginning were
every length of rhyming pattern?
really the same.
Does every de Bruijn
sequence form a cycle?
How many universal poems There’s also just one cycle for singlets (well, what
are there of a given length? would you call a one-line poem?) so our sequence
starts 𝟣, 𝟣.

(The last question is just my second question

from earlier and we are almost ready to answer To find out how it continues we need to draw
it!) some more graphs.

𝟧𝟦 ⋅ chalkdust
 feature

Here’s the graph for triplets: It’s not hard to see that the only choice here
when walking along the edges is as to the order
of the loops a and c at node AB so we get two
a Eulerian cycles.
b
a AA AB So the two cycles (starting at AA) look like this:
b AAabacb: the final b means ‘match the
c
second rhyme in the preceding couplet’
As with couplets, we have to be careful with the and that couplet is AC. So we can write
interpretation of the edges. this as AAABACC
AAabcab: the a looping at AB means
When we traverse an edge, we have to examine ‘match the first rhyme in the preceding
the last two lines of the rhyming pattern. We couplet’ and that couplet is BC. So we can
then have a set number of options depending on write this as AAABCBB.
whether the last two lines of rhyming pattern
rhyme of don’t rhyme: So our sequence of number of Eulerian cycles in
rhyming graphs starts 𝟣, 𝟣, 𝟤.
Here’s the graph for four-line poems:
START

Are
No Yes AAA
the rhymes b b
the same? a

ABB AAB
b

Pick an Pick an
c
option option c a a
b
ABC ABA

d c
Introduce Introduce
a new rhyme a new rhyme b
a
c
Repeat the Repeat
rhyme in the the rhyme
first line Counting the Eulerian cycles in this seems a bit
daunting, but it can be simplified. Firstly, the
loops. The loop labelled b at node ABA has to be
Repeat the traversed, and so it can be traversed at any stage
rhyme in the
second line when we arrive at node ABA. Since there are two
paths that lead to ABA, we must visit it twice and
so we can choose either of those visits as the time

spring 2025 ⋅ 𝟧𝟧
 to traverse the loop. So the presence of that loop So our sequence of number of Eulerian cycles
means that we get twice the number of Eulerian goes:
cycles that we would get if we simply ignored it. 𝟣, 𝟣, 𝟤, 𝟦𝟥𝟤.
Similarly the loops at ABC can be traversed one of
To get the number of rhyming patterns, multiply
six ways since there are two ways to order them,
each by the number of edges in its graph:
and then three ways to assign an ordered pair to
the two times we visit the node: either both on 𝟣 × 𝟣, 𝟣 × 𝟤, 𝟤 × 𝟧, 𝟦𝟥𝟤 × 𝟣𝟧 = 𝟣, 𝟤, 𝟣𝟢, 𝟨𝟦𝟪𝟢.
first visit, one on each (in their order), or both
on the second. So we get a factor of 6 here. The Neither of these sequences is in the OEIS!
loop at AAA doesn’t affect things since we only
visit AAA once in a cycle. Therefore, removing all The real challenge, though, is to write a poem
the loops simplifies the graph at the expense of that matches one of these patterns. Or possibly
a total factor of 12. all of them—a universal universal poem would be
one that had every possible universal poem (for
When looking for Eulerian cycles, it is easier to
a given length) contained within it!
pick a starting edge. For aesthetic reasons, I like
to start with the edge from ABB to AAA. The first Still, as Sarah Hart tells us, someone’s written a
decision is then at AAB where we have three out- book with 100,000,000,000,000 sonnets. So why
going edges. We’ll see that we can pick these in not challenge your English department?
any order, leading to a factor of 6. Each of these
exits from AAB must return to it via ABB (except
for the last, which ends the cycle there) and there
are two routes left from ABB to AAB, another fac- A universal poem for triplets
tor of 2.
I started this article with a poem, so it seems
Now of the exits from AAB, it turns out that the only fitting to end it similarly. Here is a possi-
crucial choice is as to whether the edge labelled ble universal poem for triplets using the pattern
a happens before or after that labelled c. Sup- AABACCC which starts at the AA node and then
pose it happens first, then there are three routes follows the b edge and then the a edge:
from ABA to ABB: direct (along a), via ABC (along
c and c), or via ABC and then back to ABA (along
b). Once one of those routes has been chosen, I thought I might
the route from AAB that starts along c is fixed. Write a poem tonight
So this gives us three options. A similar analysis But before I began
for if c happens first gives us also three options. To my utmost delight
So for each ordering of the exits from AAB there
I learnt of a way
are three possible ways to arrange the edges be-
tween AAB and ABB. To count the lines of a lay
And I spent so much time doing
Putting all of these factors together gives:
that and not writing a poem
𝟤 × 𝟨 × 𝟨 × 𝟤 × 𝟥 = 𝟦𝟥𝟤. that I’ve nothing to say.

Andrew Stacey
Andrew Stacey is head of mathematics at Oxford High School, part of the Girls Day School Trust.
Prior to that, he was an academic mathematician specialising in differential topology. He likes to
enter poetry competitions but has never yet won anything.

𝟧𝟨 ⋅ chalkdust
 flowc ha r t

Which sport are you?

Do you take yes
START a break every You are American football
2 minutes?

You are diving

no
∇⋅

What is your What is

fluids calculus ∇
favourite area your favourite
of maths? operator?

∇×
geometry

𝜀 How far You are curling

from water are
surprisingly far

you?
You are gradminton
quite close

Do you like
You are shapes? yes
swimming
You are kite foil
no

You are rowing

no Do you like I like tying knots

knots?

What is You are rock climbing

𝟥 × 𝟣𝟫 + 𝟥 × 𝟩 + 𝟤 × 𝟧?
my favourite checkout

19 penalties, 3 converted tries and 2 tries

I like untangling knots You are maypole

dancing
11 reds and 11 blacks

You are cricket

oval

What is
your favourite
You are snooker shape?

diamond
You are darts
You are baseball
Pictures American Football house: BanjoZebra,
CC BY-SA 4.0; Maypole dancing: Owain.davies, You are rugby union
CC BY-SA 4.0.

spring 2025 ⋅ 𝟧𝟩
 RE V IE WS
Books, films, games, anything mathematical...

Digital Fortress
Dan Brown
The incredibly beautiful and intelligent Susan Fletcher finds it really diffi-
cult at the NSA, where every single one of her colleagues is a man. When
the agency encounters a code that even its supercomputer can’t crack,
she’s not sure where to turn. Luckily, her incredible intelligence and
beauty come in handy as she tries to work out who the traitor is: her
boss (who’s in love with her), or her colleague (who’s also in love with
her)? Maybe if she namedrops her expensive shoes again, she’ll get some-
where.
This is a 510-page novel about cryptographic software, and prime num-
bers first get a mention on page 494.

iiiii
Prime Target Maths Fest
Doesn’t get anything right about An excellent one day event for
maths or history. sixth formers who love maths.

iiiii ggggg
Block Blast!
February’s planetary alignment A bafflingly popular combination of tetris, tris,
Always nice when a system collapses onto a dis and monis. Touted by its makers as both
one-dimensional space. addictive and relaxing.

ggggi gggii
𝟤𝟣𝟥𝟨𝟤𝟩𝟫𝟪𝟦𝟢 × (𝟤𝟣𝟥𝟨𝟤𝟩𝟫𝟪𝟦𝟣 − 𝟣) 𝟤𝟣𝟥𝟨𝟤𝟩𝟫𝟪𝟦𝟣 − 𝟣
A number that’s truly perfect. A number that’s truly in its prime.

ggggg ggggg
MathsJam Gathering Thunderbirds on Blu-ray
2024 Released to celebrate its 60th anniversary.
We had an excellent time. Best Pros: finally back in the original 4:3 aspect ra-
weekend of the year. tio. Cons: Thunderbirds Are 60 was right there.
ggggg ggggi
𝟧𝟪 ⋅ chalkdust
 bo o k of the yea r

BOOK of the YEAR 2024

Winner Readers’ choice

The winner of Chalkdust Book of the Year 2024 The winner of Chalkdust Readers’ Choice 2024
is Think Twice by Alex Bellos. is Mapmatics by Paulina Rowińska.
This books were selected by the editors of Chalk- This book was voted by our readers to be the
dust to be the Chalkdust Book of the Year 2024, Chalkdust Readers’ Choice 2024.
based on our four judging criteria: style, control,
damage and aggression.

Shortlisted

The winners were selected from our shortlist of seven books released in 2024. The five other books are
all also very good. They are: Everything is Predictable by Tom Chivers; Much Ado About Numbers by Rob
Eastaway; Maths 100 Ideas in 100 Words by Sam Hartburn, Ben Sparks & Katie Steckles; Love Triangle
by Matt Parker; The Art of Uncertainty by David Spiegelhalter.

spring 2025 ⋅ 𝟧𝟫
 A DAY
IN THE LIFE…

I
N this issue’s edition
of our day in the life f Megan Tarry works in audit in London,
series, we hear from
f Bayley McNevin works in audit in Liverpool,
three recent graduates
who work in finance: f Anna Lee works in tax in London.

𝟨𝟢 ⋅ chalkdust
 feature

Megan Tarry
Hey! I’m Megan and I’ve worked in audit in
London in a graduate role since August. I’m do-
ing an apprenticeship to become ACA qualified
which has involved studying for exams (so far
I’ve done 6 out of 15) alongside full time work.
Despite the controversial opinions, I’ve really en-
joyed my time in audit so far. Every day, I’ve
worked on something different and I’ve learned
so much in such a short space of time. Although
the maths that I use barely goes beyond GCSE
level, the application challenges you to think in
different ways which I have found really inter-
It’s fun to stay at the AAAA
esting! Welcome to a random day in my life!
My first challenge of the day is trying to find
the cheapest, least congested, most reliable way onto supplier statement testing that I made a
to get to work… it’s rare to get the luxury of all start on yesterday. This involves ensuring that
three! Today I decide to take the tube, which all purchase amounts issued by suppliers of
gets me to work around 8.40am. This means I goods to the client have been correctly recorded
have 20 minutes to settle myself in, make a cup in the client’s purchase records (AKA a lot of Ex-
of tea and catch up on emails and admin. cel work). I then spend a quick 20 minutes be-
Yesterday I sent out some requests for support- fore lunch sending an email to our data analyt-
ing documents to the client I am currently audit- ics team to resolve some formatting issues with
ing, so I spend the beginning of my day checking a document (it has over a million rows of data!).
our client portal to see if anything has been sub- We break for lunch for an hour and I spend it
mitted. heading out for a walk in the city before sitting
Then, at 9.30, I have my morning catch up down to chat to others in my intake. The nice
with the rest of my team where we discuss our thing about a grad scheme like this is that there
progress and plans for the day, before moving are a lot of us in the same position—a few years
out of uni with a lot to learn about the job.
After lunch, I speak to my assistant manager
who briefs me on my task for the afternoon.
I then work on some other areas of the audit,
checking that the accounts figures we have for
this year reconcile with some supporting docu-
mentation that we have been provided.
My day wraps up at 5.30pm and I rush back
home to get ready to head out again. Having to
balance both work and study has made it harder
to prioritise a work–life balance so I joined a lo-
cal competitive cheerleading team to ensure that
I step away from work for at least a couple of
hours a week in my busiest times and get some
Rate my meal deal (we give it a 6) endorphins!

spring 2025 ⋅ 𝟨𝟣
 Bayley McNevin
My day starts at 7am, with the usual scramble
to get up and ready. By 7.45am, I’m on the
Merseyrail, heading into the office in the Liver
Building (I know, I know—fancy). The early train
is a little bit busy, but it’s worth it to get a head
start in the morning. I mostly spend the journey
scrolling through emails, mentally preparing my-
self for another busy day.
I get to the office, get myself set at my desk,
and head to my 9am catch-up meeting. To-
day we’re discussing progress on our current
client. It’s the usual—everyone gives updates, Spot the former DITL contributor

talks through any problems, and someone in-

evitably brings up Excel shortcuts that no one ings and discussing next steps. It’s always a
has actually mastered yet. But that’s the audit bit tense, even when everything’s going well. I
life—a mix of progress, spreadsheets, and always try to keep my nerves in check—presenting your
a dash of confusion. work is never as easy as you think it’s going to
be, no matter how many times you’ve done it.
After the meeting, I do some of my work before I
But we get through it without any disasters.
grab a quick lunch. Today, it’s soup that I made
last night. I also grab a free fizzy drink from The highlight of the day? Getting my new lap-
the ‘can fridge’—a bit of an office legend at this top. My old one finally died last week, the screen
point. The new office was specifically designed just went black and never came back on. I swear,
with this fridge in mind, so it’s basically a rite of I heard it give a sigh of relief when it gave up.
passage to raid it when the midday hunger hits. But today, the shiny new replacement arrives,
It’s a small comfort in the middle of the hectic and I can’t say I’m not a little excited. Setting it
day. up takes far too long—but at least I’ll have some-
thing that works now.
Post-lunch, I’m in a meeting with the client.
We’re walking them through some of our find- By the time 5pm rolls around, I’m definitely
ready to switch gears. I meet up with an old
uni friend for a pizza at Rudy’s, which is just the
break I need. We catch up on life and try not to
talk about work too much. It’s the perfect way
to unwind after a full day of client calls, dead-
lines and technology drama.

After tea I head home, where the real work be-

gins: ACA exam revision. I’ve got a law exam
coming up, and I still have three exams left to
tackle after that. It’s not glamorous, but I know
it’s part of the process. One step closer to fin-
ishing the qualification—and one step closer to
knowing what it’s like to sleep again. And hope-
fully, one step closer to being able to drink from
POV: your train is arriving and it’s a class 777 the can fridge without feeling guilty about it.

𝟨𝟤 ⋅ chalkdust
 feature

based companies with operations abroad. Busi-

nesses often need guidance on navigating in-
ternational VAT regulations, and we help them
through our global network.

By 9.30, the office is buzzing. Once a week,

our team—12 people strong—gets together for
a catch-up on recent tax updates and ongoing
projects. Given our small size, we don’t need
daily meetings, but we always stay in the loop
on each other’s work.

As the day progresses, I tackle emails, client

queries, and VAT return reviews—especially as
Do you think Anna’s textbook was typeset in LATEX? deadlines approach. There’s a steady flow of
work, but the variety keeps it interesting.
Anna Lee At midday, I break up the routine with a
Hi, I’m Anna! Last June, I graduated from lunchtime run alongside a teammate, getting
Durham University with a BA in liberal arts, fo- some fresh air before diving back into work. If
cusing on history and economics. Fast forward I’m not running, I’m catching up with colleagues
to today—I’m living in London and working as from my training cohort. One of the perks of
an indirect tax analyst on a graduate scheme. As working in a big firm is meeting loads of people
part of my role, I’m also studying for two pro- my age who, like me, are navigating their first
fessional qualifications: the Association of Tax corporate job post-university.
Technicians (ATT) and the Chartered Tax Advi- Once the workday wraps up, I either hit the of-
sor (CTA). That means a few months of the year, fice gym or meet up with friends. Thursdays,
I swap spreadsheets for study leave, diving deep in particular, are big social nights—grads often
into tax law and compliance. gather for a post-work drink at a nearby pub be-
Since joining my graduate scheme, no two days fore heading home.
have been the same. The nature of my work Life as a graduate in London is fast-paced, but
keeps things fresh, and while I’ve tried to cap- I love the balance of work, learning and socialis-
ture a typical week, every day brings something ing. No two days are ever quite the same, and
new. that’s what makes it exciting! f
We have a hybrid working model, but we’re en-
couraged to come into the office at least three
times a week. Personally, I find I’m most pro-
ductive in the office, so I usually go in Monday
to Thursday. My day typically starts between
8.30 and 9.00am—cycling in when the weather
allows (which is a great way to wake up!). With
a coffee in hand, I check my inbox, follow up on
queries, and review any responses to emails I’ve
sent to our member firms.
Most of the client accounts I work on involve
overseas trade and VAT compliance for UK- The grindset never stops

spring 2025 ⋅ 𝟨𝟥
 THINGS WE
LEARNED AT
MathsJam Gathering 2024
Last year’s MathsJam Gathering took place in Staffordshire in November. Lots of us went,
and while we were there we learned a lot of new things including...

You need 11 cuts to turn one The answer to ‘Is πe or π + e

thing into 12 smaller things transcendental?’ is ‘yes’
With a little help from a guillotine, Numbers are called algebraic if they are
Christian Lawson-Perfect showed us that the solution of a polynomial equation
(if you’re not allowed to cut multiple with integer coefficients (for example,
pieces of paper at once) each cut of the √𝟤 is a solution of 𝑥 𝟤 − 𝟤 = 𝟢) and tran-
guillotine increases the number of pieces scendental if they are not (for example,
of paper by one, so making 𝑛 pieces will π and e are transcendental).
always take 𝑛 − 𝟣 cuts.
It is unknown whether πe is transcen-
dental and it is unknown whether π +
e is transcendental, but Ben Handley
showed us that at least one of πe and
There’s a lot of maths on
π + e must be transcendental.
buildings in Leiden
We were so excited when Hannah Gray
showed us this that we are turning it into
two pages of magazine for issue 22. You should fight one
horse-sized duck
Nicole Cozens finally settled the age-
The new prime number is big old argument of whether you should
fight 100 duck-sized horses or one
Ayliean MacDonald started her quest to horse-sized duck by working out which
get people to say the largest known prime option could hit you with the most
number (see page 13). There’s a lot of dig- force: the 100 mini-horses would hurt
its to say. you a lot more.

𝟨𝟦 ⋅ chalkdust
 th i n gs we lea r ned at. . .

It’s impossible to cut a pizza If you roll 7 dice, you will

into nine equal pieces with almost certainly roll a
two horizontal and two multiple of 7
vertical cuts James Grime showed us a game he made
Michael Gibson asked for a pizza cut up that he calls dice hard: you pick a num-
into nine slices and received a pizza cut ber from 2 to 7, then roll 7 dice. If you
like this: can rearrange the dice to make a 7-digit
number that is a multiple of the number
that you picked then you win, with more
points available if you picked a larger
number.
The chance that you can rearrange the
numbers to make a multiple of 7 is sur-
prisingly high: 99.997%. James, however,
After eating the pizza, Michael worked didn’t tell us how to work out which or-
out that it was impossible for all nine der of the dice gives you the multiple of 7,
of the slices to be the same size, so you so we’ll leave you to puzzle over that.
can’t share a pizza cut like this fairly
between nine people, but that it is pos-
sible for three people to share a pizza
cut like this fairly by having three slices Gluing the bottom of a Rubik’s
each if the lines are in just the right cube is really inefficient
places.
Katie Steckles showed us that you can
solve a Rubik’s cube without turning one
of the faces by gluing one of the faces in
place and doing a sequence of 17 moves
Infinite mancala isn’t a very
that was equivalent to rotating the glued
good game face. Impressive, but a lot slower than just
Jonathan Welton told us about man- turning that face.
cala on an infinitely long board start-
ing with 𝑛 pebbles in every 𝑛th hole. He
found that for lots of values of 𝑛 the re-
sults were periodic, but often it was a Negative infinity is very big
while before the repeating started.
One of the tasks that we set people tak-
For 𝑛 = 𝟤𝟣, Jonathan isn’t sure if the ing part in Taskmathster was to make
result are periodic—but if it is periodic, the largest small number. The team that
the repeating doesn’t start for at least made a very large negative infinity out of
21 trillion moves. chairs did not score highly.

spring 2025 ⋅ 𝟨𝟧
 The crossnumber
#21, set by Humbug sponsored by

1 2 3 4 5 6 7 8

9 10

11 12 13 14 15

16 17 18 19

20 21 22 23 24 25 26

27 28 29 30

31 32 33 34

35 36 37 38 39

40 41 42 43 44

45 46 47 48 49 50

51 52 53 54

55 56 57 58

59 60

61 62 63

An over-zealous copy editor has ‘corrected’ this crossnumber. In the original puzzle, every clue was of the
form 𝑎 † 𝑏 A or 𝑎 † 𝑏 D, where 𝑎 and 𝑏 are positive integers, and † is +, −, ÷, or × (and A and D stand for
across or down); the copy editor then took every clue, calculated 𝑎 † 𝑏 , and replaced 𝑎 † 𝑏 with the result.
For example, if a clue now reads ‘6A’, then the actual clue could be:
𝟢 + 𝟨A, 𝟥 + 𝟥A, 𝟣𝟤 − 𝟨A, 𝟣 × 𝟨A, 𝟤 × 𝟥A, 𝟨 ÷ 𝟣A, 𝟣𝟤𝟢 ÷ 𝟤𝟢A, or many other things...
There is only one solution to the completed crossnumber. Solvers may wish to use the OEIS, Python, a team
of 300 volunteers with abacuses, etc to (for example) obtain a list of prime numbers, but no programming
should be necessary to solve the puzzle. As usual, no entries begin with 0.
To enter, send us the sum of all the digits in the row marked by arrows by 25 September 2025 via
the form on our website ( d chalkdustmagazine.com). Only one entry per person will be accepted. Winners
will be notified by email and announced on our blog by 1 November 2025. One randomly-selected correct
answer will win a £100 Maths Gear goody bag, including non-transitive dice, a Festival of the Spoken Nerd
DVD, a dodecaplex puzzle and much, much more. Three randomly-selected runners up will win a Chalkdust
T-shirt. Maths Gear is a website that sells nerdy things worldwide. Find out more at d mathsgear.co.uk

𝟨𝟨 ⋅ chalkdust
 cro s s num ber

Across Down
1 997A. (3) 1 1D. (5)
3 81A. (4) 2 6172344439D. (5)
6 1D. (5) 3 3D. (4)
9 286A. (4) 4 4D. (6)
10 16D. (3) 5 81707A. (5)
11 2A. (3) 6 3A. (4)
12 6A. (4) 7 72D. (3)
13 1D. (5) 8 60A. (3)
16 16A. (4) 14 15094A. (4)
18 360D. (4) 15 4D. (4)
20 2A. (4) 17 5901148A. (4)
22 8A. (4) 19 19893A. (4)
24 15104D. (4) 21 725D. (3)
27 4A. (4) 23 23D. (4)
29 40A. (6) 25 3038A. (4)
31 30A. (5) 26 3661706D. (4)
33 6A. (4) 28 1925D. (4)
35 4A. (4) 30 1009262A. (4)
37 668997D. (5) 31 1061A. (4)
40 34459245389A. (6) 32 41A. (4)
43 7A. (4) 34 7568A. (4)
45 4A. (4) 36 11727A. (4)
46 8A. (4) 38 6468D. (4)
48 6A. (4) 39 705159A. (3)
51 4A. (4) 41 2302929D. (4)
53 5A. (4) 42 766A. (4)
55 14750A. (5) 44 584828A. (6)
57 4A. (4) 47 47D. (5)
58 56D. (3) 49 59921D. (5)
59 256D. (3) 50 21D. (5)
60 2A. (4) 52 52D. (4)
61 4A. (5) 54 54D. (4)
62 5A. (4) 55 98D. (3)
63 64D. (3) 56 944D. (3)

spring 2025 ⋅ 𝟨𝟩
 Skimming potatoes
Ryan Palmer gives his chips tips for
finding the perfect skimming stone

A
S a mathematician, it is rare that I get ming, ‘ducks and drakes’, skipping (if you are
to speak to the media. My papers don’t American—the internet made sure we knew this),
always ‘catch the eye’. Yet one did. On or ‘piffing a yonnie’ (no idea!? That’s on you, Aus-
4 January 2023, a random Wednesday in an arbi- tralia).
trary year, my inbox blew up and my phone kept
Perhaps it was the time of year. A slow news day.
ringing. That day, a piece of my research with
A piece of new year’s ‘fluff’ at a time when no-
Frank Smith, a professor in the department of
body wants heavy news—‘a scientist redefining
mathematics at UCL, was published. As a result
rock skimming?’—that’s the stuff people want!
of this article, 83 news stories appeared around
the world. I even had the chance to speak to vari- I had expected it to all blow over quickly, but
ous radio presenters and a couple of TV reporters then I made a fatal error. At some point, I men-
to boot. And in all honesty, I was surprised! tioned the joy of trying to skim potato-shaped
stones. And, well, that was it. They got ob-
What did we do to receive such global interest, I sessed. In one small sentence, I soon became
hear you ask? Well, we studied skimming—how globally (but not widely) known as the scientist
objects hit water and bounce back off. Much who suggested potato stones are better for
like the humble pastime we all know and love skimming. Their spud-based, mish-mash of inter-
of stone skimming, also known as rock skim- est and enthusiasm boiled over. After that one

𝟨𝟪 ⋅ chalkdust
 feature

story, more reporters came, eager to hear about and mass have here?
the mysterious potatoes that skim so well.
If an ice crystal sinks into a water layer, freez-
I learned two lessons that day: ing may ensue and, without protection, an ice
block may form. If the crystal skims, we want
1. Be careful what you say when the microphone
to know where it goes and whether that changes
is on you (lest you become forever known as
how we protect an aircraft. So, we studied the
the ‘expert’ of ‘potato skimming’—ouch), and
relationship between ice crystal mass and shape
2. Don’t take the internet too seriously (people to show how various types of ice crystal behave—
are passionate about skimming... skipping— it’s either sink or skim.
sorry America).
Mathematical modelling allows us to peer be-
While it was nice to receive some light-hearted yond the constraints of experimentation. Yes,
positive recognition, we didn’t really set out to experimental evidence is vital, but it is limited.
study potatoes... in reality, we were interested If you want specific answers to specific ques-
in something much more important. tions, experiments are great. But if you are af-
ter something a little more general and want to
unpick the underlying physics, this is where the
Aircraft icing beauty of maths reveals itself. With well-defined
physics and a set of equations to hand, mathe-
All this potato waffle and online roasting be-
matical modelling enables us to evaluate and un-
gan with the study of aircraft, specifically air-
derstand hard-to-measure problems and, in our
craft icing. In the upper reaches of the atmo-
case, inform the efficient and effective design of
sphere where the air gets cold and clouds get
aircraft ice protection systems.
thin, aircraft can meet ice crystals and super-
cooled droplets (unfrozen water drops that are
colder than freezing). Relatively harmless on
their own, when lots of these particulates hit an
Making it mathematical
aircraft they can cause ice to form on the wings, To model what happens when the ice meets wa-
engines and other vital components. Without ad- ter, we need some equations.
equate protection, this icing can become a sig-
Modelling fluids is tricky. Most of fluid mod-
nificant hazard. Historically, icing events have
elling starts with a classic set of equations
led to disaster, with two well-known instances
(known as the Navier–Stokes equations) but
together claiming the lives of almost 300. To pre-
these are a complex beast to solve. To make life
vent future incidents, the aerospace community
easier, we can employ a classic applied maths
continues to work to understand the physics of
technique called asymptotic reduction—which is
this hazard and protect against it.
really a fancy way of saying, we compared the
This problem was the essence of the motivation big and small factors in our problem and kept
for our study. When ice and water particles im- what was most influential. To do this, we make
pact warm aircraft surfaces, such as wings and some assumptions about different parts of the
engines, water layers form. Once formed, other ice–water interaction, like the water layer being
ice crystals may follow up and hit these layers. shallow (eg thin and long), the horizontal veloc-
So, what happens next? Does the ice crystal skim ity of the ice being much greater than its vertical
much like a rock on a lake or sink like a potato velocity (as you would expect when skimming!)
in a pan? In the case of aircraft icing, the skim- and that the fluid has certain physical properties
ming ice isn’t some nice flat rock but potentially —then our problem simplifies! With some mas-
all sorts of shapes and sizes. So, the natural ques- saging, our set of equations is no longer quite so
tions to ask is what influence does the ice shape beastly. We now have a new set called the shal-

spring 2025 ⋅ 𝟨𝟫
 low water equations (included below for awe-
inspiring effect):
ston
e
𝜕𝑢 𝜕𝑢 𝜕𝑝
+𝑢 =− ,
𝜕𝑡 𝜕𝑥 𝜕𝑥
𝜕ℎ 𝜕 liquid layer
+ (ℎ𝑢) = 𝟢.
𝜕𝑡 𝜕𝑥
𝑥𝟣 𝑥𝟢 = 𝟣
If you are unfamiliar with this notation, don’t solid wall
worry: broadly these two equations together de-
scribe how the water layer’s horizontal velocity, A stone skimming the surface of a liquid layer. If
𝑢 , and height, ℎ, relate to each other and vary 𝑥𝟣 reaches the body’s leading edge, we deem it
to have sunk, while if it reaches the trailing edge,
in time and space. With that, we can model the this is a skim.
water layer.
As the contact region grows larger, the pressure
Now, what about the ice (which we will refer to between fluid and body increases also, pushing
as the ‘body’ from now on)? When the body back against the body and potentially lifts it back
first descends into the water’s surface, pressure out of the water. If the force is sufficient, the
𝑝(𝑥, 𝑡) builds up underneath as the water resists body skims! Yet, depending on how the body ini-
the body’s entry. This pressure along the body tially makes contact with the water, it may well
produces a force, hence we can turn to Newton’s sink! If the body is too heavy or descends too
second law, the classic 𝐹 = 𝑚𝑎 (force equals mass quickly, that the water will stretch past the end
times acceleration) to understand how the body of the body and it will flood, claiming the body
reacts, lifting (changing its vertical position 𝑦 ) to the water’s murky depths.
and rotating (changing its angle 𝜃 ) as follows:

What does this mean for the

𝑥𝟢
d𝟤 𝑦
∫ 𝑝(𝑥, 𝑡) d𝑥 = 𝑚 ,
𝑥𝟣
𝑥𝟢
d𝑡 𝟤
avid amateur skimmer?
d𝟤 𝜃
∫ 𝑥𝑝(𝑥, 𝑡) d𝑥 = 𝐼 d𝑡 𝟤 . Through our study we uncovered a relationship
𝑥𝟣
between a body’s shape and mass that deter-
Note that 𝑚 and 𝐼 (the body’s moment of iner- mines its ability to skim. Greater mass can lead
tia) here are multiplied by vertical acceleration to a ‘super-elastic’ response (a most enticing
(d𝟤 𝑦/d𝑡 𝟤 ) and the angular acceleration (d𝟤 𝜃/d𝑡 𝟤 ), name)—an almighty leap from the water, with
just as in Newton’s second law. the body leaving at a greater height than it en-
tered. Have we broken physics? Did we find
a way to overcome the pesky constraints of the
Now let’s think through what happens while the
conservation of energy? Fortunately, not!
body skims. Before entering the water, the body
is ‘dry’. Then, as soon as the body first touches Bodies with larger masses sink deeper into the
the water, only a single point is wet. Let’s call water and for longer; this increases the pressure
it 𝑥𝟢 . As the body plunges further still into the under the body and its contact time with the wa-
water, more water contacts the body—as shown ter. In doing so, the large horizontal velocity of
above right. Underneath the body then, there is the body (much like your thrown stone) is con-
a contact region (the ‘wet bit’) where the body verted into vertical velocity. Hence the almighty
meets water, which we denote as being between leap. But, interestingly, if the same body be-
two points 𝑥𝟢 and 𝑥𝟣 , the trailing and leading comes more curved—this effect is inhibited. The
edges of this wet region, respectively. body rotates more and leaves the liquid layer

𝟩𝟢 ⋅ chalkdust
 feature

It’s a
chip-wreck

The body might sink...

...or skim.

Byeeeeeeee!

faster. A shorter, sharper skim. Altogether then, orange), the body undergoes a successful skim,
curvature enables heavier bodies to skim that descending into the liquid layer and rising out
would otherwise sink if they were flatter. For again, leaving the liquid layer at a lower height
our aircraft icing problem (and your holiday by than it entered (this is due, in part, to how the
the beach), this means a broader range of bod- body rotates!). The most interesting case here is
ies may skim than expected with some possibly 𝑐 = 𝟣 (green) in which the so-called super-elastic
rather dramatic results: response is seen: the body is ejected from the liq-
uid layer at a far greater height than it entered!
height
So, while rock skimming is fun—aircraft icing is a
6 𝑐=𝟣
more significant application our findings. Thank-
4 fully, our paper is only part of the varied re-
search on skimmers and not the single authority
2
(though one Redditor did question whether we
0 time had any authority given ‘they’d never seen a sci-
2 4 6 8 10 entist throw anything further than 5 feet...’—it’s
-2 𝑐 = 𝟣𝟢
𝑐=𝟢 the ones closest to the truth that hurt the most).
-4 Still, skimming is as complex as it is fun. So,
when you next stand by that body of water, glis-
The graph shows the vertical position during a tening in the sun, ready to skim—you can leave
skimming motion of three bodies with different your King Edwards, Maris Pipers and Russets be-
amounts of curvature (𝑐 ). All three bodies en- hind. Pick up a weighty stone and see if you
ter the water layer at the same height. The flat can get that almighty super-elastic leap. And
body (𝑐 = 𝟢) sinks and so the blue curve stops when you do, remember, that the phenomenon
mid-skim. For the largest curvature case (𝑐 = 𝟣𝟢, of skimming goes far beyond the lake.

Ryan Palmer
Ryan is a generally curious applied mathematician who, by his own admission, finds it hard to say
‘no’ to new research topics. . . His research spans sensory biology, electrical ecology, fluid dynamics
and healthcare modelling. He can be found at the University of Bristol’s school of engineering
mathematics and technology where he is a lecturer in mathematical and data modelling.

spring 2025 ⋅ 𝟩𝟣
 THINGS TO GET IN
THE POST !

At number 10, it’s

a gas bill.
At 9, it’s another bank
statement for the previous
tenants.
𝟏𝟎
𝟏𝟎
𝟗
At 8, it’s this year’s bin
collection timetable.
At 7, it’s the new Screwfix
catalogue. 𝟖𝟖
𝟕
At 6, it’s the copy of
Clown Town that you
A new entry at 5, it’s preordered.
your ballot papers
from UCU.
𝟔𝟔
𝟓𝟓 Down 3 places to 4 as
it’s really not the right
time of year:
a Chalkdust
At 3, it’s the stuff you
ordered from
Christmas card. 𝟒𝟒
Maths Gear.

𝟑
At 2, it’s a copy of
Think Twice by
Alex Bellos (see
A new entry at page 59).
number one this
issue: a copy of 𝟐
Chalkdust NEXT ISSUE
issue 21. We rank the top ten

𝟏 mathematical pets.
Vote now at chalkdustmagazine.com
Pictures UCU strike: LornaMCampbell, CC BY 4.0.

𝟩𝟤 ⋅ chalkdust
ON MY
BLACKBOARD… What’s taking up
the most space?
We get to know the What’s the overall vibe? Advection–diffusion
editors of Chalkdust Squishy gels! I share the equations: they de-
by asking them blackboard with my group and scribe the transport of
about the most we all work on modelling how fluid through the gel
personal subject hydrogels swell and shrink. as it swells and shrinks.
possible: the Hydrogels are gels which can
contents of their absorb large volumes of wa- Any drawings you’re
blackboards. ter and are what soft contact especially proud of?
lenses and Orbeez are made We draw gels a lot,
This issue, we talk to
of, among other things. which tends to involve
Ellen Jolley, who’s
been on the team drawing a square or
since Issue 12. Is there anything on there circle and writing ‘gel’
you’d never get rid of? in the middle. Today
Not on the blackboard, we we also have a bend-
tend to erase it quite often. ing gel bilayer—if two
But we do have a picture layers of gel are stuck
of our supervisor permanently together and one
on our pinboard for inspiration. swells and the other
And an origami frog. doesn’t it can cause
bending (as depicted!)

Finally… chalkboard or
whiteboard?
Whiteboard for practicality
but chalkboard for vibes.