ILLUSTRATING

MATHEMATICS
Edited by Diana Davis
ILLUSTRATING
MATHEMATICS
Edited by Diana Davis
 Copyright Information CONTENTS

etc Introduction 3
Drawing 4
Designed by Margo Angelopoulos
Paper & Fiber Arts 14
Laser Cutting 32
Graphics 48
Video and & Virtual Reality 66
3D Printing 82
Mechanical Constructions & Other Materials 128
Multiple Ways to Illustrate the Same Thing 148
INTRODUCTION It is said that an interpreter ‒ a person who translates from one language to
another ‒ does not exchange words for words, but rather exchanges meaning for
meaning. As mathematicians, a large part of our job is to explain things to others.
We may use words and symbols to do this, but our job is not primarily to convey
words and symbols; like an interpreter, it is to convey meaning.

There are many ways to communicate meaning. As the great mathematician

William Thurston said, “Mathematics is an art of human understanding. Math-
ematical concepts are abstract, so it ends up that there are many different ways
that they can sit in our brains. A given mathematical concept might be primarily
a symbolic equation, a picture, a rhythmic pattern, a short movie ‒ or best of all,
an integrated combination of several different representations.”

The purpose of this book, then, is to help you find a good representation of the
mathematical concept you wish to illustrate. This book does three things:

• Showcases the great variety of materials for illustrating mathematics,

• Gives voice to people’s stories about illustrating their mathematics,

so that we can learn from their experience, and

• Shows the variety of ways that different people use the same materials
in very different ways.

In addition, it will introduce you to many of the amazing people who spent
time at ICERM in Fall 2019 for a program on Illustrating Mathematics, in an
attempt to capture some of the creative and generous spirit that flowed through
our days there.

Diana Davis
May 1, 2020
Bures-sur-Yvette, France

3
DRAWINGS
The adage that a picture is worth a thousand words is certainly
true in mathematics, where one carefully chosen figure can
eliminate the need for many lines of equations and exposi-
tion. For most of us, our most frequent way of illustrating
mathematics is by hand – on chalkboards, scrap paper, paper
napkins or, increasingly, whiteboards and electronic tablets.

4 5
A B These drawings are from my first solo paper “Bundles, handcuffs, and local
freedom.” They illustrate the existence of a knot in the 3-sphere whose
complement is hyperbolic, that admits a fibration over the circle, and whose
group contains a subgroup that is locally free (finitely generated subgroups
are free) and not free. This answered a question of James Anderson: He had
observed that if such a thing could not exist, then there would be counterex-
amples to Thurston’s Virtual Fibration Conjecture (which we now know to
be true by remarkable work of Ian Agol and Dani Wise).

The fundamental group of the 2-complex X (A) contains a subgroup that

is locally free and not free. The method to construct the knots is to find
a fibered knot in the complement of X so that X’s group injects into the
C D knot’s. The complement of X is the handlebody (B). The knot is inside
the handlebody (C). The knot is built by taking a square knot (which is
fibered), cabling (which is again fibered), and then “plumbing.” Plumbing is AUTUMN KENT
the process of gluing a twisted band along a square in a fiber, as illustrated University of Wisconsin
in (D–E). The rest of the figures are part of the proof that the knot is hyper-
pen drawings
bolic, and that you can perform surgeries to get closed hyperbolic fibered
manifolds with groups having locally free non-free subgroups.

One year while I was in graduate school, Bob Gompf was teaching a course
in 4-manifolds that I had lost the thread of. I didn’t want to be rude, so I
E kept attending. So there I was in class, doodling and thinking about Ander-
son’s question. I knew that connect sums, cabling, and plumbing preserved
fibering. I sketched the complement of X and saw the square knot there. I
cabled it for fun. I wanted the result to be hyperbolic, so I needed to do
something else. So I plumbed a little band on to get rid of the essential
torus that was ruining hyperbolicity, and I left class with a theorem. It took
a while to prove that the examples were hyperbolic, but I had found them.
Getting lost in a lecture isn’t always bad!

I drew the pictures by hand since I didn’t know how to do it any other way!

The original paper: Autumn Kent, Bundles, handcuffs, and local freedom,
Geometriae Dedicata 106 (2004), 145–159.
7
In December of 1971, the grad students at Berkeley invited me (I was also
heavily bearded with long hair) to paint math frescoes on the corridor wall
separating their offices from the elevator foyer. While milling around before
painting, one of the grad students, Bill Thurston, came up to ask, “Do you
think this is interesting to paint?” It was a complicated maze-like looking
smooth one-dimensional object encircling three points in the plane. I asked,
“What is it?” and was astonished to hear, “It is a simple closed curve.” I said,
“You bet it’s interesting!”

So we proceeded to spend several hours painting this curve on the wall. It

was a great learning and bonding experience. For such a curve drawing to
look good, it has to be drawn in sections of short, parallel, slightly curved
strands that are subsequently smoothly spliced together.

When I asked how Thurston got such curves, he said by successively applying
to a given simple curve a pair of Dehn twists along intersecting curves. The
“wall curve painting,” two meters high and four meters wide, dated and
signed, lasted on that Berkeley wall with periodic restoration for almost four
decades before finally being painted over.

WILLIAM THURSTON &

DENNIS SULLIVAN
University of California, Berkeley
painted mural

Further information: Lee Mosher, What Is... A Train Track? Notices of the AMS
50(3), March 2019, pp. 354–356.
9
In his work on 3-manifolds, William Thurston proved that there exist exactly
eight “geometries” that serve as models for any compact 3D object. The
Euclidean geometry that children learn at school is one of them. Others, like
spherical geometry or hyperbolic geometry, are well known to mathemati-
cians. Still others are more exotic, e.g. the Nil and the Sol geometries.

The underlying topological space of the Nil and the Sol geometries is
standard 3D Euclidean space. However, the way of measuring distance
between two points is not the usual one from the Euclidean geometry. A
natural question is:

What is the shape of a ball in each of these geometries?

This is a picture of a sphere in Sol geometry. In other words, every point on

the surface of this shape is the same distance from the central point, using RÉMI COULON
the Sol distance metric. CNRS / Université de Rennes 1
charcoal on paper
This project originated from a course given by Rich Schwartz during the
“Illustrating Mathematics” semester program at ICERM in Fall 2019. Rich
showed us beautiful animations of geodesics and balls in the Nil and Sol
geometries. I became so intrigued by these objects that I decided to make
physical models of them. As the shapes of the balls can be generated by a
computer, 3D printing was the easiest and fastest way to achieve this goal.
The shape of the balls are so elegant, that I used them later to practice
charcoal drawing!

Later, these 3D-printed models turned out to be very useful for another
project that we worked on at ICERM. The goal was to produce a virtual
reality program of the eight Thurston geometries. Having the 3D-printed
balls in our hands helped us to check that the pictures generated by the
computer made sense, and track the bugs in the code.

An article by Matei P. Coiculescu and Richard Evan Schwartz about the balls in Sol:
http://www.math.brown.edu/~res/Papers/sol.pdf
11
This animated gif shows a construction of truncated hyperbolic space, or
neutered space. I made it using Notability on an iPad, hand drawing all 50
pictures individually. I like drawing, and I can do it easily on my iPad while
traveling and on the go.

For a non-compact hyperbolic 3-manifold with fi nite volume, the universal

cover is hyperbolic 3-space, and the fundamental group of that manifold acts
on it but not geometrically, as the quotient is that non-compact manifold.
When we cut off the cusp, since the cusp retracts on a slice of itself, it does
not change the fundamental group. Cutting the cusp, we see that the slice is
a flat torus, whose universal cover is the flat plane.

This flat plane, in hyperbolic space, sits as a sphere centered at infi nity, on
the horizon: a horosphere. Cutting the cusp amounts to, at the universal
cover level, removing an infi nite collection of disjoint horoballs, which are INDIRA CHATTERJI
the interiors of those horospheres. Now the fundamental group acts on the University of Nice
truncated hyperbolic space geometrically.
hand-drawn animated gif image
The terminology “neutered space” was coined by Benson Farb in 1994, and it
is close to perfect: short, illustrative and easy to remember, and it encodes in
two words the whole construction and most of the assumptions. The violence
associated to that word, however, made it difficult to use and is at odds
with the actual precision of the procedure: the horoballs’ removal is precisely
encoded by the group action. In 1999, Martin Bridson and André Haefl iger
used the terminology “truncated hyperbolic space,” which is in fact quite
accurate in the upper half plane model.

View animation: https://math.unice.fr/~indira/GIFS/Horo.gif

13
PAPER &
FIBER ARTS
Paper is by far the most abundant resource in most mathemati-
cians’ offices, and is our most readily available tool for creating
mathematical illustrations. Most of us created the Platonic
solids out of paper at some point in our early education, and
have created numerous other models throughout our math-
ematical careers. For creating objects that people will handle
and manipulate, and especially for those that require some
stretch, sewn or knitted fabrics work better than paper.

14 15
The Fabric of Spacetime is an installation serving as an interactive model
of a young universe, combining crochet and electronics to create a dynamic
and luminescent experience. The main physical component is a large, hand-
crocheted hyperbolic manifold, where stitches were added at an exponential
rate. In this way, the circumference grows exponentially faster than the
length, introducing negative curvature and resulting in the many folds. This
technique was pioneered in 1997 by Daina Taimina, a mathematician at
Cornell. The curvature is analogous to the geometry of a very young universe
(much less than one second old), where the spacial dimensions grew expo-
nentially with respect to the time dimension, introducing curvature in the
geometry of spacetime itself.

Sewed into the fabric model are 264 individually programmable neopixel
LEDs, forming a spiral pattern around the inside, and mounted throughout
the room are 6 servo motors, each connected by fishing wire to a fold in the
crocheted model. There is also a PIR motion sensor directed toward the
underbelly of the piece. All of these are wired to an Arduino MEGA micro-
controller.

While undisturbed, the servos pull the model open and closed in a regular
breathing motion, shifting to red while opening, and to blue while closing.
This is in homage to the cosmic red shift and blue shift of the universe: since
red has the longest wavelength of the visible spectrum, things moving away
from us at relativistic velocities gain a slight red tint in color, and since blue
has the shortest wavelenth, things moving towards us gain a slight blue tint.
In this way, the color shift of the piece represents the actual shift that would
GABRIEL DORFSMAN-
be visible in an expanding and contracting universe. When the viewer walks
under the piece, they trigger chaotic motion and lights. HOPKINS & MEGHAN
MAYNARD
I find crochet to be a very flexible medium, both literally and figuratively, and ICERM; Independent artist
I am fascinated at the way it can create various geometries and symmetry
patterns, and also how easy it is to combine with electronics. Using crochet, crocheted yarn and LEDs
I really tangibly felt the intensity of exponential growth. To create hyper-
bolic crochet, you add stitches in such a way that the circumference of the
crocheted piece grows exponentially. In practice, this means the workflow
really starts to slow down. This piece took hundreds of hours of crochet, and
weighs 10 pounds. The circumference of the piece at the base is over 80 feet
(all hidden in the folds).

17
Essentially, my project illustrates a version of the Gauss-Bonnet Theorem.
The question I explored (with Bill Thurston), was the following:

How can we make well-fitting “clothing,” out of rigid material, for a

sphere and other surfaces?

Our first try at making clothing for a sphere was the version using “fat
triangles” to make an octahedral pattern (top left). This is the pattern made
from 8 “triangles” that have edges made from circle arcs, that make right-
angled triangles. This sort of pattern works great when made from cloth,
which has some stretch, but when made from paper, created a “sphere” with
large flat faces. By making longer, meandering seams, the curvature was
distributed over a larger portion of the sphere, which made a very satisfying
sphere (bottom). We made an intermediate sphere to further illustrate how
seams can distribute the curvature (top right). KELLY DELP
Cornell University
Paper was a fun, inexpensive medium to work with. Also, craft paper cutters
electronically cut paper
are relatively inexpensive, and can easily cut complicated curves. Through
this project, I learned about planar curvature of curves, and Gaussian
curvature of surfaces.

Further information: Kelly Delp and Bill Thurston, Playing with surfaces:
spheres, monkey pants, and zippergons. Proceedings of Bridges 2011:
Mathematics, Music, Art, Architecture, Culture (2011), pp. 1–8.
19
To make a torus from a rectangle of paper, glue a pair of opposite sides
together to form a cylindrical tube, then glue the two ends of that tube
together. The second gluing requires squashing the cylinder, resulting in a
doubled-up cylindrical tube. It is somewhat unsatisfying, as it encloses no
volume, making it far from the “typical torus” formed by a tire or a lifebuoy,
which could be described as a “circular tube around a circle.”

The mathematician’s dream is a non-squashed polyhedral torus, enclosing

volume like the “typical torus,” yet everywhere flat in the sense that each
vertex has total angle 360°. A common belief is that these wishes cannot be
satisfied simultaneously. Yet Dmitry Burago and Victor Zalgaller showed
that it can be done, as we learned from Henry Segerman and his 3D-printed
“hinged flat torus”! Having learned that, we started exploring paper models
for such tori.
ALBA MARINA MÁLAGA
One description of our model starts from a prism on a regular n-gon, with SABOGAL & SAMUEL
its bottom and top faces removed. For some integer k less than half of n, LELIÈVRE
consider triangles joining top vertices to bottom vertices: each top vertex is
ICERM; Université Paris-Saclay
joined to the next top vertex, and to the bottom vertex k steps ahead; each
bottom vertex is joined to the previous bottom vertex, and to the top vertex folded paper
k steps back. These triangles are all congruent. Together with the outside
of the prism, they form a polyhedral torus where at each vertex the three
angles of congruent triangles meet on the inside, and two right angles meet
on the outside, making a total of 360°.

For a paper layout, cut along the outer prism’s half-height line, along one
inner triangle edge and along the two outer vertical half edges from the ends
of that inner edge to the first cut. The pictures show examples of such tori
for various choices of n and k, and paper layouts at various stages of folding
up. Extra flaps allow us to assemble the tori without glue or tape.

Instead of cutting by hand, we wrote a script using SageMath, a piece of free

software for mathematics. Given a choice of n and k, it outputs an SVG file,
from which a mechanical or laser cutter can cut along some lines (to cut out
the shape) and score along others (to make it easier to fold). Given the cut
and scored piece of paper, it still takes some practice to assemble the torus.

Further information: https://im.icerm.brown.edu/portfolio/paper-flat-tori/

21
Due to some fortunate events, I found myself with a laser cutter. I was
happily using it to make tiles and other flat things, but what I really wanted
was to take the leap to making 3D things. Bent and folded paper was an
obvious option, and I worked on many different things, some more successful
than others – different objects, and different ways of connecting them. Then
I woke up one morning and somehow several things had come together in my
head: I had an idea for a connector, and something to try to make. I have to
confess that this happens quite a lot and most of the time the ideas turn out
to be wrong or not quite work, but this time they were perfect. Curvahedra
was born!

This process of initial play and discovery meant that I had something; the
next question was to ask what I could illustrate with it. The way that the
system connects up to make loops controls the holonomy (“turning”), and
thanks to the Gauss-Bonnet theorem, the holonomy controls the curvature EDMUND HARRISS
of the surface. University of Arkansas
laser-cut mylar
This illustration is, of course a direct result of my initial motivation, to use
2D cutting to produce true 3D objects. I hope this illustrates that sometimes
it is worth playing and developing, rather than pushing directly for a desired
outcome.

For more information and to purchase curvahedra: http://curvahedra.com

23
I make crocheted models of basic hyperbolic surfaces such as a disc or a
cylinder. The models carry triangulations of surfaces, and demonstrate
several properties: first, that the sum of angles in a hyperbolic triangle is less
than 180 degrees, and second, that the circumference growth is faster than
that of a Euclidean circle.

I prepared the first crochet model to demonstrate some ideas in a Rieman-

nian Geometry class I was teaching. It was a solid hyperbolic disc, and I
wanted to demonstrate the growth of a ratio of the circumference to the
radius of a disc. I found that the process of crocheting a solid hyperbolic disk
took too long, because it is exponential in time and in yarn consumption.
So since I had lots of experience with crochet, I started experimenting with
different stitches and techniques. I suddenly realized that using a triangular
pattern gave a triangulation of the surface that showcases many other visual
properties. MARIA TRNKOVA
University of California, Davis
I made these objects to illustrate the Nash-Kuiper C 1 embedding theorem,
crocheted yarn
which roughly implies that if you have a 2D surface that is oriented and closed,
then there is a way to embed it into an arbitrarily small ball of Euclidean
3-space, in such a way that the surface is not stretched or compressed at all.

The crocheted objects serve as models of a hyperbolic disk that is embedded

into 3-space in a way that preserves all of its intrinsic properties. It is
clear that a crochet model cannot be placed inside of an arbitrarily small
Euclidean ball, because the yarn has some thickness and volume. Still, each
model demonstrates smooth “curving” inside of some ball, and “tries” to fill
the whole space inside of that ball. As a crochet model is extended by adding
more rows, it will look more and more like a ball, and will be increasingly
thick and dense.

For these models, the stretching properties of yarn are important. I tried
several types of yarn, and acrylic and wool work best. Making the pattern
was the most time-consuming part. I experimented a lot with which pattern
returns a look that is most similar to that of equilateral triangles.

Further information and crocheting instructions:

Maria Trnkova, Hyperbolic flowers, preprint (2019).
25
The triacontahedron is a Catalan solid that has thirty congruent rhombic
faces. Its dual is the icosidodecahedron, which is the Archimedean solid
obtained by deeply truncating the dodecahedron or, alternatively, deeply
truncating the icosahedron.

To make a paper model of the triacontahedron, one can easily find a net on
the internet – a flat connected version of all the faces that, when assembled,
gives the polyhedron – to download, print, cut out, and tape together. In this
case, I wanted a version where the faces are decorated, and where I could
print out or laser cut each piece. To be able to do that cleanly, I affixed
tabs to the edges. Yet tabs introduce an issue, because they require space.
Acute angles between adjacent faces of any given net might not allow for
the insertion of tabs. I therefore dissected the net into pieces so that on each
piece no adjacent faces have exterior edges making acute angles. To make
matters more mathematically interesting and less demanding to produce, I CAROLYN YACKEL
made all of the pieces congruent. Mercer University
laser-cut paper
Note that this dissection condition came from the medium: paper is an
inexpensive way to construct a solid, and it is easy to decorate. I added
the congruence condition so that the number of unique files created for the
printer or laser cutter is kept to a minimum. Each face could be cut indi-
vidually with tabs on all four sides, but it is preferable to have more faces
connected, because then the paper is automatically hinged between these
faces, which constrains their relationship in space and creates structural
integrity.

With a ten three-face pieces cut and ready, it seems simple to assemble the
triacontahedron. Surprisingly, there are many ways to begin assembling with
the pieces and get stuck so that the desired polyhedron cannot be completed.
The reader might wonder why we don’t use five- or six-face pieces, for then
only six or five copies would have been needed, respectively. As with many
projects in mathematical art, this enterprise has proven to provide math-
ematically interesting questions as well as illustrations.

For more information on the rhombic triacontahedron:

http://mathworld.wolfram.com/RhombicTriacontahedron.html
27
We are making paper with constant negative curvature, in order to demon-
strate various features of the hyperbolic plane. A flat piece of paper has
zero curvature, a sphere has constant positive curvature, and our paper has
constant negative curvature. Our project builds on earlier work of Roger
Alperin, Barry Hayes, and Robert Lang.

We hope that this paper can serve as an instructional tool for college-level
geometry courses. Our goal is to explore hyperbolic versions of many of
the things that we can make from flat paper: straight edge and compass
constructions, tilings, origami, and so on. Paper is a natural medium for this,
as it has very little elasticity, so the intrinsic geometry is preserved when it
is smoothly deformed; yet unlike fabric it holds a crease so we can generalize
ideas and methods of flat paper folding to get new hyperbolic designs.

We learned a lot of math in the process of creating this paper! For example,
we needed to learn how to isometrically parameterize the pseudosphere
(top right) and related constant-curvature surfaces such as Dini’s surface
(left) from the upper half-plane model of hyperbolic space. This project
also enabled us to experience hyperbolic geometry in a physical way, which
offered interesting surprises despite our existing theoretical understanding.
For instance, we knew from local isometry theorems that the paper should
fold over onto itself and fit perfectly; yet when holding the paper in our
hands it was hard to believe it would actually work. (It does!)

We are still trying to figure out the best way to make this paper. Our first
attempt used papier mâché, but it was too stiff. Following advice of John AARON ABRAMS &
Edmark, Rotem Tamir, and others, we improved our process by making flat STEPAN PAUL
strips from pulp, then introducing curvature by pressing the strips onto a
pseudospherical mold (top right) while they were still wet. We hope to make Washington & Lee University;
larger, sturdier, more robust paper by starting with higher quality fibers. Harvard University
homemade paper, 3D-printed
plastic, molded silicone

Video showing local isometries, rotating and translating paper on the surfaces:
https://www.youtube.com/watch?v=JRd928WpY9w
29
For many years, I have used sewing and quilting as ways to understand
geometric structures. The first structures I made were crocheted hyperbolic
planes. I was really fascinated with the idea of making them homogeneous
and isotropic, meaning that there is no special direction or location. I was
crocheting at a conference and a friend of mine, Stu Ramsden, suggested that
I try making a Klein Quartic, which is a bit like a platonic solid. A Platonic
solid is a solid body made of copies of a regular polygon (a triangle, square
or pentagon) with the same number of polygons meeting at each corner. For
example, the dodecahedron is made of 12 pentagons where three pentagons
meet at each corner. The Klein Quartic is made of heptagons – a polygon
with seven sides – that meet together in groups of three at each corner. This
quilt (not pictured) gave me the ability to touch the Klein Quartic, and also
visualize symmetries in a way that digital models don’t.

As part of the Illustrating Mathematics program at ICERM in Fall 2019, SABETTA MATSUMOTO
Rich Schwarz, a math professor from Brown, gave a course called Geometry Georgia Institute of Technology
and Illustration. He covered many topics in this course, including translation
hand-sewn fleece, and rice
surfaces, which are surfaces that are made from a flat (Euclidean) polygon
with a set of rules that determine which sides are glued together. During
a discussion over lunch, Pierre Arnoux mentioned that his colleague Maki
Furukado had made a beanbag of a translation surface that has genus 2,
meaning that it has two holes. It was very delicate, so I offered to make him
another one out of fleece, so it could be easily manipulated. Working with
Pierre, I created the surface from four rectangular strips of fleece. This type
of zippered rectangle construction is called a suspension surface, because it is
the suspension of an interval exchange transformation.

There are other ways to make a genus-2 translation surface. Several of my

colleagues at ICERM were studying billiards problems on pentagons (see
Multiple ways to illustrate the same thing: Diana Davis and Samuel Lelièvre),
using the double pentagon translation surface. This surface is made by taking
two pentagons with one pair of sides glued together, forming a polygon with
eight sides, and gluing the other pairs of edges that are parallel. I made this
surface, too, but it was much harder to manipulate than the zippered rect-
angular surface. It only has one vertex, which means that all of the corners
of both pentagons all meet at a single point. These two physical versions of
genus-2 translation surfaces helped me to understand not only what transla-
tion surfaces are, but how and why mathematicians find them useful when
they are studying very different types of problems.
31
LASER CUTTING
Since mathematical objects are infinitely precise, it is always
a little disappointing to cut out some pieces with scissors
or a bandsaw, fit them together, and find that they are not
quite right. The laser cutter eliminates much of this issue, by
cutting in precision of up to 600 DPI, creating cuts, folds,
curves, tabs, holes, and other mathematical elements in, for all
practical purposes, exactly the right spot.

Of all the techniques presented here, the laser cutter is the

most dangerous, as it can melt and warp materials, create
toxic fumes, and start fires. It must be watched at all times
when in use, and it is not suitable for children. And yet, if you
wish to create a physical manifestation of any 2D object, the
laser cutter is almost certainly the best tool.

32 33
This is a tesselation with twin dragons. The twin dragon is a 2-reptile fractal,
meaning that it can be replicated from two smaller copies of itself. It illus-
trates that all pieces of the same size are translations of each other. We look
for things that are mathematically well-known, beautiful, and accessible and
engaging with the general public, and the twin dragon fits our criteria well.

We chose, as we often do, painted wood. People like the feel of wood, and
we love the flexibility of an unlimited color palette, which we get by painting
the wood before we cut it.

We first made terdragons, which are closely related 3-reptiles. Tesselated

terdragon pieces of a given size appear in up to six orientations. We were
surprised to discover that twin dragons of a given size all have the same
orientation.
HEIDI ROBB &
We could only find descriptions of how to construct the interior of the twin PETER BENSON
dragon, and we used Adobe Illustrator to trace the boundary, but it wasn’t Cherry Arbor Design, LLC
good enough to create interlocking tiles. We were fortunate to be introduced
to Dylan Thurston via Sarah Koch while Dylan was visiting the University laser-cut painted wood
of Michigan. We discussed the problem, and Dylan devised a clever substitu-
tion scheme for constructing the twin dragon starting with a single skewed
hexagon. From there, it was obvious how to use an L-system to directly
construct a boundary with whatever level of detail we wanted, and the pieces
interlocked perfectly.

Our first efforts at cutting actually had too much detail. The pieces had
fantastical lacy edges that interlocked, but as you tiled with them, the lacy
bits would break off. We then dialed in just enough detail to make the pieces
sturdy.

Further information:
https://www.cherryarbordesign.com/tessellating-with-twin-dragon-fractals/
35
There are many possibilities for the shape of the universe in which we
live. To understand these better, topologists often imagine a surface that
divides the universe into simpler regions. Imagine a surface that is somewhat
translucent, and is sitting in the middle of one of these possible universes.
As we look out, we see an amazing picture. Because of the way both the
universe and the surface fold back on themselves, we see different shades of
light and dark wherever we look, creating an amazingly complex and self-
similar pattern.

In this piece we show one such view of one such possible universe, laser cut
in acrylic and side-lit with LED strips. In particular, this picture represents
a small part of the boundary of 3D hyperbolic space. We view this space as
coming from the universal cover of some 3-manifold, containing a topologi-
cally significant surface. We created a depth map based on how many lifts
of the surface you pass through as you move out a fixed distance from some DAVID BACHMAN
point of hyperbolic space toward the boundary. This is joint work with Saul Pitzer College
Schleimer and Henry Segerman.
laser-cut acrylic
This is a laser-etched piece of acrylic, with carefully designed lighting to
illuminate the etching. I had never tried edge-lighting an acrylic relief, and
thought this would be a perfect way to create a visualization of our results.
I was surprised at how beautiful it turned out!

There were a lot of experiments to get the lighting effect correct. First, I
built a custom frame that hides an LED strip light. I then learned that the
acrylic needed to have a clear boundary where it met the frame so that light
propagated through it. Then I needed to cut the acrylic again, make modifi-
cations to the frame, and so on.

37
These lizards illustrate a tiling of the plane by congruent pieces, following
a similar picture created by the mathematical artist M.C. Escher in 1943. I
have loved tilings for years, so when I learned to use a laser cutter, I wanted
to create this beautiful tiling, as a sort of puzzle for myself and my students.

This tiling is based on the tiling by regular hexagons: As in the hexagon

tiling, you can check that here, every tile touches six others. Also, around
each lizard there are three points of rotational symmetry of order 3: at the
left cheek, the left foot, and the right knee.

This tiling is three-colorable, meaning that if you have three colors of tiles,
you can assemble the tiling in such a way that no two tiles of the same color
touch each other. I chose to make lizards in three contrasting colors, so that
it would be possible to assemble them in a three-colored way, as shown.
DIANA DAVIS
I created a file to cut a sheet of acrylic plastic into 66 lizards, and then Swarthmore College
ran the same file on plastic sheets of three different colors. I engraved some
laser-cut acrylic
details on one side of the lizards, which both gives them some personality
and helps with orientation: in order to fit together, all of the lizards must
either have the engraved side up or the engraved side down. The width of the
laser (the kerf) is small but nonzero, so there is a bit of wiggle room between
the pieces, making them easy to put together.

I knew that the tiling was three-colorable, but I didn’t realize until I put it
together that all of the lizards of a given color would be in the same orienta-
tion: for example, the pink lizards in the picture are all facing to the left. I
also didn’t expect how delightful it would be to assemble the tiling. In the
same way that coloring a coloring book is said to approximate a medita-
tive experience, fitting the lizards together is also a calming, satisfyingly
tactile process. This is especially true when assembling them in a 3-coloring,
because after you choose how to put together the first two tiles, the rest of
the tiling is determined.

Escher’s original artwork: M.C. Escher, Reptiles. Lithograph print, 1943.

39
These tiles originated in a project with Jonah Ostroff and Lucas Van Meter
where we created variations of the classic card game SET for non-abelian
groups. Such a game requires illustrations of group actions that can be easily
understood, and so we made these tiles, along with versions for several other
non-abelian groups, in order to experiment with the design of these non-
abelian SET games.

This medium allows for tactile interaction with what is usually a purely
abstract concept. By physically rearranging the wooden pieces, the players
can test and visualize the multiplication in this group; small wooden pieces
are ideal for this type of manipulation. We have also tried playing with paper
cards, and while the mathematical content is the same, the laser-cut wooden
tiles are much more aesthetically pleasing to touch and move.

This particular collection of wooden tiles is a way to visualize the semi-direct CATHERINE HSU
product of the symmetric group S3 acting on three copies of Z/2Z. Each tile University of Bristol
represents one group element: the three lines give an element of S3, and the
laser-cut wood
three dots give an element of (Z/2Z)3. We can illustrate the group action by
concatenating a series of tiles, and then reading off the resulting element from
the lines formed between the tiles and the dots, modulo 2, along each line.

I learned lots of interesting details about permutation groups and their group
actions by creating this illustration. For example, I discovered that if you
order the three transpositions in S3, then the fi rst two transpositions always
compose to the same 3-cycle as the second two transpositions. More generally,
I found that playing with these tiles gives an intuitive understanding of the
algebra used to construct their underlying groups, especially in the case of
semi-direct products.

More variations on the classic game SET:

https://people.maths.bris.ac.uk/~zx18363/set.html
A video of my talk about this project:
https://icerm.brown.edu/video_archive/?play=2063
Further information: Cathy Hsu, Jonah Ostroff and Lucas Van Meter, Projectivizing
Set. Math Horizons, 27:4 (2020), pp. 12–15.

41
This is a regular dodecahedron composed of twelve regular pentagonal faces,
three meeting at each vertex. It is one of the five Platonic solids. Plato
claimed that it was “used for arranging the constellations on the whole
heaven,” and so it was associated with the universe, representing mystery
and meditation. During the time I was building the piece, I was learning
about how to think of its group of symmetries – the alternating group A5,
which I was teaching in class as an abstract entity – in geometric terms.

My dodecahedron is built from twelve flat pieces of fiber board cut in a

laser cutter. I chose laser-cut wood because it is an organic material, and as
such it has a very different, warmer presence than, say, 3D-printed plastic
or metal, inviting one to touch the piece. Wood presented certain chal-
lenges, given the objective: a construction where the pieces are assembled
without glue, held by the springiness of the wood that desires to return to
its flat state. The solution uses a technique called kerf cutting where special GREG MCSHANE
incisions have been made to allow the wood to bend, in order to form the Université Grenoble Alpes
shape of the solid in three dimensions.
laser-cut fiber board
I coded all the cuts myself using elementary geometry – scalings, transla-
tions, rotations. Even now I am amazed at the strange outcomes of applying
transformations in the wrong order.

This piece is the result of much frustration through trial and error – an
application of the dichotomy method for finding a solution among infinitely
many choices. There are several parameters one must tweak in order to
achieve the desired result: the width of the strip between pentagons, the
number and size of the kerf cuts, the size and position of the jigsaw joints
where two pieces are attached.

43
This mirror box illustrates billiards on a square billiard table. Here, instead
of a ball bouncing off of the sides, a laser beam reflects off of mirrors. The
nice thing about using a laser beam instead of a ball is that you can see
many segments of its path all at once, whereas with a ball you would have to
remember where it went after you have seen it bounce around. The picture
shows two periodic paths, of period 4 (upper right) and 6 (lower left). We
have made several versions of this table over the years, and we have brought
them to many math fairs for children.

Since we wanted to be able to pack up the box and carry it easily, we made
it in seven pieces: four mirrored walls, and three interlocking pieces that fit
together to make a large square base (not shown). The corners of the walls
have rectangular protrusions that interlock, and there are the same rectan-
gular protrusions at the bottom of each wall (shown here at the top) that
fit into rectangular holes in the base. Fitting the walls into holes in the base ALBA MARINA MÁLAGA
ensures that the box is perfectly square, and not just a rhombus. Unfortu- SABOGAL & SAMUEL
nately, we made the base out of cheap plywood, which warped significantly LELIÈVRE
after a few weeks, making it unusable with the laser.
ICERM; Université Paris-Saclay

There are several considerations when cutting mirrors with a laser cutter. laser-cut mirrors, carpenter’s laser
First, if you cut them mirror side up, the laser beam is reflected back into
the machine, and may damage the machine, so best practices say to cut it
mirror side down.

Second, there are two kinds of mirrors. Back mirrors, which have a mirrored
surface coated with glass, are by far the most common; we used them for the
box in the picture. They are cheap and widely available, and the mirrored
surface is protected by glass, so if you touch it, you can easily clean it. The
downsides are that (1) passing through the glass dilutes the strength of the
laser, so it becomes invisible after a few bounces, and (2) the laser is not
bouncing off of the surface of the wall, but is bouncing off of the mirrored
surface a few millimeters behind the wall, which introduces error that
accumulates after several bounces. Front mirrors, which have the mirrored
surface on the face of the material, are much more expensive and you must
take care to keep them clean, but they provide an excellent reflective surface
that avoids the two issues described above.

Further information and laser cutting files:

https://imaginary.org/hands-on/mirror-rooms
45
This project illustrates real projective triangle tilings. Much as squares, equi-
lateral triangles or regular hexagons tile the 2D plane in a regular way,
we can make a regular tiling of other spaces as well. The mathematical
artist M.C. Escher famously made pictures showing triangle tilings of hyper-
bolic space, imagined as a disk in which straight lines are given by arcs of
circles (upper right). Here we look at similar tilings, in projective geometry
instead of hyperbolic geometry. I illustrated these as a stack of coasters,
with each coaster displaying a way to tile a portion of projective space with
triangles. Furthermore, the tilings in the set are all deformations of each
other, meaning that mathematically, we can interpolate from one tiling to
the next through a continuous path of tilings. Some sense of this deformation
is given by laying all the coasters out sequentially (bottom).

These tilings form a continuous family, so it would be best to experience

this via 2-dimensional “stills” from different times during the deformation. STEVE TRETTEL
Coasters turned out to be much a fun way to do this. I also created an ICERM
animation video showing the continuous process (see link below).
laser-cut wood
Compared to working only via computer, there is definitely something to
be said about producing physical objects. Looking at some members of the
woodcut family as others were printing allowed me to notice new features
about the differences between nearby tilings, and get a better sense of what
is geometrically happening when one tiling deforms into another.

It took some trial and error to get the mathematical image in my mind
and the result of the laser cutter to match up. As in hyperbolic space, the
triangles here get smaller and smaller as they approach the outer rim. This
leads to complications when a triangle is smaller than the laser beam width!
After making a few test cuts that were either burned around the border (too
many triangles) or had gaps in the interior (too few triangles) I took a new
approach, and rewrote the code to automatically discard triangles with area
or height below a certain threshold. This allowed me to generate a large
number of triangles, making sure to get all of the important information for
the picture, while ensuring that the machine was only given instructions to
engrave triangles big enough to be resolved by the laser.

Animation showing the continuous path of tilings: https://vimeo.com/368068637

The original procedure for producing the symmetry groups used to create these
images comes from: Anton Valerievich Lukyanenko, Projective deformations of
triangle tilings, master’s thesis (2008). 47
GRAPHICS
One of the most tantalizing mathematical illustrations ever
is what I like to call the “asterisk picture” of the Mandelbrot
set. Mathematicians had the idea of applying the function
f(z) = z2 + c over and over, starting with the input z = 0,
and using each output as the next input. They wanted to
know what happens to the image, for different points c in the
complex plane. Using a very early computer, they set it to
work on this problem, and for each representative point in the
box [-2, 2] x [-2, 2], they had the computer print an asterisk
if the image stayed bounded, and print nothing if the image
went off to infinity. The result was a shape that no one had
ever seen before. To this day and beyond, exploring math-
ematics through computer graphics continues to elicit a sense
of surprise and awe.

48 49
I am interested in the dynamics of the Riemann zeta function and its
relatives. I wanted to prove their fractal nature, and to visualize and charac-
terize them, particularly the neighborhoods of their complex zeros.

This is a unit square of the iteration fractal of the Riemann zeta function
centered on the non-trivial zero located at approximately 0.5 + 28979.4096i.
This is one of many examples where the non-trivial zero lies on an “island” of
fractal surface close to, but separated from, the main “arms” that extend up
and down the imaginary axis. Each of the protrusions emanating from the
“island” (and the protrusions emanating from those protrusions) are scaled
and distorted reproductions of the main fractal arms. This is to be expected
for a fractal that exhibits self-similarity at all scales.

The iteration fractals of the Riemann zeta function, and its close relatives
(the Hurwitz zeta functions, and certain Dirichlet L- f unctions), all exhibit DAVID RAINFORD
the same essential details: they have discrete clumps of fractal surface, with Senior Counsel at Taylor Vinters
occasional separated “islands,” all connected by points with predictable
computer-generated graphic
values, with their complex zeros situated in neighborhoods with predictable
characteristics.

Working with computer graphics permits me to make arbitrary changes

in scale in order to zoom out or zoom in, concealing or revealing detail, at
will. There are two key classes of pixels rendered in the final image: those
corresponding to starting values that blow up upon iteration (blue), and
those that converge to a fixed point upon iteration (orange). Once I figured
out the classification, it was straightforward to choose a color scheme that
classified the pixels and reflected their particular iteration counts. Another
obvious optimization would be to use interpolation to smooth the transitions
between patches of color, but I’m happy with the hard transitions for now.

Blog posts about this and similar images: https://primepatterns.tumblr.com/

51
For years, I have struggled to find good techniques for coloring mathematical
surfaces with patterns. Coloring a sphere is easy, because it’s so round and
regular, but the bands in this image have not been easy to color well. My
goal is to find conformal coordinates on this shape, which means that all
the angles in the source pattern are portrayed correctly on the surface.
Conformal coordinates make patterns more recognizable. I heard Steve
Trettel’s talk, “What does a torus look like?” and that sent me back to an old
project, which involved coloring this set of bands with a wallpaper pattern.

The innovation shown here comes from making the shape in software called
Grasshopper (a plug-in for an architectural design software called Rhino).
The good news is that I was able to paste a nice wallpaper pattern onto the
surface and then make it look nice in Photoshop. The bad news is that the
coordinates are not conformal: the pattern is indeed distorted.
FRANK A. FARRIS
Grasshopper has a growing community of mathematicians who support each Santa Clara University
other in solving problems, which was very helpful: I had been coloring this
computer-generated graphic
shape by coloring a gigantic mesh, with approximately one vertex for each
pixel of the coloring pattern. It really bogged down the computer. When
I asked the Grasshopper community about it, Daniel Pinker, the creator
of Kangaroo (another Rhino plug-in), solved my problem perfectly. I’ll be
working with this new technique for a long time.

It is still a long road to finding conformal coordinates on this particular

structure, but I learned a lot about coloring tori. Surprisingly, the best way
to find coordinates on donut shapes is to look for them in four dimensions,
where they lie nicely on the 3D sphere.

The alert reader will see that there’s an error in the image shown here: the
pattern fails to match on the left side of the higher horizontal band. After
all my Facebook friends had told me how beautiful it was, two mathematical
artists looked at it for two seconds before saying, “Oh, but there’s a mistake
right there!” It’s wonderful to be working in a community of people who take
time to look closely at things.

Further information:
Wilder Boyden and Frank A. Farris, Polyhedral Symmetry from Bands and Tubes,
Journal of Mathematics and the Arts, to appear.
53
This image shows part of an orbit of a group of matrices over the ring of
integers in an imaginary quadratic field. In particular, it is the projective
special linear group over the the Gaussian integers. These matrices act as
Möbius transformations, taking circles to circles. Asmus Schmidt used such
arrangements of circles as a way to break up the complex plane into pieces,
for the purposes of a continued fraction algorithm.

Schmidt’s original work took place before the advent of powerful computer
graphics systems, and his papers included some figures showing only half a
dozen or a dozen circles, illustrating only one level of the recursive structure.
I wanted to more fully understand the structure, and I wanted to find out
what happened in other imaginary quadratic fields. Using computer graphics
allowed me to work interactively, changing parameters and redrawing
frequently. It was also important to me to be able to print out the pictures
and work with colored pencil, ruler and compass on top of them. For example, KATHERINE E. STANGE
in some pictures the human eye picks up “ghost circles” that appear to fit in University of Colorado Boulder
the geometry but are missing. I measured these with a ruler and compass to
computer-generated graphic
conjecture their exact form, and then proved that they exist.

Easy access to exploring these pictures led me to conjecture and prove

a variety of things about the relationships between the arithmetic of the
imaginary quadratic field and the pictures themselves. For example, the
circle arrangement is connected if and only if the field is Euclidean.

The experience of programming the pictures was itself enlightening: for

example, a naive algorithm exploring the Cayley tree results in varying reso-
lution in different parts of the picture, and inadvertent congruence conditions
add or remove extra layers of circles. Often, errors in coding led to useful
insights. Later on, I worked harder to make the pictures more beautiful, and
learned about color choices and vector graphics, among other things.

Further information: Katherine E. Stange, Visualizing the Arithmetic of

Imaginary Quadratic Fields, International Mathematics Research Notices, Volume
2018, Issue 12, June 2018, pp. 3908–3938.
55
While studying properties of graphs of complex functions, specifically level
curves of the real part of iterates of complex functions, we stumbled on the
first image shown here (top left). Our initial reaction was that this image
looked like something from a coloring book for adults. As we examined this
image more closely, we noticed that the Julia set for the function f(z) = z2 – 1
appears to lie in the middle. In addition, the decorations inside the Julia set
correspond to critical points for the real part of the function, all of which
are saddle points.

From there we decided to explore functions with interesting Julia sets to see
if we could find more artistic images, which we made into a coloring book
(see below). The image in the top right comes from a function whose Julia
set resembles the Sierpinski triangle. The flowers we see are from the poles
of the sixth iterate of this rational map. Similarly, the image in the bottom
left comes from a function whose Julia set consists of concentric circles, and
the poles of its third iterate appear as flowers.

For the image in the bottom right, we started exploring the interplay between
zeros and poles of different orders. By experimenting with the window, which
cuts off parts of flowers created by poles, we were able to create this image
of flowers in a window box.

Functions used:
top left: 6th iterate of z2 – 1, level 0 contours
top right: 6th iterate of z2 – 0.6/z, level 0 contours
bottom left: 3rd iterate of z3 + 0.0001i/z3, level –2 and 2 contours
bottom right: 5th iterate of 2(z – 3)2(z + 10)/z3, level 0 contours JULIE BARNES &
BETH SCHAUBROECK
Western Carolina University;
U.S. Air Force Academy
computer-generated graphics
Our coloring book:
Julie Barnes, William Kreahling, and Beth Schaubroeck, Coloring Book of
Complex Function Representations, Mathematical Association of America, 2017.
Further information:
Julia Barnes, Clinton Curry, Elizabeth Russell and Lisbeth Schaubroeck,
Emerging Julia Sets, Mathematics Magazine, Vol 88 (2), pp. 91–102, April 2015.
Julia Barnes and Lisbeth Schaubroeck, Any Way You Slice It, It Comes Up Julia
Sets, Mathematics Magazine, Vol 92 (1), pp. 3–16, February 2019.
57
This print, titled “Prime Goose Chase,” has a structure based on a tradi-
tional quilt pattern called Wild Goose Chase, while its content relates to the
Fundamental Theorem of Arithmetic. The integers from 1 to 256 are the
“geese,” and the prime decomposition of each integer is shown using colored
triangles. There are 8 columns of numbers, starting with a black triangle
representing 1 at the upper left. Solid triangles are used for primes, and each
prime is assigned a unique color: 2 = red, 3 = gold, 5 = yellow-green, and so
on, until 19 = magenta. As larger primes are needed, more colors are created
by adding white to these basic 8 hues. Composite numbers are represented
by subdivided triangles. For example, since 6 = 2 x 3, it is half red and half
gold. Powers of primes are shown using horizontal shades of the base color.

I believe that integers have personalities, largely based on their prime factors.
In this design, my goal was to produce a visual table that would show the
factoring of individual integers, and also make it easier to observe number MARGARET KEPNER
patterns. Independent artist
computer-generated graphic
This work was produced as a digital print using Adobe Illustrator. This
medium provided the precise shapes and range of colors my design required.
Developing this piece improved my appreciation of the integers and their
inherent rhythms.

In my first draft, the black spaces between the columns seemed too static. I
solved this by adding sloping gray bands connecting multiples of 6. This also
illustrated the fact that all primes, after 2 and 3, are plus or minus 1 from
a multiple of 6. I enjoy finding visual elements that reinforce mathematical
concepts.

Gallery of my mathematical artwork: http://mekvisysuals.net

59
I added a feature to the 3-manifold software SnapPy that shows you the
view you have when living inside a hyperbolic 3-manifold. I just tried out
some random census manifolds to see whether I get an image that would
make a nice demo. I was pleasantly surprised to find this one (s431). The
image shows the edges of the triangulation and a cusp neighborhood as you
look down a closed geodesic in the hyperbolic 3-manifold. Walking down the
closed geodesic once rotates the hyperbolic 3-manifold by 2/5 of a turn, and
you see a beautiful helix formed by one of the edges.

A 3D print of a tiling by fundamental domains is always distorted in some

way because we live in a Euclidean world (at least on human scales).
However, the image you see here is the true, undistorted inside view of the
hyperbolic 3-manifold.

This is still work in progress. For example, I am still trying to come up with MATTHIAS GOERNER
a good way to show which tetrahedra are the same in the manifold. I first Pixar Animation Studios
tried to do this by coloring a small ball about their incenters, but it turns
computer-generated graphic
out that our Euclidean intuition about where we the expect the incenter of
a tetrahedron to be is completely off in the hyperbolic world.

To download SnapPy: https://snappy.math.uic.edu/

61
Chip firing is a process on a graph, which is a network of vertices connected
by edges. Initially, a stack of “chips” is placed on each vertex. A vertex is
allowed to “fire” its chips if it has at least one chip for each vertex to which it
is connected (its “neighbors”). When a vertex fires, it sends one chip to each
neighbor. We repeat this process until no vertex has enough chips to fire,
which we call a “final configuration.”

The illustration shows chip firing on a cubical grid in 3D. Every vertex
has six neighbors, so it needs six chips in order to fire. In this example,
every vertex except for the origin started with 4 chips, and the origin
started with 100,000 chips. The final configuration forms a cube. Vertices
with 0, 1, 2, 3 or 4 chips are rendered as small cubes of different colors.
Most of the vertices have 5 chips in the final configuration, so we rendered
them as transparent, to exhibit the funnel-like shapes going into the cube
from each of its faces. The octant of the cube facing the viewer is also MARTIN SKRODZKI,
transparent, and only half of the lower octant is shown. This is to show CAROLINE J. KLIVANS,
the inner structure of the final configuration, which resembles “leaves.” AND PEDRO F.
FELZENSZWALB
A computer-generated rendering allows us to explore both the outer
ICERM; Brown University;
shape and the inner structure of the pattern interactively. Our original
plan was to 3D print the patterns inside the object, but they turned out Brown University
to be too fragmented, so a 3D print would fall apart. Having a digital computer-generated graphic
model allowed for switching on and off the visibility of certain vertices,
which made it easy to identify larger structures within the geometry.

Initially, we represented each vertex by its own cubical geometry object

in the software Blender, but this yields too many objects, and crashed the
Blender GUI. Therefore, we decided to create only six objects, each repre-
senting one of the numbers of remaining chips in the final configuration.
Using the clipping plane built into the Blender camera, we could also clip the
object and thus show any planar slice of the illustration. However, we wanted
spherical, not planar clipping. Thus, we decided to build more geometries
again: six geometries for each “shell” around the origin. That is, all cubes of
a fixed radius distance to the origin and of the same color were collected in
one object. By hiding the outside shells, it became possible to interactively
show the internal layers of the structure.

Further information:
Caroline J. Klivans, The Mathematics of Chip-Firing, CRC Press (2018).
63
This illustration shows several tilings of the hyperbolic plane, in various
different projections, or models. Hyperbolic geometry uses a set of axioms
that are slightly modified from the Euclidean geometry that students learn
in high school. It enables constructions of tilings that would be impossible
in the Euclidean plane: for example, the alternating regular hexagons and
octagons in the lower-left image. I’ve been interested in 2D and 3D tilings
for many years, and I find the repetition of shapes with varying sizes and
curving distortions in hyperbolic tilings to be particularly mesmerizing.

These images are all generated by a fragment shader that runs on a comput-
er’s graphics processing unit (GPU). I got into this style of programming
through the web site shadertoy.com, which hosts shaders for a community
of computer graphics hobbyists and professionals. Generating images like
these in a pure fragment shader is challenging, because it inverts the typical
paradigm of graphics programming: Instead of starting from a description of MATT ZUCKER
a polygon’s vertices, and coloring in all of the pixels it covers, the program Swarthmore College
must instead decide independently for each pixel which of the many possible
computer-generated graphics
polygons it lies in, and color it accordingly. I wrote the shader that generated
these images, which you can find in the link below. Figuring out how to
write it was a really enjoyable puzzle that unfolded over many months.

One of the biggest lessons I learned along the way was that I had started
out using the wrong projection of the hyperbolic plane to do most of my
math. Initially, I was using Poincaré disk coordinates to represent points on
the plane. In the Poincaré disk model, there are two types of lines: circles
orthogonal to the disk, and diameters of the disk. The coordinates I was
originally using required inefficient and over-complicated code to correctly
handle the two different types of lines. Another big problem I ran into was
lack of precision close to the edge of the disk. Switching everything over to
the upper hyperboloid model led to much simpler code and improved perfor-
mance, but it took me a while to figure out the best way to program every-
thing. Between a couple of math papers I dug up online and some Twitter
correspondence with other graphics programmers, I learned how to perform
all of the necessary geometric operations both efficiently and accurately.

The shader that generated these images: https://www.shadertoy.com/view/3tsSzM

65
VIDEO AND
VIRTUAL REALITY
People like to say that hyperbolic space is “cold, dark and
lonely,” but what would it really be like to wander around in
hyperbolic 3-space? As soon as computers had the requisite
capabilities, mathematicians were making videos to commu-
nicate what mathematical objects would look like in real
life, and to visualize what it would be like to move through
different geometries. Now that processors are much faster and
computers have excellent graphic capabilities, it is possible to
not only make a video, but also for people to direct their own
movement through these spaces in real time, in virtual reality.

66 67
In his video “Knots to Narnia,” William Thurston uses large wire knots to
demonstrate his concept of knots as portals, where he actually steps through
the knot to move back and forth between, in his conception, Earth and
Narnia (see previous page). Virtual reality gives us the ability to bring to
life this experience of actually stepping into other worlds and seeing, not just
imagining, what it looks like on the other side.

These pictures show the six-fold branched covering of order two of the trefoil
knot, generating the dihedral group of symmetries of the triangle. The three
outer loops of the knot correspond to reflections, and the inner region corre-
sponds to the rotation by 120°.

To create this virtual reality experience, I had to learn a lot about cyclic
branched covers of knots, as I had to construct them in the software imple-
mentation. This led me to discover a construction from Poul Heegaard’s MORITZ L. SÜMMERMANN
dissertation from 1898, which could be implemented to simulate these portals University of Cologne
in virtual reality. It consists of gluing a cone to the knot, which serves as the
virtual reality
branch cut, along which the different worlds are glued together.

At first, I thought about implementing portals as surfaces somehow spanned

by the knot. The most obvious choice was to try Seifert surfaces, but this
turned out to be a dead end. The next attempt was to not construct the
branched covering, but only simulate it through the knot projection on the
screen. This approach used a variant of Reidemeister moves to keep track of
the worlds the regions led to, but it turned out to be quite complicated and
unstable at the crossings. Luckily, I then found the reference to Heegaard’s
construction in John Stillwell’s “Classical Topology and Combinatorial
Group Theory,” and I was able to implement the virtual reality world.

It was a challenge to reduce the computational load of the software, which

has to compute which world to show for each pixel of both screens in the
head-mounted display. I was able to achieve this by using shaders to offload
much of the calculations on the graphics processing unit, thanks to some tips
from Roice Nelson. The main computational load is now the rendering of the
worlds. I wanted to create worlds that are interesting to look at, but not so
interesting as to take away the focus from the knot itself.

Software download and information: https://imaginary.org/program/knotportal

Thurston’s original video, “Knots to Narnia” :
https://www.youtube.com/watch?v=IKSrBt2kFD4
69
In 2018, Olga Frolkina of Moscow State University published a proof of the
impossibility of packing an infinite number of Möbius strips into a finite
3D space. Evelyn Lamb wrote about the proof for Quanta Magazine, and
I created this animated illustration to go along with the article. This is
visual storytelling rather than an accurate simulation: I had the idea to show
Möbius bands being forcibly crammed into a container and bursting back
out, hinting at Frolkina’s result. Since it’s impossible to render an infinite
number of anything, I thought “a lot” should do just fine, and cloned as many
Möbius bands as my computer would handle.

Working in 3D simplifies and expedites certain processes – I would not have

tackled this project if I had to draw out dozens of Möbius strips and then
animate them frame by frame, manually! Working in Cinema 4D allowed me
to model just one strip, clone it, and then run a simulation. Eliminating the
grunt work allows for more ambitious ideas, and more time for creativity. OLENA SHMAHALO
Also, in 3D software it’s possible to both simulate real-world dynamics accu- Quanta Magazine
rately and ignore them at will, so you can create situations that are believ-
computer-generated animation
able, but physically impossible, like magic!

Despite being a particularly mathematically-inclined artist, I’m not a math-

ematician. So I knew nothing about this kind of packing problem prior to
reading Evelyn’s article. Although I wouldn’t expect anyone to learn much
just from viewing my illustration, I hoped that it would spark intrigue, and
invite people to read about this fascinating research.

Coincidentally, the difficulties I encountered in creating this illustration

echoed the conclusions of the mathematical result – it really is challenging
to cram Möbius strips into a 3D space (albeit for different reasons). In theory
this should have been a relatively simple simulation: just pour a bunch of
shapes into a container, then have a piston come down and stuff them in.
In practice, though, all sorts of weird stuff happened. The Möbius strips
would glitch and jiggle around wildly, sometimes they’d pop right through
the container like ghosts through walls, and getting them to look sufficiently
smushed and then pop out dramatically was no picnic either. I had to keep
tweaking various parameters to get the whole system to behave. But this is a
common enough occurrence with projects; there might as well be some kind
of pseudo-scientific “law” saying that “things that seem easy to do are usually
much more complicated than one might anticipate.”

The article with the animation: Evelyn Lamb, Möbius strips defy a link with
71
infinity. Quanta Magazine, February 20, 2019.
 These images are inside views of famous 3-manifolds, with their geome-
tries modeled by Thurston geometries. Such spaces date back to the famous
Thurston geometrization conjecture, proved in 2003 by Grigori Perelman.
The theorem states that every compact 3D manifold decomposes into pieces
whose geometry is modeled by Thurston geometries. The even more famous
Poincaré conjecture is a corollary of this theorem.

It is difficult to visualize such beautiful spaces due to complications imposed

by topology and geometry. The main effort for visualization such spaces was
at the Geometry Center from 1994 to 1998 under the leadership of William
Thurston. This program studied and disseminated modern geometry using
interactive visualization based on the traditional rasterization pipeline (from
computer graphics). Two problems arise: the scene must be replicated to
“unroll” spaces with complicated topology, and to rasterize a scene it is
necessary to compute rays between the scene points and camera position TIAGO NOVELLO DE BRITO
(a hard problem). To produce our images we use a ray-tracing framework, Instituto de Matemática Pura e
RayVR, that overcomes such difficulties by operating intrinsically in the Aplicada (IMPA)
HYPERBOLIC SOL
geometry and topology, searching for the scene objects by tracing rays from
virtual reality
the viewer towards the pixels. RayVR is a project of the VISGRAF Labora-
tory, which I developed with Vinícius Silva and Luiz Velho.

Our images illustrate four 3-manifolds with their geometries modeled by

hyperbolic, spherical, Nil, and Sol geometries – we think these are the most
interesting Thurston geometries. Such manifolds are extremely important in
the study of 3-manifolds, and we can visualize being immersed in a scene
where powerful modern mathematics was developed.

The hyperbolic and spherical manifolds are both obtained by gluing,

with appropriate rotations, opposite faces of special regular dodecahe-
drons in hyperbolic and spherical spaces. A monkey’s hand, or face and
hands, are added to the scene to give more context. The Nil and Sol
manifolds are obtained by identifying faces of the cube. In both cases,
two faces are identified by translation, and we identify the remaining
faces in such a way to give rise to the Nil and Sol manifolds. The face
pairing makes the rays iterate, giving a tessellation of Nil and Sol spaces
by cubes.

For associated videos, papers, technical reports and texts:

https://www.visgraf.impa.br/ray-vr/
73

NIL SPHERICAL
Algebraic surfaces are the 2D solution sets of systems of polynomial equations.
For example, a cylinder is an algebraic surface, since it can be described as
the set of all points (x, y, z) so that x 2 + y 2 =1. Algebraic geometers like to
fill in some missing points “at infinity” by considering these surfaces inside
of projective space. For the cylinder, this adds one extra point, infinitely far
away, where all of the straight lines on the cylinder meet up.

When we learn to compute in projective space by hand, we work with several

sets of compatible equations in different “affine charts.” It can be unclear
how the pictures in these charts relate to each other, or how they form a
geometric whole. The goal of my illustration is to make the projective nature
of algebraic surfaces viscerally intuitive, by allowing the user to “click and
drag” from one affine chart to another. I do this in virtual reality with an
HTC Vive headset.
SEBASTIAN BOZLEE
There seem to be two kinds of geometers: those who imagine that their University of Colorado Boulder
mathematical objects are something that could fit on your desk, and those
virtual reality
who imagine their mathematical objects are room-sized, something you
could travel through. I have always been the first kind of geometer, but I
have been curious to experience the second point of view. A computer visu-
alization on a monitor reinforces the first point of view – after all, the image
on the monitor is only a few inches wide. Virtual reality, on the other hand,
allows us to experience the second one.

Originally, I displayed the algebraic surface in 2D, and the user manipulated
it using a keyboard and mouse. The result was completely unsatisfactory:
it was too difficult to understand and control. Switching to virtual reality
makes it much easier to get a sense for the shape of the surface, and allows
the user to interact with the surface by moving in 3D, which is a much
more natural way to interact with something in 3D. Also, when I first got
the program running in VR, the surfaces didn’t have the coordinate grid on
them, which made them into big blue blobs that were hard to read. Adding
the grid makes it easier to get a sense for the shape of the surface when it
is far away.

Video of VR experience and further information:

https://www.youtube.com/watch?v=MYfUbeiHkfU
75
How do you illustrate high-dimensional spheres? I wanted an image that
would convey a sense of interleaving, as equatorial discs are revolved to
spin out higher-dimensional balls. I also wanted to capture the beauty and
mystery of these objects: the “music of the spheres” that animated the imagi-
nation of cosmologists of old.

In my multivariable calculus video course, one of the capstone projects is

the computation of volumes of balls of radius r in arbitrary dimensions. The
sequence, starting with the familiar πr 2 in 2D and 4/3πr 3 in 3D, becomes
an elegant interlaced sequence of recursive formulae in arbitrary dimensions.

Computer graphics has no limits. All the geometric and topological entities I
imagine can be cast into colors and forms on the screen. Color is the ultimate
hard choice. Given that my calculus YouTube series is called “Calculus
BLUE,” I changed the initial greyscale images to an oversaturated blue- ROBERT GHRIST
themed scheme. University of Pennsylvania
computer-generated graphic
As a result of rendering and, especially, animating this sequence, I fell in love
with high-dimensional spheres in a more sensory as opposed to intellectual
manner. It’s one thing to “do the math” and another altogether to see the
thing unfurl.

Calculus BLUE videos: www.youtube.com/c/ProfGhristMath/

77
Mathematics, unlike other sciences, does not forget its ancestry. My purpose
was to reanimate and preserve, if only for one more generation, the math-
ematical experiments programmed long ago, but on now extinct computers
in dead languages, and with obsolete graphics libraries. The Italians say
“traduttore, traditore” – translators are traitors. But David Eck’s library
makes WebGL read like OpenGL, and some Javascript is reminiscent of C.

I chose to create interactive web animations for two reasons. Most impor-
tantly, it has been said that the web browser is the most powerful software
running on any device. Furthermore, learning how to download, install and
recompile code has fallen out of fashion. In re-programming the mathemat-
ical ideas in Javascript/WebGL from code that no longer compiles, I am
realizing anew just how ingenious those original translations from algebraic
expressions to computer code by my students and collaborators really were.
GEORGE K. FRANCIS
The collage shows four real-time interactive computer animations (at the University of Illinois at Urbana-
link below) of classical homotopies from the 1990s. Each tells a story of Champaign
its mathematics and graphical genesis, the better understood if you can
manipulate the camera, homotopy, and shapes from your web browser. You interactive web animation
are welcome to steal the code, to experiment and improve the animations.

(upper left) This immersed cylinder is almost halfway through its eversion. It
is the central topological detail in five homotopies, based on the pioneering
work on sphere eversion by Bernard Morin. It was originally programmed by
Apéry and Chris Hartman in C/IrisGL for the CAVE at the NCSA (1992).

(upper right) This image of the Etruscan Venus, a Klein bottle parametrized
by a Whitney Excellent mapping, conveys little of the ingenuity of its design
and its pinchpoint-cancelling homotopy by François Apéry. Watch it morph
through deformation into the Roman (Steiner) surface and the Boy’s surface.

(lower left) John Dalbec worked out a contraction of an unusual cell-complex

embedded in space, Zeeman’s Duncehat, which is contractable but not
collapsible. The surface has one vertex, and a single disk-like sheet attached
three times to the one edge.

(lower right) A Möbius Band, with a circular edge (yellow) morphs through
other bubble-like shapes.

Interactive animations: http://new.math.uiuc.edu/glsimEck/Collage

79
Tilings have shaped my journey through mathematics, in my research,
outreach, and artistic endeavors. Tilings are also commonly found in our
daily lives, making them a convenient starting point for bringing broad and
diverse audiences to mathematics.

The tilings that you see on a bathroom floor, or a terrace paved with bricks,
or in a wallpaper pattern, are periodic: you can find some part of the tiling
(often a rectangle), that you could cut out and use as a stamp, repeating the
same pattern over and over again to create the entire tiling. Said another
way, such tilings are invariant under a translation of the entire plane: you
could shift the whole thing by some vector, and the resulting tiling would be
the same as the original.

By contrast, the tilings I study and illustrate are aperiodic, which means
that they don’t repeat in any direction. In order to illustrate this property, DARÍO ALATORRE
I use animations, which not only show the difference between periodic and Institute of Mathematics,
aperiodic tilings – an aperiodic tiling will never overlap with a translate of Universidad Nacional Autonoma
itself, while a periodic tiling will – but also appeal to their dynamic nature.
de México (UNAM)
This dynamic nature of aperiodic tilings is one of my main research interests.
It turns out that one can talk about the space containing all of the tilings animated aperiodic tilings
that can be possibly constructed with a given set of tiles. This kind of space,
together with the translations of tilings mentioned above, constitute very
interesting examples of hyperbolic dynamical systems. They are laminated
spaces similar to solenoids (see 3D printing: Dina Buric), and there are still
many aspects about them that remain unknown.

When I started creating graphics and animations for illustrating aspects

of my studies, I initially used frame-by-frame animated gifs. Soon after, I
started exploring different methods, like Processing software and LiveCode
tools. This, in turn, led to a collaboration with a community of digital
artists, which I perceive as a great symbiosis.

These graphics and animations were created with Python matplotlib, Python
mode for Processing, and ImageMagick.

Animations of aperiodic tilings:

http://penelope.matem.unam.mx/media/photologue/photos/00a.gif
http://penelope.matem.unam.mx/media/photologue/photos/00c.gif
Gallery of more images: http://penelope.matem.unam.mx/galeria/
81
3D PRINTING
The best way to understand a 3D object is to hold it in
your hands. But how can we create an accurate model of a
surface, or another mathematical object? A hundred years
ago, mathematicians made models of surfaces out of plaster,
supported by metal rods. Many mathematics departments
and mathematical institutes are fortunate to have collections
of such surfaces, which are beautiful and instructive. Exactly
how people constructed the molds to create these surfaces is
somewhat mysterious.

These days, it is possible to 3D print objects that represent

mathematical surfaces, by using math software to plot the
surface, and then using 3D printing software to transform the
plot into a solid, printable object. Although the process is
simple in theory, 3D printing is a relatively new technology,
and as such there are usually technical hurdles to resolve
along the way, so any project usually requires several
failed attempts before the result succeeds in realizing the
mathematician’s vision.

82 83
This set of 3D prints illustrates singular algebraic surfaces. I work on
solutions to the particular problem of physically visualizing nodal singulari-
ties, where two or more pieces come together at a single point. My main moti-
vation for 3D printing them is to illustrate the output of the algorithm for
numerical real cellular decomposition implemented in my computer program
Bertini_real. It computes a union of “cells,” each equipped with a generic
point and homotopy that can be used to compute additional points on the
real part of a complex variety. This output is naturally 3D printable.

Fused fi lament fabrication (FFF) using thermoplastic polyurethane (TPU) is

the right material for these objects for a number of reasons. FFF can be done
at home or in the office with an inexpensive 3D printer, with no chemically
dangerous materials or supplies. Post-processing consists only of removing
support material, and since TPU is flexible, printed surfaces can have much
thinner connections at nodal singularities without inevitably breaking. SILVIANA AMETHYST
Furthermore, the very small earring-sized prints are robust enough to be University of Wisconsin‒Eau Claire
worn daily, and the skeleton-like object is seemingly delicate, but readily
3D-printed plastic
able to be carried in a backpack or pocket. I continue to learn about how to
engage the public in my research by having cheap non-breakable objects to
show them while wearing earrings at the grocery store. TPU is a challenging
material to work with, requiring a high quality hotend and a commitment to
printing very slowly, but it’s incredibly rewarding.

My early versions of prints of the Barth Sextic and other nodal surfaces, such
as those coming from the Herwig Hauser gallery of algebraic surfaces like
Octdong, broke either during support removal or transport for show. Many
of my most experimental polylactic acid (PLA) prints are thus doomed to
live forever trapped in their support material, since it’s so far inside a cavity
in the surface I would destroy it in completing it, or the material connection
between pieces is so small.

Gallery of more images: https://silviana.org/gallery

Further information: Silviana Amethyst, Daniel J. Bates, Wenrui Hao,
Jonathan D. Hauenstein, Andrew J. Sommese, Charles W. Wampler, Algorithm
976: Bertini_Real: Numerical Decomposition of Real Algebraic Curves and
Surfaces, ACM Transactions on Mathematical Software, 44(1), [10]. https://doi.
org/10.1145/3056528
85
Stereographic projection is a map from the sphere to the plane. Place a
sphere on the plane, and draw a straight line from the north pole of the
sphere down to the plane. The line goes inside the sphere, hits the sphere,
then continues on to hit the plane. Stereographic projection sends the point
on the sphere to the point on the plane. This is precisely what the light rays
do, so the grid pattern on the plane is the stereographic projection of the
pattern on the sphere.

I chose to illustrate this to fill a gap in my book, “Visualizing Mathematics

with 3D Printing.” I already had some prints showing the “3D shadow” you
get as the result of a higher dimensional version of stereographic projection:
from the sphere in 4D space to 3D space. So I needed to explain what stereo-
graphic projection is, and the lower-dimensional version seemed like it would
be easier to understand.
HENRY SEGERMAN
3D printing has a lot of nice features – it can be very precise, which I needed Oklahoma State University
in order to get a good shadow. A print is also physical – you can touch and
3D-printed plastic
play with it, unlike a computer graphics render. Also unlike a computer
render, there’s no way to “cheat.” People are rightly suspicious that some ...plus a flashlight.
effect could be faked with computer graphics, but they (think they) under-
stand everything about a piece of plastic and a flashlight – so it must be a
real phenomenon.

I started with the design on the plane, then made cones from the design
up to the north pole. Then I used the cones to cut out the windows. There
are some tricky issues to do with the thickness of the sphere. Ideally, the
sphere would have zero thickness, but of course that would be impossible
to print. Cutting out the cones works, but leaves very sharp angles near the
north pole of the sphere. So I need to cut a second time to remove these
sharp angles.

Video of the object in action: https://youtu.be/VX-0Laeczgk

Further information: Henry Segerman, Visualizing Mathematics with 3D printing,
Johns Hopkins University Press (2016).
87
More than anything, my work illustrates the human brain’s remarkable
ability to fill in missing information as it works to make sense of the data
it receives from the surrounding world. By varying the thickness of piping
generated along the sixth iterate of Hilbert’s famous space-filling curve, I
have rendered Hilbert’s 2D portrait as a 3D object. Using a Python script
within the modeling program Rhino, my code reads in an image and uses
pixel data to generate rectangular piping along the curve. The width of the
rectangular profile of the piping at a point depends on the intensity of the
pixel at that point; it is wider when the intensity of the pixel is lower (where
the image is darker) and thinner when the intensity of the pixel is higher
(where the image is brighter).

I am new to 3D printing, so I chose this medium in large part to gain

expertise in the modeling and printing processes of a 3D object. I worked in
stages, first learning how to print a Hilbert curve with square piping, before JUDY HOLDENER
incorporating a rectangular profile with variable width. I used a Python Kenyon College
script within Rhino to carry out the process, reading in image data from the
3D-printed plastic
famous (cropped) picture of Hilbert in his white hat.

I created this portrait using the sixth iterate of the Hilbert curve, which
I constructed in pieces – printing off 16 subsquares of the curve on an
Ultimaker 3. Each subsquare is approximately 5 inches in width, and was
generated using the fourth iterate of the Hilbert curve. Hoping to render
a higher resolution of the image, I was initially planning to generate the
portrait using the seventh iterate of the curve. However, such an approach
would have involved well over 140 hours of printing.

I decided I was happy with a lower resolution, which may require some
squinting on the part of the viewer. However, the precision improves dramat-
ically when the image is scaled-down (see thumbnail to the right). An unex-
pected discovery was the difference in finish and texture between the top
of each printed curve segment and the bottom. Preferring the shiny finish
of the bottom, I ultimately decided to print each subsquare of the curve
upside down.

Further information: Judy Holdener, Hilbert’s Portrait via his Space-Filling

Curve. Proceedings of Bridges 2020: Mathematics, Music, Art, Architecture,
Culture (2020), to appear.
89
This series of models gives an example of sphere eversion, which is the
problem of turning a sphere inside-out smoothly, without creating any cusps
or creases along the way.

After designing a movie with Jos Leys showing an original way of turning the
sphere inside-out, I decided to 3D print a selection of frames of this animation
(bottom). Unlike most physical models of self-intersecting surfaces, I decided
not to use cut-out windows or to print a wireframe, but to have a surface
that is completely closed. To see what is inside, the viewer would open the
object, which is held together with magnets (top). These objects have been
designed and parameterized layer by layer, in a tomographic way, so it was
natural to slice them horizontally. I ordered a print on Shapeways, glued in
the magnets, hand-painted it, and then varnished it to protect the paint.

The 3D printing technology I chose was selective laser sintering on nylon ARNAUD CHÉRITAT
powder. This material is resistant, thanks to a strong bond between the Institut de Mathématiques
grains, and has some degree of flexibility. The thickness of the surface is de Toulouse
around 1.5 mm, chosen to balance durability and price. The overall size is
a balance between price, ease of handling, and the amount of space needed, painted 3D-printed nylon
given a thickness, for the high-curvature parts to render correctly. with magnets

Hand painting is pleasant but tedious, as there are three colors (one for the
outside, one for the inside, one for the cuts), and the boundaries between two
colors require a lot of precision. This certainly was the most time-consuming
part, and was done in my leisure time in the summer break.

I chose the magnets to be as small as possible while still holding the objects
together. After my institution purchased a set of these objects, I created a
second series, where the shape looks better, and with a better selection of
key moments. This time, I used bigger magnets, because while handling the
first set, they would often come apart spontaneously. On the other hand, the
material and varnish did not need adjustment: the models often fell on the
floor when we used them for outreach, and they barely got damaged.

Further information: Arnaud Chéritat, Yet another sphere eversion, preprint (2014).
John M. Sullivan, Sphere eversions: From Smale through “The Optiverse,” Bridges:
Mathematical Connections in Art, Music, and Science, pp. 265–274 (1999).
91
This is a 3D-printed object that illustrates card shuffling. I have previously
made 2D visualizations of card shuffling and permutations, so extending this
to 3D was a natural continuation.

This particular object illustrates two perfect out-shuffles with thirty-six

cards and six piles. A perfect shuffle with two piles is when a deck of cards
is split exactly in two halves, and the cards are alternated, such that every
other card comes from the same half of the deck. Additionally, an out-shuffle
is when the top card remains the top card. The principle is the same for
six piles.

Each curve, from top to bottom, represents the path of a single card.
Together, these curves represent the paths that all of the cards take from
start to finish. Notice that each curve starts and ends at the same position
along the horizontal bars that hold the curves together. This reflects the ROGER ANTONSEN
fact that the cards are restored to their original order after two perfect University of Oslo
out-shuffles. Also observe that the outermost curves remain in the same
3D-printed plastic
position, which means that the top and bottom cards each remain in the
same position throughout all of the shuffling.

While 2D visualizations of shuffling have unavoidable curve intersections,

the third dimension provide the opportunity to avoid these crossings as
much as possible. This object is a compromise between avoiding intersections
between the curves, and making the final object look nice.

I was originally inspired to create these visualizations after attending a talk

by Perci Diaconis. I wanted to know if I could better understand the math-
ematics of card shuffling by aesthetically exploring the various permutations
underlying the shuffling methods. My motivation was to make these invisible
structures visible, and to create elegant and interesting art in the process.

In this case, the curves are calculated with Processing, imported to Rhino,
and produced using Shapeways.

Further information:
Roger Antonsen, Card Shuffling Visualizations, Proceedings of Bridges 2018:
Mathematics, Art, Music, Architecture, Education, Culture, pp. 451–454 (2018).
Eight Ways to Shuffle Sixty-Four Cards: https://rantonse.no/en/art/2018-07-25
93
The first picture shows a 3D-printed press for folding a simple Yoshizawa
butterfly model, and the second shows a press for constructing an origami
model with interesting geometrical properties known as the Mirua map
fold. We made these presses to enable pre-creasing of the paper for origami
models with challenging folds, and to avoid unnecessary creases in the final
paper models.

For example, the classic method of creating a Miura map fold requires the
folder to make some “universal folds” that bend both ways, to aid in the final
construction of the model. With the 3D-printed press, we can avoid these
universal folds, and produce a stronger model. Another benefit of the Mirua
press is that it allows the folder to create folds at exact angles – in this case,
84 and 96 degrees. We also made a press for the Yoshizawa butterfly origami
model; the classic origami design involves some partial folds and unnecessary
creases, but with the origami press, we can achieve a cleaner and curvier LAURA TAALMAN
final butterfly. James Madison University
3D-printed plastic
The press has two pieces that fit together, with the origami paper in between.
We printed a star in one corner of each piece, to indicate how to orient the and folded paper
two pieces to fit them together correctly. Along each of the lines where we
want to make a crease in the paper, one piece has a raised creasing line, and
the other has a lowered creasing line, to create a “mountain fold” or “valley
fold” in the desired location. 3D printing is a natural fit for these presses,
since a laser cutter would not be able to produce both raised and lowered
creasing lines.

The 3D design for the press went through many iterations so that it could
most effectively produce creases in Washi origami paper without ripping
holes in the paper. The corners where mountain and valley folds intersected
were the most challenging in this respect.

For more information, free downloadable files, and folding instructions, see:
https://mathgrrl.com/hacktastic/2019/07/3d-printed-origami-press-miura-fold/
https://mathgrrl.com/hacktastic/2019/08/origami-press-2-yoshizawa-butterfly/
95
To communicate about mathematics, it is very effective to create physical
objects that people can pick up and touch. Complex trees are mathematical
objects that live in the plane, but if at each branching level we add a certain
height, this gives the object a third dimension, which illustrates some extra
properties that are not visible in the 2D pictures. I 3D printed fractal trees
to make them tangible, to be able to share my work with more people.

I was surprised by the shape of the plane-filling structure of the “canopy”

part of the tree that touches the surface of the 3D printing bed (lower
left picture). It is a Rauzy fractal, an object that comes up in many areas
of mathematics (see Pat Hooper’s pictures, referenced below). The Rauzy
fractal is made of smaller copies of itself, and it tiles the plane, and so this
3D fractal tree is also a modular design that can be used to tile the plane.
This example illustrates how the Rauzy fractal can arise as the limit set of an
iterated function system of two linear maps, f1(z) = 1 + cz and f2(z) = 1 – cz. BERNAT ESPIGULÉ
Wolfram Research
3D printing with CURA software is an art. This object pictured here is a
3D-printed plastic
high-quality print, created at ICERM, with an Ultimaker 3 printer. It is the
result of a sequence of failures and successes. This tree was printed upside-
down with a thin raft (an solid base layer of plastic) one millimeter thick.
Applying Magigoo (an adhesive for 3D printing) to the build plate before
starting the print was crucial to force the raft and the tree’s canopy to stick.
Magigoo made the final step of removing this delicate print effortless, since
its adhesive properties are designed to release the 3D print once the build
plate has cooled.

Further information: Bernat Espigule, Families of connected self-similar sets

generated by complex trees, preprint (2019).
360-degree video of object: https://photos.app.goo.gl/kkjqAM2Q6XZyyUEw5
Pat Hooper’s artwork showing iterates of the Rauzy fractal:
http://wphooper.com/gallery/2019/arnoux_yoccoz.png
97
Each horizontal slice of this 3D-printed bottle is the Julia set of some quadratic
polynomial. Let me explain what this means: In the field of dynamics, one
seeks to explain how systems evolve, and in particular to predict their
long-term behavior. For example, when we take a complex number z and
repeatedly apply a polynomial z2 + c to it, where will this take us? For the
vast majority of points, one can answer this question with certainty. On
the rest of the points, which comprise the Julia set, the question cannot
be answered in the same way, because the behavior is “chaotic,” just as the
weather is inconveniently unpredictable over long periods of time.

In the 1980s, Julia sets were appropriated as a symbol of the inherently

chaotic nature of even the simplest systems. Tiny differences in the choice of
the initial complex number z yield widely different outcomes, after applying
the polynomial often enough. And so, though it is the ethereal frontier of
a pair of regions where the future of z is perfectly stable and deterministic, GREG MCSHANE
long-term predictions on the Julia set are quite impossible. Université Grenoble Alpes
3D-printed plastic
Benoît Mandelbrot, who was working for IBM, had access to sophisticated
computers, which allowed him to explore the world of Julia sets of quadratic
polynomials. In his book, he illustrates the way that the Julia set of the
complex polynomial z2 + c changes with the parameter c, by stacking Julia
sets in three dimensions. Inspired by this, we have imitated what he did to
obtain our Mandelbrottle, stacking 150 Julia sets from a short trip in the
complex plane. Starting at zero, we walk around a small circle completely
contained in the Mandelbrot set. We return to zero, and pause a short
moment there before taking another walk around a slightly smaller circle,
and again take another rest at zero. Finally, we hop to another point and
walk quickly straight back to zero to make the bottle cap. Our path has to
be carefully chosen to obtain just the right profile for a swirling Coke bottle.

Appropriately, the design of the Coca-Cola bottle dates to about the same
time as the work of the French mathematician Gaston Julia a hundred years
ago. In Andy Warhol’s studio, it was transformed into part of the vocabu-
lary of Pop Art. Through the work of Mandelbrot and others, Julia sets
became an icon of what one is tempted to call “math consumption.” Repre-
senting unavoidable uncertainties and abrupt bifurcations in the apparently
banal, Julia sets have appeared on teenage bedroom posters, T-shirts, album
covers, even on iTunes.

Benoît Mandelbrot, The fractal geometry of nature, W.H. Freeman (1977).

99
In many different disciplines, we study how families of mathematical objects
change, when one of their parameters is changed a little bit. If these objects
are all 2D, one can vary the parameter over time, and display how the object
changes as an animation. Alternatively, with a 3D printer, we can replace
the time dimension with the z-axis, stacking each successively deformed layer
atop the last. This allows us to hold the entire parametrized family of objects
in our hands, giving an entire new dimension to the deformation (literally!).
I like to call these “static animations.”

My first experiment with static animation was studying the deformation

of the Pythagorean tree fractal (bottom). It is constructed by attaching the
long edge of a right triangle to the side of a square, then attaching squares
to the legs of the right triangle, and repeating this process forever. A right
triangle (up to scaling) is determined by one acute angle, and a small shift in
that angle corresponds to a small change in the triangle, but the movement GABRIEL DORFSMAN-
reverberates more and more as you progress deeper into the fractal. In the HOPKINS
static animation, I vary the angle from 45° to 60° slowly, over the course of ICERM
more than 300 layers.
3D-printed plastic
A second experiment is with deformations of Julia sets (top), following
a workflow developed jointly with Bernat Espigulé. Defining Julia sets is
actually quite easy: Let c be a complex number, and consider the function
f(z) = z 2 + c. If I start with a point p in the complex plane, and I iterate
f – that is, I consider p, f(p), f(f(p)), f(f(f(p))), ... and so on – one of two
things happens: either it gets infinitely large and escapes to infinity, or else
at each iteration it stays relatively close to the origin. The points that stay
close to the origin are those in the Julia set Jc .

It turns out that if you change the constant c just a little bit, the set Jc only
changes a little bit as well, but it can change in quite interesting ways. Given
a path through the complex numbers, we stack the Julia sets corresponding
to the points in the path: If we are at a point c in the path, we print the layer
Jc . Then if c’ is a point just a little bit further down the path, we print Jc ‘
as the next layer. In this way, we turn a path through the complex numbers
C into a 3D-printed deformation of Julia sets. One can view the combination
of Mandlebrot and Julia sets as a sort of fractal subset of C x C (which is
4D), and these prints are 3D slices of this mysterious and beautiful 4D space.

101
The “Perko Pair” is a famous pair of knots that for a long time were thought
to be different, but later were famously revealed to be the same by Kenneth
Perko in 1973. When these knots originally appeared in Rolfsen’s Knot
Table, they were known as 10161 (now “Perko A”) and 10162 (“Perko B”). The
knot we printed is a morph about halfway between Perko A and Perko B,
with spikes just because we felt like it, and because we could.

The large size of this model made it too expensive to print with a service,
so we had to 3D print it ourselves on a desktop FDM 3D printer. In this
size, this model only prints well with either dissolvable supports or so-called
“breakaway” supports. Water-soluable dissolvable supports are an option,
but they can take days to dissolve, and create a lot of goopy mess. Breakaway
supports are really difficult to remove, but they break off with an amazingly
good finish on the final model, as you can see in the upper right image.
LAURA TAALMAN
Being able to examine this knot from all directions in physical space gave us James Madison University
a much better understanding of this knot, as well as its well-known “A” and
3D-printed plastic
“B” conformations.

We went through many failed prints of this model before getting the stars to
align on this six-day print job for a successful result. Then a year later our
cat knocked it off the shelf and we had to do it all over again!

Perko’s 1973 paper: Kenneth A. Perko Jr., On the classification of knots.

Proceedings of the American Mathematical Society 45 (1974), 262—266.

For more information and downloadable files:

Perko Knot Reprint: Dissolvable vs. Breakaway Supports:
https://mathgrrl.com/hacktastic/2018/12/8132/ and
Giant Spiky Perko Knot:
https://mathgrrl.com/hacktastic/2017/05/giant-spiky-perko-k

103
These objects solve the following minimization problem:

Among all polyhedra with given number of vertices and given volume,
find the one with the least surface area.

We call them Akiyama polyhedra, after Shigeki Akiyama, who discovered

them. From left to right, bottom to top, these are the Akiyama polyhedra
with 3, 4, 5, ..., 12 vertices. All of the polyhedra in the picture have the same
volume.

I do a lot of mathematical outreach for children, and I made these to bring to

math festivals. When you can hold an object in your hand, you get a better
understanding of it. For example, in some cases, the symmetry of the shape
is obvious (the tetrahedron and icosahedron) but several of them are not at
all obvious, so it helps to be able to look at it from all sides to find its axis ALBA MARINA MÁLAGA
or plane of symmetry. SABOGAL
ICERM
When I 3D printed these, I printed them with 100% infill, so that they
have not only the same volume, but the same weight. Visually, some of the 3D-printed plastic
polyhedra look a lot bigger than others, so making them the same weight
emphasizes that they all have the same volume. This was easier to do using
3D printing technology than with any other method.

For downloadable printing files: https://imaginary.org/hands-on/minimizing-polyhedra

The original paper: Shigeki Akiyama, Minimum polyhedron with n vertices, preprint (2017).
105
This object illustrates a small piece of an infinite periodic surface, originally
studied by Dami Lee. It’s part of a multimedia piece of art called “Example/
Exception.” The surface can be made by gluing octahedra of two colors (here,
grey and white), where each white octahedron is glued to 4 greys, in such a
way that no two grays share an edge; and each grey is glued to two whites,
on opposite faces. The process of building the surface out of polyhedra can
be continued forever, to obtain an infinite surface that has three directions of
periodicity. Hence, it is called a triply periodic polyhedral surface.

When you quotient by the three directions of periodicity, you get a genus-3
Riemann surface, where you can explicitly write down the automorphism
group, hyperbolic structure, algebraic equation (it is defined over Q), and
several distinct flat structures. This work made up much of Lee’s PhD thesis.

The combination of ease of printing octahedra with a 3D printer and the JAYADEV ATHREYA
ability to glue them made it feasible to make this at an appropriate scale. University of Washington
The ability to print in multiple colors also made it possible to show off the
3D-printed plastic
two distinct families of octahedra.

In making the physical object, I learned that the number of octahedra in a

large ball grows cubically in the size of the ball, and I started to think about
what it would be like to be an ant living on this surface. It leads to thinking
about the geometry of cubic differentials. I also discovered that glue is hard
to work with, and that sometimes, something bigger is more fun. I have
subsequently made a much bigger version out of Rubiks Octahedra.

107
This object shows the five platonic solids. We have chosen to nest them like
a Russian nesting doll. We specifically directed this object at getting children
excited about the Platonic solids, and the fact that there are only five of
them. In addition, the method of nesting one in another is based on a pairing
(or duality) between them, which gives rise to a natural way to embed one
in another.

The five Platonic solids are classical, and I probably learned of them in high
school, but I did not know the fact that they come in pairs with a natural
nesting of one into the other until graduate school, and I did not decide to
illustrate them until now. This type of thing is ideally suited for 3D printing,
in that it involves multiple interlocking 3D objects.

This project is joint work with Jack Love at George Mason University. On
our first attempt, we did not put any padding between the outside and inside EVELYN SANDER
of the shapes, and they each fell into pieces upon printing. On our second George Mason University
attempt, we did not leave any extra space between the objects, and we could
3D-printed plastic
not close each piece to fit around the inside piece. (This 3D printing issue
would not be an issue for a laser-cut object.) On our third attempt, we
realized that there was no way to hold it up, and made a customized stand
to hold it.

For a video of the nesting in action:

https://gmumathmaker.blogspot.com/2019/11/nested-platonic-solids.html
109
These are fundamental building blocks for knotted and braided surfaces
in 4D space. My goal was to create standard-sized models that could be
snapped together in a LEGO-like fashion, so that it would be possible to
make a variety of surface braids, as children’s toys. I made these prototypes
on a 3D printer.

In the 3D-printing world, I learned to cope with unifying normal vectors.

Unsurprisingly, I learned that, in general, curves and surfaces are fictions.
My goal is to depict surfaces as they appear when projected into 3D space
from 4D space, but to do so, the surfaces that I draw must be thickened to
create 3D models that can be sent to the printer. I had to make some effort
to accommodate the transition from 2D graphics to 3D graphics. Splines and
handles of curves require more care on a 3D palette that is projected to a
computer screen. On the other hand, some needed operations became easier.
As I move forward, I will have to change my paradigm from drawing surfaces J. SCOTT CARTER
to sculpting them, as I develop facility with Blender. University of South Alabama
3D-printed plastic

Further information:
J. Scott Carter and Seiichi Kamada, How to fold a manifold. New ideas in low-
dimensional topology: Series on knots and everything, World Scientific Publishing
(2015), pp. 31–77.
J. Scott Carter and Seiichi Kamada, 3D braids and their descriptions. Topology and
its Applications 196 (2015), pp. 510–521.
111
These plastic sheets illustrate the graph of a cubic function of two variables,
as an instructional tool for multivariable calculus. Concepts in multivariable
calculus such as gradient, partial derivatives, chain rule, and optimization
can be visualized in terms of the geometry of the graph of such a function, so
I wanted to give my students a tool for engaging with these ideas in a tactile
and visual way.

I needed to make fi fty copies of the model so that every group of students
could interact with the model. Instead of 3D printing fi fty copies of the
model, which would be expensive and time-consuming, I 3D printed one
large model, and then thermoformed PETG plastic sheets onto the model to
create many copies. Unlike most 3D-printed plastic, PETG plastic is clear,
and works well with dry-erase markers.

In choosing the function, I learned some of the constraints in fi nding a cubic STEPAN PAUL
polynomial with just the right set of properties that I wanted to illustrate. Harvard University
thermoformed PETG plastic
I had to experiment with the kind of plastics used for the 3D-printed mold
and the thermoplastic sheets used for the surfaces, and also with the design
of the mold. The fi rst mold I made tore the thermoplastic along the edges and
corners, and got stuck inside the cooling plastic. I had to redesign the mold
with rounded edges and with a 2-degree draft along the sides, and I switched
from polystyrene thermoplastic sheets (think single-use hot beverage lids) to
PETG sheets (think soda bottles). Also I discovered that typical 3D printing
plastics like PLA melt when they come into contact with the heated ther-
moplastic sheets; Alba Málaga and I destroyed a large model that had taken
hours for Alba to 3D print, when it melted on contact with the hot plastic
sheets. I then switched to 3D printing with ABS, the plastic that LEGOs are
made of, which has a sufficiently high melting point.

113
Regular pentagons do not tessellate (tile) the plane, but it turns out that
there are 15 families of non-regular pentagons that do. These pictures show
3D-printed copies of a representative of the “Type 5” family of tessellating
pentagons (see tessellation in upper left), discovered by Karl Reinhardt in
1918. The design for these models comes from our open source “Pentomizer”
OpenSCAD code, which can produce all of the representatives of each of
the 15 families of tessellating pentagons. The code we wrote for this model
allows us to vary the parameters, and visually explore the entire space of
each tiling pentagon family.

To print the pictured models, we used a 3D-printing service with a profes-

sional selective laser sintering (SLS) 3D printer. Rather than extruding a
thin stream of melted plastic layer by layer like a desktop 3D printer, the
SLS method starts with a 3D volume full of plastic powder, and a laser
fuses the plastic in the regions required by the 3D object. Afterwards, each LAURA TAALMAN
resulting 3D printed object must be sifted out of the remaining powder. To James Madison University
minimize per-part costs, we designed a thin loop into the model so we could
3D-printed plastic
connect the pieces together as one unit (upper right and lower pictures).

The models print well on any desktop 3D printer, or can be laser cut. 3D
printing is not necessarily the best medium for these flat models; it is easier
and faster to laser cut this design if all you want is a handful of pentagons.
However, when printed in plastic on a desktop printer, the individual
pentagons are slightly flexible and particularly nice. When printed on an
SLS printer, the pentagons have a very professional and finished look.

Fun story: One time we printed a full gallon Ziploc bag of one of these
pentagons, for an event. Only after printing them all did we realize that our
pentagons were degenerate representatives of the family we were printing,
and in fact had only four sides. :)

For additional design tips and downloadable files:

https://mathgrrl.com/hacktastic/2019/07/stacking-3d-models-for-bulk-printing/
https://mathgrrl.com/hacktastic/2019/07/print-all-the-pentagons/
115
These are two so-called cubic surfaces: the points satisfy a certain equation
in x, y, z of degree three. These are some of the most fascinating objects in
the intersection of algebra and geometry, from the 19th century.

I have been creating many different 3D-printed versions of such objects over
the past 20 years. Classical sculptures show them in plaster, and my modern
versions, which are exhibited in museums, are usually 3D printed in white
plastics. But these objects in gold-plated brass visualize their fascinating
geometry in a much more obvious way, mainly because of the reflections of
the material. In particular, in some interesting light situations with lighter
and darker colors around, these reflections yield interesting visual effects on
the surfaces of the sculpture. The curvature of these surfaces is much more
interesting than you might think when just computing it, and printing them
in brass allows us to observe these features.
OLIVER LABS
Over the years, I have been creating many different versions of cubic surface MO-Labs
sculptures using 3D printing, until fi nally, now, I like them. Earlier versions
3D-printed gold-plated brass
I created consisted of the part of each surface lying inside a certain sphere.
This is a natural way to cut a fi nite part of an infi nite object – at least
when one does not know much about the object, and wants to focus a region
around the origin. But I worked with these objects for a long time, and
thought a lot about which representation in space I prefer. My fi nal choice is
close to the classical one: cutting by a cylinder. But the decision about the
position and size of the cylinder is not always an obvious one, and making
this decision has been quite time-consuming for some of the cubic surfaces,
as has the choice of the exact equations.

For more sculptures, and to purchase them: https://math-sculpture.com/

Further information: Oliver Labs, Straight Lines on Models of Curved Surfaces,
The Mathematical Intelligencer, Volume 39, (2017), pp. 15–26.
117
The 2-adic solenoid is quite the elusive pop star, featured on the covers of
many dynamical systems textbooks but rarely seen as a tangible object in
our world. Fortunately, this piece can be held in the palm of your hand! It
consists of three solid tori 3D printed in nickel, representing the geometric
construction of the 2-adic solenoid. The second torus is stretched and
wrapped inside the fi rst, and the third is stretched, twice, and wrapped inside
the second. The process can be repeated ad infi nitum. The intersection of all
transformed solid tori gives the solenoid. Taking a look at the cross-section of
the piece shows a system of intersecting disks, where each disk contains two
smaller disks. A point in the cross-section of the 2-adic solenoid corresponds
to choosing one of two disks at each step, ad infi nitum. Hence the cross-
section must be a Cantor space.

I chose 3D printing because I wanted the model to be as precise as possible.

I chose nickel because I wanted the object to act as a paperweight, sitting DINA BURIC
heavy atop some papers or in a person’s hand. University of Victoria
3D-printed nickel
When making the piece, I knew very little about the actual mathematical
object, but the making of the tangible piece spurred a lot of research and
discovery. I created several talks explaining the different constructions,
which I labeled as geometric, topological and algebraic. Through this research
I learned that the solenoid arises in several mathematical fields, and takes
on many forms that seem different but are ultimately equivalent dynamical
systems.

Creating the design in Mathematica did not take too long, once Ian Putnam
helped me with the surface formula. Printing a rough sketch model at the
University of Victoria’s Design Scholarship Commons also happened fairly
quickly. But making the model printable in metal through Shapeways was
a process of trial and error over the course of many frustrating months.
Rendering the model in Mathematica and uploading to Shapeways often took
several minutes, with failure being the fi nal result. Ultimately, tinkering and
persistence won Shapeways over and my design was printed!

Further information:
http://www.math.uvic.ca/faculty/putnam/ln/ Smale_spaces.pdf.
R. Clark Robinson, An Introduction to Dynamical Systems, Pearson Prentice Hall (2004)
Files for printing: http://www.math.uvic.ca/~buricd/projects.html.
119
3D hyperbolic space H 3 allows for lots of interesting group actions, transfor-
mations that return a space to itself. These give rise to periodic tessellations
(tilings) of that space, in which one has a finite number of building blocks
that are repeated over and over again, with each one occurring infinitely
many times. For a 2D analogue in our standard Euclidean geometry, think of
an infinitely extended bathroom wall with a periodic pattern of tiles, or the
work of M.C. Escher (see Laser cutting: Steve Trettel).

Many of those 3D hyperbolic tessellations arise from number theory. For each
natural number d, we can form a Bianchi group Gd of 2 x 2 matrices. Specifi-
cally, the entries of Gd are of the form a + b √-d with integers a, b. Each of
these matrices maps hyperbolic 3-space onto itself. Since this group action
has a special property (it is discrete), it provides a periodic tessellation as
advertised, with finitely many basic tiles as its building blocks. Imposing a
further technical condition results in a tessellation for which the vertices of HERBERT GANGL
all the tiles lie at the boundary of H 3 (it is an ideal tessellation), and thereby Durham University
ensures a close connection to our standard 3D Euclidean geometry. Each of
3D-printed precious metals
our polyhedra displayed in the pictures is associated to some polyhedron
in an ideal tessellation that results from the action of a group Gd as above.

The choice of tessellation arose from my diploma thesis. Don Zagier gave
me the topic, and Walter Neumann crucially suggested that I use work of
David Epstein and Robert Penner. I computed the first dozens of cases on
an Atari ST+ in Omikron Basic (in the late 80s). In recent years, the topic
became a nice playground for summer undergraduate research projects in
Durham. With the advent of affordable 3D printers, an obvious task was
to find ways to visualize—and materialize—the objects. This became easier
after my student Josh Inoue discovered a simple way to recover an even
more symmetric version of each such polyhedron having all its vertices on a
sphere (via a suitable affine transformation, dependent on d). For computing
and displaying them, we invoked software like Pari/GP, Mathematica and
OpenSCAD.

Eventually I was tempted to get the objects materialized in precious metals

through a 3D-printing service, a process that turned out to necessitate many
iterations. I was so amazed by the beauty of the outcome that I decided to
create a whole collection of mathematical jewelry.

121
The Klein bottle is one of the simplest examples of a fiber bundle: a space
built by putting copies of one shape (the fiber) at each point of another space
(the base space) in such a way that adjacent copies glue together nicely to
form the total space. The usual embedding of the Klein bottle into 3-space
illustrates how to construct the Klein bottle: glue one pair of sides to form
a tube, and then glue the other pair with a flip. In contrast, the shape of
the Klein bottle that I chose here highlights the fiber bundle structure itself,
using a wireframe so that we can see and follow the base space around the
figure, and we can see the circular structure of selected fibers as well.

3D printing was a natural choice, since the construction I envisioned was

relatively straightforward to implement algorithmically. Where a hand-
constructed or sculpted approach would have suffered from variations across
the piece, a 3D-printed approach gave it the uniformity I sought.
MIKAEL VEJDEMO-
I printed my first attempt (lower right) on an extrusion printer. A combina- JOHANSSON
tion of extreme overhangs and issues with temperature control failed that CUNY College of Staten Island /
print. Based on that failure, I switched to sintered-type printing methods,
CUNY Graduate Center
and sent my models to Shapeways rather than trying to print them myself.
With sintering, the overhangs are supported by the unprinted materials, 3D-printed plastic, alumide
and cease to be a problem. The first instance of that, seen here in grey, was and steel
printed in alumide. I also used the features of that printing method to print
a disconnected ball in place inside the Klein bottle.

The modeling for the alumide model had aesthetic issues: along a curve
winding twice around the Klein bottle, several tubes in the model bundle up
together. Furthermore, the fibers intersect themselves, somewhat obscuring
the fact that they are all topological circles. Revisiting the modeling script, I
created a new version that includes the base space as one of the longitudinal
wireframe wires, with the fibers dodging out of each other’s way. This way,
the circularity of the fibers is more clearly visible, and the self-intersection
of the Klein bottle along the base space is visible in the alternating intersec-
tions of fibers with the base space circle.

Further information: Mikael Vejdemo-Johansson, 3D printing identification spaces

of the square, preprint (2019).
123
This piece is an illustration of elementary cellular automata, made from
3D-printed ceramic. A cellular automaton consists of a grid of cells, where
the state of each cell depends on the states of those around them. Simple
examples of cellular automata are Pascal’s triangle, which has rows of
numbers where each one is the sum of the two above it, and John Conway’s
Game of Life, where each square in a grid is colored white or black at each
time step, depending on the colors of the cells around it in the previous
time step.

I’m not a mathematician, so I learned everything about cellular automata

through this project. Once I was doing the work and the research, I started
to be even more interested in exploring complex systems that are based on
iterating simple rules. I think about these rule-based mathematical systems
as games that can be played for an infinitely long time.
TIMEA TIHANYI
My workflow from the math to the object can be summarized by the following Slip Rabbit Studio
steps: After picking the game rules, we coded them in Sage, using CoCalc
3D-printed ceramic
(see link below). I worked on coding this with mathematician Sara Billey
from the University of Washington, then altered Billey’s code-kernel with the
help of Daria Micovic. The code generated a 2D plot of a distribution matrix
in black and white. We then used a 3D modeling program, Rhino, to turn
this matrix into printable 3D geometries of cubes and cylinders.

This process required careful calculations and many tests, which were
reverse engineered from the relationship between the printing layer height,
nozzle size and cell size, and the desired scale of the finished object.
A precise reflection of the original matrix was my main goal, reflecting on the
algorithmic nature of the first half of the process. On the other hand, I enjoy
seeing how physical reality – the qualities of the material and the physics
of gravity, time, etc. – make subtle unexpected changes on the otherwise
perfect model.

I work by testing and constantly refining until I have a very solid prototype
that can be 3D printed reliably. Still, I might lose 50% of the pieces because
what I’m asking from the fragile porcelain is kind of difficult already.

More of my work: https://www.timeatihanyi.com/

Information about 3D printed sculpture: https://www.sliprabbit.org/studio
CoCalc is a web-based platform for computational math, founded by mathematician
William Stein: http://cocalc.com/
125
I am interested in curved 3D spaces. We’re used to flat space, where two
straight lines that are parallel in one location remain parallel forever. In
curved spaces, two straight lines that start out parallel will intersect. On
the curved surface of the earth, imagine two people on the equator, one in
Ecuador and one in Kenya. They start walking due north at the same pace.
Initially, their paths are parallel, but they will get closer to one another, and
eventually their paths meet at the north pole.

In 3D, there is a space called the 3-sphere that, like the surface of the
familiar 2D sphere, has uniform positive curvature. There are many beautiful
geometric structures that live naturally in this space. The three sets shown
here are a pair of interlocking Möbius strips (bottom left), a twisted Scherk
saddle tower minimal surface (bottom right), and a woven pair of Hopf fibra-
tions on the Clifford torus (top left). Both the Möbius strips and the Scherk
saddle tower started as objects in 3D Euclidean space with finite extent SABETTA MATSUMOTO
in the z-direction. I then placed them along an equator of the 3-sphere, Georgia Institute of Technology
which closed them up into a loop. For the third piece, I took the Hopf
metals cast from 3D printed molds
fibration – which is a series of circles that are all interlinked with each
other – and looked at the circles that live on the Clifford torus. I took two
of these Hopf fibrations that are orthogonal to one another and wove them
together along the Clifford torus. I then projected all of these objects to 3D
Euclidean space.

I use the cast metal process used by Shapeways because it accurately repro-
duces geometry and creates professionally finished pieces. These pieces are
3D printed in wax, and then cast into precious and semi-precious metals.
The choice of materials depends on the type of geometry. For instance, iif
there are multiple interlocking pieces, the piece cannot be cast in or plated
with precious metals. My experience with interlocking pieces is that they
turn out beautifully, but they require more upkeep to keep them looking new
than ones that are made from or plated in gold.

I wanted to bring many of these objects to life in a medium that comple-

ments their natural aesthetic simplicity. I chose jewelry for a few reasons:
it is simple and elegant while still able to show mathematical details. I love
the notion of creating something elegant and feminine out of mathematics.
Those of us who are mathematicians see the beauty in everything we do, but
it is hard to communicate that outside the community. I believe jewelry is a
way to universally share the beauty we see with others.
127
MECHANICAL
CONSTRUCTIONS &
OTHER MATERIALS
Linkages, tensegrities, and other mathematical constructions
are plenty interesting to study in theory, but it’s far better to
construct them, hold them in your hands, and see how they
move. As you will see, you can make an illustrative mathemat-
ical object out of just about anything.

128 129
This is a hyperboloid, cut from a brick.

This item was the result of a project with Nick Bruscia and Dan Vrana from
University at Buffalo and OMAX corporation, who make the water jet. The
challenge was simply to explore what was possible with a 5-axis water jet.
As the cutting “tool” is a straight (well, spiraling) jet of water, ruled surfaces
are the natural thing to start experimenting with.

A surface is ruled if, for every point on the surface, there is a line through
that point that is completely contained in the surface. A hyperboloid is a
ruled surface, so to cut it out, we need the jet of water to follow the ruling
lines. This object therefore starts as an illustration of an application of
geometry: how to control and think about the forms that can be made, by
controlling the form of a general tool.
EDMUND HARRISS
It also presents a compelling version of a classic mathematical object. The University of Arkansas
surprising material (brick) draws you in, with the question of how it could
Waterjet-cut brick
be made. Even knowing the tool used might not help, as the curved surface
seems impossible to make with such a straight tool. This leads directly to an
understanding of the behavior of ruled surfaces, and the hyperboloid.

131
This piece shows rational points on a cubic surface. Rational points are the
points (x, y, z) that lie on the surface, such that x, y and z are all rationals.
This is related to Fermat’s famous last theorem.

Away from the 27 straight lines contained in the surface, there is only a small
finite number of rational points with numerator and denominator below 100
(on the part of the surface shown). To visualize these points in real life, we
needed some support material to hold them at their position, because the
points themselves are not connected – there is an empty space between them.
Laser-in-glass allows us to do this with a support material (glass!) that is
almost invisible. This is perfect for getting a good idea of the fascinating
structure and geometry of these points.

There is some complicated hidden structure in the positions of the points.

Understanding it mathematically is not easy, but visually, one quickly gets OLIVER LABS
a rough feeling for it. MO-Labs
laser-in-glass
Choosing the correct size for the small balls representing the points was a
challenge. Also, choosing a good bound for the numerator and denominator
(100) was a lengthy experimental process. Working with Ulrich Derenthal,
we essentially had to choose a compromise between having enough points to
see certain geometrical features, without the points being too numerous or
the balls being too large. Too many tiny balls almost cause the whole cubic
surface to be filled.

For more sculptures, and to purchase them: https://math-sculpture.com/

133
The Cantor set is one of the simplest fractal objects. Its ternary representa-
tion is obtained by an iterative procedure: start with a segment of length 1
and remove the middle third; for each of the two remaining segments, remove
the middle third; and so on. After an infinite number of steps, the remaining
object is the Cantor set, and looks like dust. For a long time, topologists
were intrigued by the embeddings of a Cantor set in 3D Euclidean space. In
particular, they asked:

Is it possible to find an embedding of the Cantor set in R3 so that there

exists a loop that cannot be shrunk to a point without crossing the Cantor
set? or, said differently: Does there exist an embedding of the Cantor set
in R3 whose complement is not simply connected?

The ternary Cantor set described above does not satisfy this property.
Nevertheless, the question was answered positively by Louis Antoine at the RÉMI COULON
beginning of the 20th century. This embedding is now known as Antoine’s CNRS / Université de Rennes 1
necklace. It is a necklace, whose links are themselves a necklaces, whose links
Brass links
are themselves necklaces, whose links...

This piece represents the first four levels of the Antoine’s necklace. Louis
Antoine was a professor at the Université de Rennes 1, where I currently
work. When my colleagues discovered that I was interested in illustrating
mathematics, they challenged me to realize an Antoine’s necklace.

Because the links of the necklace are so intertwined, it was not possible
create this object with a 3D printer or a laser cutter. Inspired by jewelry, I
decided to build the object from brass chain. This piece is made of 10,000
links, which I opened and closed individually. The whole process let me
experiment with the self-similar structure of fractals. In particular, at the
end, a single link is simultaneously used to close one copy of each level of the
necklace. The first versions were unicolor. However, it was not easy to distin-
guish the smallest levels. Using two different platings solves this problem.
The next challenge is to build a stand to display the necklace.

Antoine’s original paper:

Louis Antoine, Sur l’homéomorphisme de deux figures et leurs voisinages.
Journal de Mathématiques Pures et Appliquées 4 (1921), pp. 221–325.
135
This assembly provides a large-scale model of Weaire-Phelan foam, which is
a candidate answer to the question:

How can we enclose and separate infinitely many unit-volume regions

(“bubbles”) in 3-space, with the minimum possible surface area?

The structure arises because the surface films of bubbles naturally tend to
minimize their surface area. For a single bubble, the minimizing shape is
a sphere, but for many tightly-packed bubbles, it’s still not clear what the
optimal structure should be. I chose to illustrate the Weaire-Phelan foam for
a public address on the nature of mathematics:

• I wanted to bring out the multi-layered nature of mathematical discovery:

each time a pattern is discovered, it may lay the groundwork for deeper,
more-encompassing patterns. By analogy, the audience could discover GLEN WHITNEY
that their “line segments” – wooden dowels distributed prior to the talk – Studio Infi nity
could fit together into pentagons and hexagons, that the pentagons and
acrylic plastic sheets, dowels,
hexagons could fit together into polyhedra, and that those polyhedra fit
together perfectly into a potentially infinite lattice. 3D-printed brackets, and metal
• This lattice acts as a gateway to an unsolved problem – is the Weaire- hardware
Phelan foam really the optimal shape for packed bubbles? – revealing that
in mathematics, there are always more patterns to be discovered.
• I wanted the audience to have a satisfying tactile experience assembling
the model. There’s nothing like touching and manipulating material
objects to form a cognitive connection with a subject matter. Also, the
scale of the model was important: I wanted the finished product to be
visible throughout the auditorium, and provide the audience with a sense
of accomplishment.

My plan was that faces of adjacent cells joining together in the final lattice
would have the same color, whereas adjacent faces of the same cell would
have different colors. Unfortunately, I discovered that it’s mathematically
impossible to accomplish this goal with only four colors, and in fact it may
not even be possible with five! The biggest problems I encountered were
about stretching and bending: First, the acrylic sheets don’t stretch and
bend exactly like the surface of a bubble would. Second, the dowels forming
the edges of the faces do bend, but my model for the dowels’ lengths did not
take the bending into account, so I cut the dowels to approximate lengths. I
made the holes in the acrylic sheets extra-large to compensate, but they still
only fit with difficulty, and some cracking of the acrylic occurred. 137
A traditional kaleidoscope is made from three long thin rectangular mirrors,
faced inward to form a long thin triangular prism inside a cylindrical tube.
Small colorful objects in various shapes are placed in the far end of the tube,
and the many reflections of the colorful objects in the mirrors creates an
image that looks like an infinitely patterned plane. In such a setup, we would
say that the fundamental domain of the kaleidoscope is a triangular prism,
and the resulting image has triangular symmetries in the flat plane.

Rather than using rectangular mirrors in a tube to create a flat reflection, I

used triangular mirrors coming together at a point (top right). This creates a
polyhedral reflection, as shown. The three polyhedral kaleidoscopes pictured
here allow one to create and view objects with icosahedral, octahedral, and
tetrahedral point symmetries.

Here, the fundamental domain of each kaleidoscope is a tetrahedron, and JOHN EDMARK
the angles between its faces determine the symmetry group of the reflection Stanford University
image. Using these mirrored fundamental domains allows one to create the
laser-cut mirrors
Platonic solids, and most of the Archimedean and Catalan solids. I created
the inserts out of card stock (top middle) that fit perfectly into the funda-
mental domains, to create the associated polyhedra, as shown. One can also
create other card stock polyhedra, or use any objects that fit inside, and
place them in the kaleidoscopes to see the resulting form.

I chose to use front-surface acrylic mirror and card stock because they are
laser-cuttable. I made several iterations of the kaleidoscopes, working to
refine the design so that they could easily be fabricated by someone with
little or no crafting abilities (see instructions at link below). I hope they will
be useful tools for others to learn about polyhedral symmetries.

Instructions for creating these:

https://www.instructables.com/id/Easy-to-Make-Polyhedral-Kaleidoscopes/
139
What happens when you attach mirror surfaces to the inside faces of a rhom-
bicosidodecahedron? We set out to answer this question by creating this
polyhedron out of Geometiles and placing a spherical camera near its center,
and then stereographically projecting the output. In this image, the “pole”
of the projection is the face opposite to the one through which the camera
was placed. Careful placement of the camera allows us to see reflections of
pentagons in other pentagons and squares.

With all the freedom that computer-generated images allow, photography

has almost a “retro” appeal. The spherical camera, of course, allows for very
unconventional photography by taking in the entire scene around it. Stereo-
graphically projecting this view onto a 2D plane gives a unique interpreta-
tion of what it’s like to live inside the intriguing mirrored polyhedron.

An unintentional byproduct of this project was seeing what happens when

there are slight deviations from the stereographic projection. When using
a spherical camera inside an almost spherical object, these deviations are
expressed as the lens of the camera being slightly below or above the center
of the object. As a result, the projections look different depending on which
pole you project from.

In order to get the best possible image, we had to experiment with the
position of the camera in the polyhedron, as well with as the ambient light.
The latter is tricky, since light can come into the rhombicosidodecahedron
only through the thin crevices between the tiles, and from the bottom face
where the camera is placed. YANA MOHANTY &
BJOERN MUETZEL
We say “we” even though Bjoern was taking the pictures at Dartmouth
Imathgination LLC;
College, and Yana was processing the raw images in California. At some
point, things got downright comical: Yana was making recommendations on Dartmouth College
where to shoot the pictures based on pictures of rooms on the Dartmouth Photography, and mirrors
campus that she had only seen online. attached to plastic tiles

Further information: https://geometiles.com/archimedean-billiards/

More pictures and videos: https://math.dartmouth.edu/~mutzel/gallery.php
141
This is a large window, about 30 feet wide (bottom image), installed at the
Mathematics Tutoring and Teaching Center at the University of Arkansas.
It shows a self-similar hierarchical pattern in the plane, made by a recursive
tiling substitution rule: all of the arcs are copies of one another, and each
arc sits beneath four smaller arcs, always in the same manner. Conversely,
each arc is one of four sitting above a larger arc, themselves a group of four
upon an even larger one. In the mathematician’s mind, this continues forever,
smaller and smaller, and larger and larger, ad infi nitum, always in a regular
manner.

I made this image in Adobe Illustrator. When drawing a mathematical illus-

tration in such a program, the main task is figuring out HOW to do the
drawing, coming up with a process that will give the desired result. The
mathematical illustrator must be willing to start from scratch several times,
refi ning this process. I drew this window four or five times over, ultimately CHAIM GOODMAN-
leading to a simple method that was accurate to 1/100th inch across a nearly STRAUSS
thirty-foot span. University of Arkansas

The complexity of the details of the drawing was severely restricted by its laminated glass with printed
exponential growth each scale of refi nement – as drawn, the illustration interlayer
nearly overwhelmed my computer!

I work with tiling substitution rules in my research, and the opportunity to

draw this window led me to explore this one. There’s a long literature on
how undecidability and other aspects of the Theory of Computation arise
as matching rules on these hierarchical tilings, and certain aspects of this
particular system seem to point in some new directions for “programming”
in these tilings.

The design was printed in three layers: a sandwich of two layers of colored
ink around a layer of white ink. The central white layer was printed in
various opacities, reflecting light and color in some areas, and translucent in
others, with bright zings of color moving through the window.

143
Each of these prints shows a planar slice of a Z3-periodic function on R3.
The function is negative in the black region, and positive in the white one.
The slicing plane is irrational, so the pattern never quite repeats. Slices like
these show up when physicists study how metals conduct electricity in strong
magnetic fields. Sergei Novikov and his coworkers realized in the 1980s that
to know how a metal conducts, you have to know how the connected compo-
nents of the black and white regions behave. Olga Paris-Romaskevich told
me about Novikov’s problem, and mentioned that she’d love to see computer
renderings of the slices. I’d been wanting to experiment with laser-cut block
prints, so I was keeping an eye out for striking black and white patterns. I
rendered a slice and thought, this is it.

I was drawn to block printing because I love M.C. Escher’s prints, and
digital reinterpretations of his work by folks like Christopher Becker and
Vladimir Bulatov have really caught my fancy. It would be wonderful to AARON FENYES
give those new digital forms all the texture and depth of the originals. When Institut des Hautes Études
I turned a hyperbolic tiling from my research into an art poster, I had it Scientifiques
screen-printed by hand at a local print shop with that hope in mind. The
printers’ lovely work encouraged me to try block printing next. laser- and hand-cut block prints

I got lots of tips from volunteers at AS220, the Providence art space where
I did most of the laser cutting and all of the printing. One was to try
different inks, papers, and block materials. Switching from Akua intaglio
ink to a letterpress ink thinned with soy oil made a big difference, and the
printing papers I sampled from the AS220 scrap drawer each made notice-
ably different impressions.

It took some care to find block materials that seemed safe to laser-cut.
Several brands of printmaking linoleum have “no dangerous decomposi-
tion products” listed on their Safety Data Sheets. However, many synthetic
rubber “easy-cut” blocks contain PVC, which is not laser-safe. I had good
luck with both printmaking linoleum and laser-safe MDF.

During a laser cutter outage, I bought a carving tool and cut the print on the
left by hand. One corner shows a downside of hand-cutting: it takes hours,
not minutes, to make a cut, but still only seconds to make a mistake.

Further information: Artur Avila, Pascal Hubert, and Alexandra Skripchenko,

Diffusion for chaotic plane sections of 3-periodic surfaces. Inventiones
Mathematicae, Tome 206 (2016), pp. 109–146.
145
We can think about prime numbers as being the “atoms” of the integers,
in the sense that every integer is the product of prime numbers, and prime
numbers themselves are indivisible, like the atoms that make up molecules.

A ring is a mathematical object that generalizes the notion of integers: it

is a group of objects such that, for any two elements of the ring, their sum
and product are also elements of the ring. Analogously to the notion of
prime numbers for the integers, prime ideals are in some sense the “atoms”
for rings. In this case, we consider the ring Z[x] of polynomials with integer
coefficients, and its set of prime ideals, denoted by Spec(Z[x]).

In The Red Book of Varieties and Schemes, David Mumford depicts

Spec(Z[x]), and together, Gabriel Dorfsman-Hopkins, Sachi Hashimoto
and I wanted to understand his picture through actually building it. We
considered the medium and technique as a representation of the underlying TASHRIKA SHARMA
abstract mathematics. University of Vienna
Polypropylene and monofilament
Translucent materials tend to have a non-formless quality to them, and have
a nice tension with the shadows created when we project light through them.
Using two types of translucent materials, we could create a tension between
the relationship of the string and the polypropylene.

By creating this object, we learned how to interpret the plane used to frame
the interplay of the prime ideals in Spec(Z[x]). This was the first version, and
can be considered an experiment of perception and information.

David Mumford, The Red Book of Varieties and Schemes, mimeographed notes from
the Harvard Mathematics Department (1967).
147
MULTIPLE WAYS
TO ILLUSTRATE
THE SAME THING
One of the great joys of a career in mathematics is the oppor-
tunity to sit with a problem for a long time. Over the months,
years and decades, our perspective on a mathematical idea may
change several times, so that the way we come to understand
it later is very different from the way we initially learned it.
When we personalize a problem, an object, or an idea in this
way, we own it, and in so doing, we give it new life. After all,
no mathematical idea can be said to “exist” in its own right;
an idea only exists in the minds of those who understand it.

148 149
The Hilbert curve is a continuous fractal space-filling curve first described by
David Hilbert in 1891 (see 3D printing: Judy Holdener).

(top left) These show four iterations of the Hilbert curve, illustrating the
recursive step from one level to the next. Laser cutting opens up an enormous
space of possibilities by its extreme precision. After a couple of minutes of
laser cutting, you have created a piece of wood that would take months or
years to create by hand, if even possible.

(top right) This is an 8 x 8 grid of wooden squares, hand-colored and arranged

into the Hilbert curve. Being able to rearrange them makes the curve and its
properties more interactive, and creates a nice game.

(bottom left) This is a Celtic knot rendition of the Hilbert curve, made with
paper. It is a 192-crossing, 3-component Brunnian link. Each component has ROGER ANTONSEN
its own color, and is equivalent to the unknot. It shares the property with the University of Oslo
Borromean rings that if any one of the strands is removed, the other two are
laser-cut wood, paper and
left unconnected. Because of the overlapping strands, I cut this from three
different pieces of paper and then braided them together. Paper is a fun mirrors
medium, but also very fragile. I made early versions with a mechanical paper
cutter, but I had more luck cutting the paper with a laser cutter.

(bottom right) This is a 52-mirror labyrinth, with a “laser beam” that traces
out the Hilbert curve. There are several different but equivalent ways of
defining such curves: One way is “external” or “plotter-based”, given by
absolute directions: up, down, left and right. Another way is “internal” or
“turtle-based”, given by commands like “turn left” or “move forward” without
any knowledge of absolute direction. The mirror labyrinth illustrates the
internal point of view, as the “laser beam” has no knowledge of absolute
direction, and the mirrors serve as the local commands.

The top layer of the base has slits precisely cut to hold the 52 mirrors so that
their reflective back faces are on the appropriate diagonals. The symbolic
“laser beam” consists of a red piece of paper, cut with a laser cutter, resting
on a wooden platform of the same shape. Originally I wanted to use an
actual laser beam to trace out the curve, but the mirrors absorbed too much
light, and the beam was invisible after a dozen bounces.

Further information: Michael Bader, Space-Filling Curves: An Introduction with

Applications in Scientific Computing, Springer-Verlag Berlin Heidelberg (2013).
151
It has always been evident to geometers that curves and surfaces are different
objects. But in 1890, Giuseppe Peano, in the famous paper Sur une courbe
qui remplit toute une aire plane, showed a very counterintuitive example of a
continuous “space-filling” curve that completely fills the unit square.

The paper of Peano has no figures; when I was a student, the conventional
wisdom was that such a curve was impossible to represent, as the figure
would have been a black square! To get intuition about this curve, the best
way was to draw a finite approximation (top left). However, this does not
give a good intuition of the curve itself, as it is very homogeneous, and makes
it difficult to understand the dynamics of the curve. I had the idea of making
a representation such that the color of each point depends on the parameter
of the curve; in this way, one can see the movement along the curve, by the
the direction of continuous change of color (top right).
PIERRE ARNOUX
A few years later, Xavier Bressaud suggested that it might be clearer to show Université d’Aix-Marseille / CNRS
the 3D graph of the function, and we made some interesting figures showing
graphics, 3D-printed plastic
perspective views of this graph for a paper in Experimental Mathematics.
But this representation was not complete: why not built the real object in
3D? There is an obvious problem: a fractal curve in R3 is too fragile to 3D
print. But laser engraving in glass is a technology perfectly adapted to an
object of this type, and in 2017, with Eltarr Loukman, an energetic under-
graduate student at Aix-Marseille University, we succeeded in creating this
object, which shows how a topological line can have Hausdorff dimension
2, and project onto a square in one direction, but not in all directions
(bottom left).

Upon seeing this curve in glass, I realized that the set of points below the
graph formed a kind of strange hill, with an ascending path passing through
all points of the hill, and that this object could easily be 3D printed, which I
did at ICERM this fall (bottom right). This gives another viewpoint on this
remarkable object. Similar models can be made for any plane-filling curve,
and it will be interesting to make them to get a better intuition about the
properties of these curves, and the differences between them.

153
These show periodic billiard paths on the regular pentagon: You imagine
that you have a pentagon-shaped billiard table, and you shoot the ball so
that it bounces around for a while, and then goes back to where it started
and repeats the same path forever. We studied and classified all of the
periodic billiard paths on the pentagon, of which there are (countably) infi-
nitely many, and we also wrote a program in Sage to draw them.

Over the past couple of years, we have tried multiple ways of illustrating
these paths. The simplest way is just to draw the path itself, as in the upper
left; this is laser-engraved in wood, which allows you to draw a very intricate
path. Incidentally, the laser follows the path of the billiard as it engraves the
path, and it’s interesting to see the pattern emerge in real time.

Next, we thickened up the path and removed the negative space, as in the
upper right; this is 1/8-inch laser-cut acrylic plastic. While this requires a
much simpler trajectory, this method emphasizes the path more. Diana has
made small versions of this into dangly earrings, which are excellent conver-
sation starters about our research.

For the picture in the lower left, you imagine that when the path hits the
table, it changes color, from red to blue or vice-versa. Alternatively, you can
imagine that you have a ribbon that is red on one side and blue on the other,
and you are wrapping it around a transparent pentagon. To do this, Diana
used clear acrylic with paper on both sides, and laser-engraved the even-
numbered pieces of the path on one side and the odd-numbered ones on the
other side. Actually, since the laser cutter beam is very narrow, we engraved
DIANA DAVIS &
ten nearby parallel paths, which removed the protective paper in strips along
SAMUEL LELIÈVRE
the path. At this point, the paper only remained in the negative space. Then
she painted one side red and the other side blue, and removed the painted Swarthmore College,
paper, to leave only the painted path. Removing the paper was much more Université Paris-Saclay
painful and tedious than expected, but the result is beautiful.
laser-cut acrylic and wood, and
Lastly, we noticed that the path partitions the pentagon in a way that is computer-generated graphics
two-colorable, so we colored the regions black and white (bottom right). This
de-emphasizes the path, but raises interesting questions, like whether the
ratio of black to white might approach the golden ratio.

Further information: Diana Davis and Samuel Lelievre, Periodic paths on the
pentagon, double pentagon and golden L, preprint (2019).
155
Surfaces of negative curvature are familiar in everyday life: a surface has
negative curvature at a point if it is saddle-like there, and the more negative
it is, the more extreme this saddle is. A surface with negative curvature is
ruffly, like lettuce or curly kale, and has a tremendous amount of surface
area for the volume it occupies. Though a lot of the mathematics of these
surfaces has been well understood for more than 150 years, there remain
many open, unexplored questions of just how these surfaces actually sit in
space, and the dramatic changes they undergo when they are manipulated.
Partly, this may be because there aren’t many ways to actually build such
a surface – crochet (see paper and fiber arts: Gabriel Dorfsman-Hopkins &
Meghan Maynard; and Maria Trnkova) is one technology – and mathemati-
cians haven’t generally played with many physical examples.

(top) This surface of constant negative curvature is made out of annular

strips connected edge to edge. Eugenio Beltrami made a model like this CHAIM GOODMAN-
one out of paper, more than 150 years ago! But that is an inefficient use of STRAUSS
material. It’s impossible to cut very many annular strips from a flat sheet University of Arkansas
without a great deal of waste. Straight strips are far more efficient, but strips
of flat materials such as steel or paper can only be bent in the direction Ethylene-vinyl acetate (EVA)
perpendicular to the surface, and the strip must remain a geodesic on any foam, steel, paper
surface it sits upon. But foam strips are squishy and can be bent in other
ways, so they are just right for this piece. The pieces for this were cut using
a router on a CNC machine, which creates a lot of dust, so it would be much
better to invest in a CNC knife that cuts cleanly.

(bottom left) This is a spherical piece of the gyroid, an infi nite periodic
minimal surface. The gyroid has one of the more fascinating of the three
dimensional Euclidean discrete symmetries – it is difficult to understand
even when you are looking at it! Steel is solid, durable, and upgrades a
piece. Mathematical illustrators should know: Welding is easy, and basic
metal work is a fast and forgiving medium! (Welding well is another matter.)
Because straight strips of steel do not bend into annular ones, these straight
strips of steel lie along geodesics on the gyroid.

(bottom middle and right) These are surfaces of constant negative curvature
made from flat strips of paper and steel, respectively. The steel object is
about ten times the height of the paper one!

Further information: Daina Taimina, Crocheting Adventures with Hyperbolic Planes:

Tactile Mathematics, Art and Craft for All to Explore, CRC Press (2018).
157
 These objects all depict the three Borromean rings, which have the property
that though no two rings are interlinked, all three cannot be unlinked. I
created these models to study how the embedding of the three rings will
change, when the surface of the handle-body is everted (turned inside-out)
by pulling the whole surface through a small puncture in its skin.

A. The Borromean rings in a solid, unmovable arrangement. Even though

no pair of ovals is interlinked, the whole assembly cannot be taken apart.
B. The Borromean rings made out of pipe cleaners, which allows one to
study how the rings can be rearranged without opening any one of them.
C. The most symmetrical way of embedding the Borromean rings in a genus-3
A B C D handlebody. This graphic shows the three entangled ovals placed onto the
surface of a Tetrus, a fattened and nicely rounded tetrahedral wire-frame.
D. A flexible version of (C), in which you can push a tetrahedron vertex
through the opposite tunnel, changing the order of the circling rings. CARLO SÉQUIN
E. A drawing showing the same embedding on a 3-hole torus, obtained by Univeristy of California, Berkeley
flattening the Tetrus, and emphasizing the 3-fold rotational symmetry.
3D-printed plastic, pipe cleaners,
F. The goal is to make a model that can by physically everted. For this
structure, it is known how to do this. Marker pen traces on a 3D-printed folded paper, computer-generated
model show the embedding equivalent to that of models (C) through (E). graphics, plastic bags
G. It was quite difficult to translate the embedding in Figure (E) onto the
prism model (F). We made this paper model to experiment with twisting
the handles, to give insight as to how many times each loop winds around.
H. This “4-Pillar Pagoda” is an alternative way to re-shape (C), resulting in a
shape that can be everted through a puncture placed at the “North pole.”
E F G H
I. A model of the prism with 3 handles made from cylindrical newspaper
bags. They are pasted together with their top and bottom ends, respec-
tively, to form two 4-way tubular junctions. In the top junction, at the
“North pole,” a star-shaped cut serves as an expandable puncture.
J. The same model after the whole skin has been pulled through the puncture.
The prism and each handle turn into half-length “reversed socks,” ending
in a “mouth” as can be found in the standard depiction of a Klein bottle.
K. A sketch of the everted embedding from (J). Eversion changes the
embedding: originally, each ring passes once through each of the three
holes; after eversion, only two of the three rings visit each tunnel.
L. The alternative embedding resulting from everting model (H), which
breaks the 3-fold symmetry. Four of the six handles carry all three rings,
and the remaining pair of opposite edges each carry just two colors.
I J K L
159
Illustrating Mathematics semester program participants, ICERM, Fall 2019.

160 161
 ACKNOWLEDGMENTS

Thanks to Eko Hironaka for suggesting this project, for recommending

me to be its editor, and for working with the AMS to make it happen.
Thanks to Marcia Almeida for being our AMS liaison. Thanks to
Katie Vokes, Curt McMullen, and Hannah Kushnick for their essential
contributions of time and ideas. Thank you to Edmund Harriss for
providing a professional-level photography setup to photograph people’s
objects, and for taking many of the beautiful pictures in this book.

Many thanks to ICERM for hosting the wonderful Illustrating

Mathematics semester program in Fall 2019, and to David Bachman,
Kelly Delp, David Dumas, Saul Schleimer, Richard Schwartz, Henry
Segerman, Katherine Stange, and Laura Taalman for organizing the
program. Thank you to the IHES for hosting me as an invited research
visitor for three months during the coronavirus pandemic of 2020, during
which I finished this book.

Thank you to everyone who passed through the ICERM semester

program, for your kindness and generosity of spirit, and especially to all
those who contributed work for this book. A special thanks to all of the
contributors for responding to my questions and returning their consent
forms, despite the global pandemic.

Finally, a multitude of thanks to Margo Angelopoulos for editing every

image, typesetting every word, and laying out every page of this book,
and for making the composing of this book a thoroughly enjoyable
collaborative experience.

162 163
PHOTO CREDITS Kepner: © 2011 Margaret Kepner. All rights reserved.
Zucker: © 2020 Matthew Zucker. Some rights reserved.
Object photographs/images courtesy of the contributor(s): Shmahalo: Illustration originally published February 20, 2019 in Quanta Magazine. Reprinted with permission.
Coulon sketch, Chatterji, Dorfsman-Hopkins/Maynard, Delp, Robb/Benson, Bachman, Hsu, McShane polyhedron, Amethyst: © Silviana Amythest. Some rights reserved.
Rainford, Farris, Stange, Goerner, Felzenszwalb/Klivans/Skrodski, Sümmermann, Novello de Brito, Bozlee, Ghrist, Segerman: © 2020 by Henry Segerman.
Francis, Alatorre, Amethyst, Segerman, Holdener, Chéritat*, Antonsen shuffling, Taalman origami, Espigulé, McShane Antonsen shuffling: © Roger Antonsen. Some rights reserved.
bottle*, Taalman knot, Sander, Carter, Paul, Taalman pentagons, Labs brass, Buric, Gangl, Vejdemo-Johansson, Taalman (3 works): © Laura Taalman. Some rights reserved.
Tihanyi, Matsumoto jewelry, Labs glass, Edmark, Mohanty/Muetzel, Sharma, Arnoux line and color graphics, Davis/ Málaga Sabogal polyhedra: © Alba Marina Málaga Sabogal. Some rights reserved.
Lelièvre graphic, Goodman-Strauss bottom three images*, Séquin Athreya: images of object and contributor by Dan Paz, used with permission.
Carter: © J. Scott Carter. Some rights reserved.
Object photographs by Edmund Harriss: Coulon necklace: © 2019 Rémi COULON Some Rights Reserved
Málaga Sabogal/Lelièvre flat tori, Harriss curvahedra, Yackel, Abrams/Paul, Matsumoto fleece, Davis lizards*, Málaga Goodman-Strauss mural picture: © 2020 University of Arkansas. Photo: Russell Cothren. Used with permission.
Sabogal/Lelièvre laser box, Trettel, Dorfsman-Hopkins Julia set, Málaga Sabogal, Harriss brick, Coulon necklace, Goodman-Strauss mural page: © 2020 Chaim Goodman-Strauss. All rights reserved.
Whitney, Antonsen curves, Arnoux glass and 3D-printed objects, Davis/Lelièvre pentagons. Fenyes: © 2020 Aaron Fenyes
Antonsen curves: © Roger Antonsen. Some rights reserved.
Contributor photographs courtesy of the contributor(s): Goodman-Strauss surface page: © 2020 Chaim Goodman-Strauss. All rights reserved.
Kent, Chatterji, Maynard, Trnkova, Paul, Matsumoto, Bachman, Davis, Hsu, Rainford, Barnes, Schaubroeck, Kepner, Goodman-Strauss top surface picture: © National Museum of Mathematics. Used with permission. ** not yet **
Felzenszwalb/Klivans/Skrodski*, Zucker, Sümmermann, Shmahalo, Novello de Brito, Ghrist, Francis, Alatorre, Photo of ICERM participants by ICERM Staff. Used with permission.
Amethyst, Segerman, Chéritat, Taalman, Espigulé, Labs, Gangl, Vejdemo-Johansson, Tihanyi, Mohanty, Muetzel,
Goodman-Strauss, Fenyes, Séquin Section background images:
Drawings: courtesy of Dover Books ** not yet **
Contributor photographs by Diana Davis: Paper & fiber arts: by Diana Davis
Coulon, Dorfsman-Hopkins, Delp, Málaga Sabogal/Lelièvre, Harriss, Yackel, Abrams, Robb/Benson, McShane, Trettel, Laser cutting: by Edmund Harris
Farris, Stange, Goerner, Bozlee, Holdener, Antonsen, Málaga, Sander, Carter, Buric, Whitney, Edmark, Sharma, Graphics: background image from Wikipedia, public domain image
Arnoux, Lelièvre Video and virtual reality: Still frame from “Knots to Narnia” video courtesy of Anthony Phillips
3D printing: by Alba Marina Málaga Sabogal
Additional photo and image credits: Mechanical constructions: by Diana Davis
Kent: Originally appeared in Geometriae Dedicata. Reprinted with permission. ** not yet ** Multiple ways: Collage © 2020 by Margo Angelopoulos and Diana Davis. Background satellite image by NASA, reuse
Thurston/Sullivan: Curve picture by Ken Ribet, used with permission. Image of Thurston reprinted from Joseph permitted. Images of octopus, spider and eightfold path symbol are public domain. Quotation from Sebastian Junger,
Malkevitch, Remembering Bill Thurston (1946–2012), AMS Feature Column, April 2017. Image of Sullivan courtesy The Perfect Storm. Photographs of flower and stop sign by Diana Davis. All other graphics by Margo Angelopoulos.
of SUNY Stony Brook Office of Communications. Reprinted with permission. ** not yet **
Coulon sketch: © 2019 Rémi COULON Some Rights Reserved Cover images:
Málaga Sabogal flat tori: © Alba Marina Málaga Sabogal. Some rights reserved. Dorfsman-Hopkins/Maynard: used with permission.
Trnkova: © 2020 by the author Tihanyi: used with permission.
Málaga Sabogal laser box: © Alba Marina Málaga Sabogal. Some rights reserved. Bachman: Artwork by David Bachman, based on mathematics done jointly with Saul Schleimer and Henry Segerman
Barnes/Schaubroeck: Federal employee; no rights to transfer. Top left and top right images first published in Coloring
Book of Complex Function Representations by Julie Barnes, William Kreahling, and Beth Schaubroeck, in 2017,
published by the Mathematical Association of America. © 2017 held by the AMS. Bottom right image appeared in
AMS Page a Day Calendar by Evelyn Lamb in 2019, published by the AMS. Used with permission of the authors. * Contributed photographs substantially modified by Margo Angelopoulos

164 165
 ILLUSTRATING MATHEMATICS
This book is for anyone who wishes to illustrate
their mathematics, which – we hope! – means
everyone. It is organized by material, and
purposefully emphasizes the process of creating
things, including discussions of failures that
occurred along the way. In this way, the reader
can learn from the experiences of those who
came before, and will be inspired to create
their own illustrations.

Topics illustrated within include knot theory,

complex dynamics, groups and prime ideals,
the Riemann zeta function, quadratic fields,
hyperbolic space, flat and curved surfaces,
hyperbolic 3-manifolds, fractals, billiards,
tilings, prime numbers, algebraic surfaces, and
space-filling curves. Everyone who opens this
book should find a type of mathematics with
which they identify.

Diana Davis