See discussions, stats, and author profiles for this publication at: [Link]

net/publication/266329130

A note on “impossible” paper folding

Article in The American Mathematical Monthly · March 1996

DOI: 10.2307/2975374

CITATIONS READS
33 3,780

1 author:

Thomas C Hull
Franklin & Marshall College
69 PUBLICATIONS 3,256 CITATIONS

SEE PROFILE

All content following this page was uploaded by Thomas C Hull on 29 March 2016.

The user has requested enhancement of the downloaded file.

 Solving Cubics With Creases: The Work of
Beloch and Lill
Thomas C. Hull

Abstract
Margharita P. Beloch was the first person, in 1936, to realize that
origami (paper-folding) constructions can solve general cubic equa-
tions and thus are more powerful than straightedge and compass con-
structions. We present her proof. In doing this we use a delightful
(and mostly forgotten?) geometric method due to Eduard Lill for
finding the real roots of polynomial equations.

1 Introduction
There are many aspects to the mathematics of origami, or paper folding. One
may study combinatorial properties that emerge from folded paper. One can
study origami as mappings from the Euclidean plane into three-dimensional
space that have certain properties. But the oldest way to study origami
mathematically is as a method for geometric constructions. The idea is
to take a piece of paper and fold it, making a straight crease line. Then
we unfold the paper and make another crease line. In doing this we start
locating points of intersection of our crease lines and thus can try to construct
geometric figures, like an equilateral triangle or the angle bisector between
two lines.
This is similar, of course, to straightedge and compass constructions, ex-
cept it is not immediately clear that the circle-making power of a compass
could be duplicated by origami since we can only make straight crease lines.
A paper-folding skeptic might thus be surprised to learn that origami con-
structions are actually more powerful than those made by straightedge and
compass. Origami can trisect angles (see [10, 12, 14, 22]) and double cubes
(see [22, 23]), as well as solve general cubic equations (see [1, 6, 9]).

1
 But who was the first person to discover the full power of paper folding as
a geometric construction tool? The credit goes to an Italian mathematician
named Margharita Piazolla Beloch in the 1930s [4]. Given this, it is perhaps
more than a little embarassing that numerous researchers since [1, 3, 6, 9, 22],
including the author [10], have failed to cite Beloch’s ground-breaking work.
(Huzita, Scimemi [13], and Justin [14] are notable exceptions.)
In this paper, we present Beloch’s proof that paper folding can solve arbi-
trary cubic equations and thus solve the classic problems of angle trisection
and doubling the cube. At the same time, we will encounter a marvelous
geometric method for finding real roots of arbitrary polynomials due to Ed-
uard Lill [19]. We finish with a more extensive accounting of the history of
origami geometric constructions, arguing that Beloch was, indeed, the first
person to discover the full power of normal paper folding.

2 Beloch’s square
Like straightedge and compass constructions, any paper-folding construction
can be described as a sequence of elementary folding moves, or axioms, as
some call them. These basic moves can be classified by enumerating all of
the possible ways a single, straight crease line can be made by aligning given
points and lines to other points and lines already made on your paper [17].
Some examples of these basic folding moves are:

O1: Given two points P1 and P2 , we can make a crease line that places P1
onto P2 when folded.

O2: Given a line l and a point P not on l, we can make a crease line that
passes through P and is perpendicular to l.

For more information on these basic moves see [13] and [17]. Note, however,
that the above two basic moves can also be done by a straightedge and com-
pass. The one basic folding move which sets origami apart from straightedge
and compass constructions is the following:

The Beloch Fold: Given two points P1 and P1 and two lines l1
and l2 we can, whenever possible, make a single fold that places
P1 onto l1 and P2 onto l2 simultaneously. (See Figure 1.)

2
 Figure 1: The Beloch origami fold.

Figure 2: Folding a point to a line creates tangents to a parabola.

One way to see what this fold is doing is to consider one of the point-line
pairs. If we fold a point P to a line l, the resulting crease line will be tangent
to the parabola with focus P and directrix l (the equidistant set from P
and l). This can be demonstrated by the following activity: Take a piece of
paper, draw a point P on it, and let the bottom edge of the paper be the
line l. Then fold P to l over and over again. An easy way to do this is to
pick a point on l and fold it up to P , unfold, then pick a new point on l and
fold it to P , and repeat. After a diverse sampling of creases are made, the
outline of a parabola seems to emerge. Or, more precisely, the envelope of
the crease lines seems to be a parabola. (See Figure 2(a).) A proof of this
can be established as follows: After folding a point P ′ on l to P , draw a
line perpendicular to the folded image of l, on the folded flap of paper from
P to the crease line, as in Figure 2(b). If X is the point where this drawn
line intersects the crease line, then we see when unfolding the paper that the
point X is equidistant from the point P and the line l. (See Figure 2(c).)
Any other point on the crease line will be equidistant from P and P ′ and
thus will not have the same distance to the line l. Therefore the crease line
is tangent to the parabola with focus P and directrix l.
In other words, folding a point to a line can be thought of as locating a
point on a certain parabola, which means that this is equivalent to solving a

3
Figure 3: Two parabolas drawn in the plane can have at most three common
tangents..

Figure 4: The Beloch square.

quadratic equation.
The Beloch fold can then be interpreted thusly: Folding P1 to l1 will
make the crease be tangent to the parabola with focus P1 and directrix l1 ,
and folding P2 to l2 will make the crease be tangent to the P2 -focused and
l2 -directrixed parabola. In other words, this origami fold finds a common
tangent to two parabolas.
Now, two parabolas drawn in the plane can have at most three different
common tangents (for example, see Figure 3), suggesting that this origami
fold is equivalent to solving a cubic equation. Straightedge and compass
constructions, on the other hand, can only solve general quadratic equations.
Theoretically, we could end this paper right here, satisfied in the knowl-
edge that origami can solve cubic equations. (Specifically, on the projective
plane, finding common tangents to two parabolas is the dual problem to
finding intersections of conics, which allows general cubic solutions to be
constructed. See [1] and [28].) But Beloch provides a constructive proof.
She considers the following problem:

The Beloch Square: Given two points A and B and two lines r
and s in the plane, construct a square W XY Z with two adjacent

4
 Figure 5: Constructing the Beloch Square using origami.

corners X and Y lying on r and s, respectively, and the sides

W X and Y Z, or their extensions, passing through A and B,
respectively. (See Figure 4.)

Amazingly, this one √ construction problem captures everything we need

not only to construct 3 2 (thus solving one of the classic Greek construction
problems, that of doubling the volume of a cube) but also to solve arbitrary
cubic equations. Furthermore, this problem is readily solved via origami!
Here is how: We are given points A and B and lines r and s. Compute
the perpendicular distance from A to r and create a new line r ′ which is this
same distance from and parallel to r, so that r lies between A and r ′ . Do
the same with B and s to construct a line s′ . (See Figure 5, left.) Note
that these lines r ′ and s′ can be constructed easily via paper folding by, say,
folding along r, marking where A lands under this fold, and then making a
sequence of perpendicular folds O2 described above. (The details of this are
left as an exercise.)
We then perform the Beloch fold, folding A onto r ′ and B onto s′ si-
multaneously. (See Figure 5, center.) This will fold A to a point A′ on r ′
and B onto a point B ′ on s′ . The crease made from this fold will be the
perpendicular bisector of the segments AA′ and BB ′ . Therefore, if we let X
and Y be the midpoints of AA′ and BB ′ , respectively, we have that X lies
on r and Y lies on s because of the way in which r ′ and s′ were constructed.
The segment XY can then be one side of our Beloch square, and since AX
and BY are perpendicular to XY , we have that A and B are on opposite
sides, or extensions of sides, of this square.

5
 Figure 6: Beloch’s origami construction of the cube root of two.
√
3
3 Constructing 2
Next we will see how Beloch’s square allowed her to construct the cube root of
two. (Actually, what follows is her construction set on coordinate axes.) Let
us take r to be the y-axis and s to be the x-axis of the plane. Let A = (−1, 0)
and B = (0, −2). Then we construct the lines r ′ to be x = 1 and s′ to be
y = 2. Folding A onto r ′ and B onto s′ using the Beloch fold will make a
crease which crosses r at a point X and s at a point Y . Consulting Figure
6, if we let O be the origin, then notice that OAX, OXY , and OBY are all
similar right triangles. This follows from the fact that XY is perpendicular
to AA′ and BB ′ .
Therefore, we have |OX|/|OA| = |OY |/|OX| = |OB|/|OY |, where | · |
denotes the length of the segment. Filling in |OA| = 1 and |OB| = 2 gives
us |OX| = |OY |/|OX| = 2/|OY |. Finally, compute

|OY | 2
|OX|3 = |OX| · · = 2,
|OX| |OY |
√
and so X = (0, 3 2).
This construction is essentially the same as the one independently dis-
covered by Martin [22] fifty√years later, although Martin takes B = (0, −k)
3
so as to construct X = (0, k).

6
4 Solving cubic equations
Beloch goes on to describe how her square construction leads to a paper-
folding method for finding real roots of arbitrary cubic equations. For this
she refers to “the famous procedure of Lill for the graphical resolution of
equations of third degree” [4]. This “Lill’s method” does not seem to be as
famous now as it was in the 1930s. She is referring to an 1867 paper [19]
by an Austrian engineer named Eduard Lill. Felix Klein describes the cubic
case of Lill’s method in his 1926 book [16, p. 267], and he refers to it as well
known as well. The general method was described in a paper by Riaz in this
Monthly in 1962 [24], but since more recent citations of Lill’s method are
rare, we will reproduce this elegant method here.
Suppose we are given a polynomial f (x) = an xn + · · · + a1 x + a0 with real
coefficients and we would like to locate a real root of f (x), if one exists. Lill
suggests doing this geometrically by creating a path in the plane based on
the coefficients of f (x).
Imagine a turtle is sitting at the origin O and facing in the direction of the
positive x-axis. (Note that Lill did not use a turtle in his original exposition,
but the analogy to modern turtle graphics makes the metaphor especially
apt.) The turtle will walk along the positive x-axis a distance equal to the
coefficient an . Then the turtle will turn 90◦ counterclockwise and walk a
distance equal to the next coefficient an−1 . The turtle will then turn again
and repeat this process until ending at a point T after traveling a distance
a0 in some direction. If any of the coefficients are negative then the turtle
will walk backwards and that side of the turtle path will be considered to
have negative length. (E.g., in Figure 7(b) the sides marked a3 , a2 , and a0
all have negative length.) If any of the coefficients are zero then the turtle
will still turn but walk a distance of zero.
After the turtle has made this path we will position ourselves at O and
then attempt to “shoot” the turtle at T in the following way: We imagine
that we are living in a universe where bullets bounce off walls at 90◦ angles.
We fire a bullet from O at the line containing the turtle path segment of
length an−1 . This bullet will ricochet off this line at a right angle, then
ricochet off the line containing the side of length an−2 , and so on. Note that
the act of 90◦ ricocheting is ambiguous, since sometimes we want it to bounce
off the line on the same side as the bullet’s approach, while other times we
want it to bounce through the line (but still at a right angle). In all cases,
we make sure to choose the option that will allow the bullet to actually hit

7
Figure 7: Lill’s method turtle and bullet paths (a) for a quintic with all
coefficients positive and (b) a quintic with a3 , a2 , a0 < 0 and a5 , a4 , a1 > 0.

the next side of the turtle path. See Figure 7. If we are able to “hit” the
turtle in this way, then the bullet path will have n sides and our turtle path
n + 1 sides.
Let θ be the angle that the first part of our bullet path makes with the
x-axis (which contains the an side of the turtle path), assuming that we are
actually able to hit the turtle.
Claim: x = − tan θ is a root of f (x).
Our proof of this will assume that all our coefficients are positive, and
so our turtle and bullet paths will be as in Figure 7(a). The cases with
coefficients negative or zero are left as an exercise.
Notice that the sides of the bullet path are the hypotenuses of a sequence
of similar right triangles whose legs lie along the turtle path. Let yk be the
length of the side opposite the angle θ in the triangle whose side adjacent to
θ is part of the segment of length ak . Then we get
yn = (tan θ)an = −xan
yn−1 = (tan θ)(an−1 − (−xan )) = −x(an−1 + xan )
yn−2 = (tan θ)(an−2 − (−x(an−1 + xan ))) = −x(an−2 + x(an−1 + xan ))
..
.
y1 = −x(a1 + x(a2 + · · · + x(an−2 + x(an−1 + xan )) · · · )).

8
Figure 8: Lill’s method, cubic case, is really constructing a Beloch square.

But y1 = a0 . Equating these two values for y1 and simplifying gives us

f (x) = 0. If no value of θ will allow us to hit the turtle, then f (x) must have
no real roots.
Lill’s method is nothing short of amazing, and it gets better. The bullet
path turns out to be similar (in the geometric sense) to the turtle path one
would obtain from the polynomial f (x) with (x + tan θ) factored out. For
example, Riaz [24] demonstrates this with the polynomial x3 − 7x − 6. This
has three real roots, and each one corresponds to a different angle θ to shoot
the turtle. If we pick one, say the one which gives the root x = 3, then the
bullet path will be a rotated dilation of the turtle path for the polynomial
x2 + 3x + 2. (If the reader has dynamic geometry software available, it can
be used to demonstrate Lill’s method very convincingly, and this is highly
recommended.)
Beloch’s stroke of brilliance in paper-folding constructions was in seeing
that Lill’s method in the cubic case is just an application of her square
construction. Indeed, in the cubic case our turtle path for a3 x3 + a2 x2 +
a1 x + a0 will have four sides, so our bullet path will have three sides. If we
think of O as the point A, and T as the point B, and we think of the lines
containing the a2 -side and the a1 -side as the lines r and s, respectively, then
a Beloch square with adjacent corners on r and s and opposite sides passing
through O and T will give us a bullet path to shoot this turtle. (See Figure
8.) Therefore paper folding can be used to perform Lill’s method in the cubic
case and thus solve general polynomials of degree three.

9
5 Origami geometry history and future
The idea of using paper folding as a geometric construction tool has un-
known origins. Ancient Japanese sangaku (mathematical problems painted
on wooden tablets and hung in shrines, circa 1600-1890) have been found that
depict paper-folding geometry problems, indicating that the Japanese have
mathematical as well as religious and artistic traditions in origami [8]. But
the first known treatise on paper-folding constructions is T. Sundara Row’s
book Geometric Exercises in Paper Folding [25], first published in 1893. This
book was mentioned by Felix Klein in one of his popular math books of the
time [15], and this seems to have helped popularize paper-folding geometry.
Beloch herself states that Klein was the first to attract students to Row’s
book with his “autorevole giudizo” (authoritative judgement) [4]. However,
Row does not try to classify the basic origami moves (axioms). He defines
paper folding very broadly, employing folding moves as needed that place
points and lines onto previously constructed points and lines, and he does
not mention or make use of anything like Beloch’s fold. In fact, Row mis-
takenly claims that it is impossible to construct the cube root of two exactly
with paper folding [25, Section 112].
Row and Klein seem to have sparked a general interest in the geometry
of paper folding in the early 1900s. A number of papers appeared around
that time focusing on solving quadratic equations elegantly via origami, all
of which cite Row as a primary influence. For examples, see Lotka’s 1907
School Science and Mathematics paper [21] and Rupp’s 1924 paper in this
Monthly [26].
In 1930 Giovanni Vacca wrote an article [27] in the Italian journal Pe-
riodico di Mathematiche whose title translates as “On the folding of paper
applied to geometry.” In it Vacca briefly describes the history of paper fold-
ing, tracing it back to Chinese and Japanese origins (although some of these
are clearly speculative), and then proceeds to describe everything that is
known about the connections between origami and geometry at the time,
including references to Row, the influential educator Friedrich Froebel, and
others. He summarizes how origami can solve quadratic equations, but makes
no mention about whether or not origami could solve cubics. Vacca does not,
however, repeat Row’s claim that origami cannot construct cube roots.
This set the stage for Margherita Piazzolla Beloch, an algebraic geometer
at the University of Ferrara, Italy. She was born in 1879 and was the daughter
of the University of Rome’s renowned historian Karl Julius Beloch. She

10
received her doctorate in mathematics in 1908 at the University of Rome
under Guido Castelnuovo. She held positions at the University of Pavia
and Palermo, working with Michele de Franchis. In 1927 she was made
Chair of Geometry at the University of Ferrara, where she remained until her
retirement in 1955. While her primary research was in algebraic geometry,
many of her papers were on the application of geometry to photogrammetry,
the study of computing three-dimensional image data from photographs, such
as x-ray or aerial images. Beloch passed away while living in Rome in 1976
[20].
In 1936 Beloch published “Sul metodo del ripiegamento della carta per la
risoluzione dei problemi geometrici” (On the method of paper folding for the
resolution of geometric problems), in the journal Periodico di Mathematiche
[4]. She describes this paper as an extract from a mathematics course she
taught at Ferrara during the academic year 1933–34. This, together with
Vacca’s 1930 paper, seems to be strong evidence that Beloch was the first to
discover that origami can find common tangents to two parabolas by folding
two points to two lines simultaneously and thus solve general cubic equations.
Beloch was also quick to note that since solving quartics can be reduced to
solving cubics and quadratics, we know that origami can find real roots of
quartic equations.
Since then it has been demonstrated that the Beloch fold is the most
complicated paper-folding move possible [17]. By this we mean that if one
tries to write a list of all possible origami moves (like O1 and O2 described
earlier) that only produce a single, straight crease line, then no other such
origami move will give us more algebraic power than the Beloch fold. In
other words, Beloch’s work does, in fact, determine the constructible limit of
normal paper folding.
To clarify, however, this work concerns only straight-crease, one-fold-at-
a-time origami. Other directions in origami constructions can and have been
explored. Folding curved creases (not straight lines) is possible, although dif-
ficult [7], and spoils the construction game completely by allowing transcen-
dentals like π to be constructed [11]. Also, Robert Lang has demonstrated
that if we allow ourselves to make simultaneous creases, i.e., an origami move
that produces more than one crease line, made in unison (such origami moves
are called multifolds), then arbitrary angle quintisections can be performed
[18]. In fact, Alperin and Lang have recently shown that if three simultaneous
creases are allowed, then arbitrary quintics can be solved [2]. They use the
quintic case of Lill’s method to demonstrate that this can be done in theory,

11
although actually performing such a complex fold seems physically impos-
sible to do in general. Furthermore, Chow and Fan argue [5] that roots of
polynomials of arbitrary degree can be found if any number of simultaneous
folds are allowed. Nonetheless, Beloch deserves the credit for first discovering
the geometric limits of origami that mere mortals are able to perform.

Acknowledgments. The author would like to thank Benedetto Scimemi

for his help in providing biographic information about Beloch, as well as for
being one of the only people in the past 50+ years to bring attention to
Beloch’s work in paper folding.

References
[1] R. C. Alperin, A mathematical theory of origami constructions and num-
bers, New York J. Math. 6 (2000) 119-133.

[2] R. C. Alperin and R. J. Lang, One-, two-, and multi-fold origami axioms,
in Origami4 —Pasadena, CA, 2006, International Meeting of Origami in
Science, Mathematics, and Education, R. J. Lang, ed., A K Peters,
Natick, MA, 2009, 371-393.

[3] D. Auckly and J. Cleveland, Totally real origami and impossible paper
folding, Amer. Math. Monthly 102 (1995) 215-226.

[4] M. P. Beloch, Sul metodo del ripiegamento della carta per la risoluzione
dei problemi geometrici, Periodico di Mathematiche Ser. 4, 16 (1936)
104-108.

[5] T. Y. Chow and K. Fan, The power of multifolds: Folding the algebraic
closure of the rational numbers, in Origami4—Pasadena, CA, 2006, In-
ternational Meeting of Origami in Science, Mathematics, and Education,
R. J. Lang, ed., A K Peters, Natick, MA, 2009, 395-404.

[6] D. Cox and J. Shurman, Geometry and number theory on clovers, Amer.
Math. Monthly 112 (2005) 682-704.

[7] D. Fuchs and S. Tabachnikov, More on paperfolding, Amer. Math.

Monthly 106 (1999) 27-35.

12
 [8] H. Fukagawa and D. Pedoe, Japanese Temple Geometry Problems,
Charles Babbage Research Centre, Winnipeg, Canada, 1989.

[9] R. Geretschläger, Euclidean constructions and the geometry of origami,

Math. Mag. 68 (1995) 357-371.

[10] T. Hull, A note on “impossible” paperfolding, Amer. Math. Monthly

103 (1996) 242-243.

[11] ———, Constructing π via origami (2007), available at [Link]

[Link]/~thull/papers/[Link].

[12] K. Hushimi, Trisection of angle by H. Abe (in Japanese), in Science of

Origami, a supplement to Saiensu (Japanese version of Scientific Amer-
ican), Oct. (1980) 8.

[13] H. Huzita and B. Scimemi, The algebra of paper-folding (origami), in

Origami Science and Technology—Ferrara, Italy 1989, Proc. First Inter.
Meet. Origami Science and Technology, H. Huzita, ed., self-published,
Ferrara, Italy, (1990) 205-222.

[14] J. Justin, Résolution par le pliage de l’équation du troisième degré et

applications géométriques, L’Ouvert: Journal of the APMEP of Alsace
and the IREM of Strasbourg, 42 (1986) 9-19.

[15] F. Klein, Famous Problems of Elementary Geometry, Ginn, Boston,

1897.

[16] ———, Elementarmathematik von höheren Standpunkte aus: II: Ge-

ometrie, Berlin, Springer, 1926.

[17] R. J. Lang, Origami and geometric constructions (2003), available at

[Link]

[18] ———, Angle quintisection (2004), available at [Link]

[Link]/science/quintisection/quintisection.php4/.

[19] E. Lill, Résolution graphique des equations numériques d’un degré quel-
conque à une inconnue, Nouv. Annales Math. Ser. 2, 6 (1867) 359-362.

[20] S. Linguerri, Scienza a due voci – Beloch Piazzolla Margherita, available

at [Link]

13
 [21] A. J. Lotka, Construction of conic sections by paper-folding, School
Science and Mathematics 7 (1907) 595-597.

[22] G. E. Martin, Paper-folding, New York State Mathematics Teachers’

Journal 35 (1985) 137-140.

[23] P. Messer, Problem 1054, Crux Mathematicorum 12 (1986) 284-285.

[24] M. Riaz, Geometric solutions of algebraic equations, Amer. Math.

Monthly 69 (1962) 654-658.

[25] T. S. Row, Geometric Exercises in Paper Folding, Open Court, Chicago,

1901.

[26] C. A. Rupp, On a transformation by paper folding, Amer. Math.

Monthly 31 (1924) 432-435.

[27] G. Vacca, Della piegatura della carta applicata alla geometria, Periodico
di Mathematiche Ser. 4, 10 (1930) 43-50.

[28] C. R. Videla, On points constructible from conics, Math. Intelligencer

19, no. 2 (1997) 53-57.

Thomas C. Hull received his B.A. from Hampshire College in 1991 and his
Ph.D. from the University of Rhode Island in 1997. He currently teaches at
Western New England College in Springfield, MA. He has been folding paper
almost as long as he has been doing math, and from 1995 to 2008 he was on
the board of directors for OrigamiUSA. He has one book, Project Origami
(A K Peters), devoted to origami math, and he is slowly working on a second
one.
Department of Mathematics, Mailbox H-5174, Western New England College,
1215 Wilbraham Road, Springfield, MA 01119
thull@[Link]

14

View publication stats