Sie sind auf Seite 1von 14

University of Missouri Science and

Technology
Opportunities for Undergraduate Research
Experience

On the Existence of Perfect


Cuboids

Author:
Seth Kitchen

Adviser:
Dr. Matt Insall

Department of Mathematics and Statistics


March 31, 2015

Abstract
An Euler brick is a rectangular prism with integer sides and integer
face diagonals. A perfect cuboid is an Euler brick with an integer space
diagonal. An infinite amount of Euler bricks exist, but it is unanswered
whether a perfect cuboid exists. We show an exponential and logarithmic
equation which could be used to prove the existence or nonexistence of a
perfect cuboid (PC). We then continue the computer search for a PC.
About the author:
Seth Kitchen is a first year sophomore at Missouri S&T. He was born and
raised in Saint Peters, Missouri and graduated from Fort Zumwalt South
High School in 2014. At Missouri University of Science and Technology,
Seth is involved with the cross country and track teams, AAVG design
team, business incubator, and is an advanced mathematics tutor for Project
MEGSSS. He is currently managing the thermocouple bay for the electronics
subgroup of the campus rocket team. He has also worked on a high school
senior research project, The Determination of an Effective Parafoil for
Human-Powered Bicycle Aircraft, with a mentor from Boeing. Seth was
captain of the Missouri team at the US Department of Energy National
Science Bowl in 2014.

Contents
1 Introduction

2 Situation Analysis
2.1 General Equations and Methods . . . . . . . . . . . . . . . . . . . .
2.2 Problem Statement . . . . . . . . . . . . . . . . . . . . . . . . . . .

4
4
4

3 Work by Leech, Bergmann, and Guy


3.1 Previous Findings . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.2 An Introduced Logarithmic Approach . . . . . . . . . . . . . . . . .

5
5
5

4 Computer Search
4.1 Version 1: . . .
4.2 Version 2: . . .
4.3 Version 3: . . .
4.4 Version 4: . . .
4.5 Analysis: . . . .

6
6
6
6
7
7

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

Nomenclature

12

Acknowledgement

13

References

13

Introduction

The author became interested in the Euler Brick problem in his senior year of
high school. Being involved with academic teams, he was constantly looking for
a way to challenge himself. Since the beginning of his mathematics education,
every question had had an answer. Examples in textbooks, examples in lessons,
questions on tests, and contest questions had always been answerable. Through
his first eleven years of education, he did not know mathematics had problems that
were unanswerable. He thought mathematicians knew how to solve everything and
their only research was to find better ways to solve these problems. So naturally
when he Googled unanswered problems in mathematics, he did not expect to find
anything. He was wrong.
Wikipedia has a very long page on the topic. The Millennium problems are
the most well-known because if anyone solves one, he/she becomes a millionaire.
However, the millennium problems are probably also the hardest to understand let
alone solve. Since the author was confident in algebra, he turned to that category.
And naturally, he understood the Euler brick problem. Anyone that has learned
the Pythagorean Theorem can understand the Euler brick problem.
It was amazing to the author that something so simply understood, could not
so simply be solved. And the Euler brick problem is not alone in that aspect. Many
problems in number theory are just like the Euler brick problem. The most recently
solved problem (with exception to the Poincare Conjecture) that the media has
been concerned with, Fermat's Last Theorem, is very closely related to the Euler
brick problem.
The Euler brick problem is also easily programmable. The Pythagorean Theorem only requires an addition, power, and square root function which is available
in every language. The only complication is precision. Since perfect cuboids have
been searched through very high numbers that exceed the built-in data types,
an unlimited precision library was needed. Since Java has a built-in unlimited
precision library it was chosen.
The purpose of the research was to locate the best strategy for computing a
perfect cuboid.

2
2.1

Situation Analysis
General Equations and Methods

A cuboid is a special type of a parallelepiped also known as a right rectangular


prism. Cuboids have been studied since the early 1700s. A pioneer in cuboid
studies was Leonhard Euler, and as such, has a special classification named after
him. An Euler brick is a cuboid with integer edges and integer face diagonals.
The first, and smallest, Euler brick was found by Paul Halcke in 1719. Its edges
are (44, 117, 240) and diagonals (125, 244, 267). The most specific classification of
cuboid is yet to be found and is known as a perfect cuboid. A perfect cuboid
is an Euler brick with an integer space diagonal. Using right triangles and the
Pythagorean Theorem, a perfect cuboid satisfies the following set of Diophantine
equations:

a2 + b 2 = d 2

a2 + c2 = e2
(1)
b2 + c 2 = f 2

2
a + b2 + c 2 = g 2
With domain:


(a, b, c, d, e, f, g)Z

where a, b, c are the edge lengths, d, e, f are the face diagonals, and g is the space
diagonal. It is one of the last unsolved problems in Algebraic Number Theory.[4]
Many teams of computer scientists have tried to solve for a perfect cuboid.
The most efficient computer algorithms are derivatives of the one found by Bill
Butler. He states all odd sides up to 1010 have been tested, however, it is impossible
to have 100 percent validity in his or his colleagues testing (programming flaws,
skipping numbers, etc.).[1] The main mathematical approaches have used modular
arithmetic, parametric equations, elliptic curves, and symmetries to narrow down
the possibilities.

2.2

Problem Statement

The problem is known as the brick problem, diagonals problem, perfect box problem, perfect cuboid problem, or rational cuboid problem.[2] To solve the problem,
one must find values that satisfy the system of equations or prove that one cannot
exist. The focus of the research is to find the best algorithms to locate a solution.
All programming is done using Java JDK1.7.0. The programs run on an ECS
Intel(R) Pentium(R) Dual CPU E2180 at 140 2.00 GHz, 2.00GB RAM.

Work by Leech, Bergmann, and Guy

3.1

Previous Findings

Modular arithmetic using small primes has been very useful in surfacing new equations and limiting possible answers. Interesting and constructive work by Leech,
Bergmann, and Guy find that the product of a perfect cuboids edges and diagonals
are divisible by 28 34 53 7 11 13 17 19 29 37.[5]
Definition 3.1. We will denote this largest-found divisor of the product of a
perfect cuboids edges and diagonals as PED.
P ED = 899231100768000

(2)

Corollary 3.1.1. And denote K as the unsolved portion of the product. Thus the
following equation:
abcdef g = P ED K
(3)

3.2

An Introduced Logarithmic Approach

It is often easier to work with addition and subtraction rather than multiplication
and division. Taking the natural logarithm of both sides we can convert 3.12 into
such:
ln(a) + ln(b) + ln(c) + ln(d) + ln(e) + ln(f ) + ln(g) = ln(P ED) + ln(K)

(4)

Now we can work to lessen our variables. We can solve for ln(d), ln(e), ln(f ),
ln(g) by taking the natural logarithm of our first set of equations.

2
2

ln(d) = ln(a 2+b )

2
2

ln(e) = ln(b 2+c )


(5)
ln(a2 +c2 )

ln(f
)
=

2
2 +c2 )

ln(g) = ln(a +b
2
Substitution and solving for K yields:


ln(a2 + b2 + c2 )
exp 1 + 2 +
ln(P ED) = K
2
1 = ln a + ln b + ln c
2 =

ln(a2 +b2 )
2

ln(b2 +c2 )
2

ln(a2 +c2 )
2

(6)

And finally, because we do not care what K is, we just need it to be an integer,
we say:


ln(a2 + b2 + c2 )
ln(P ED) Z
(7)
exp 1 + 2 +
2
1 = ln a + ln b + ln c
2 =

ln(a2 +b2 )
2

ln(b2 +c2 )
2

ln(a2 +c2 )
2

Computer Search

Four programs were used on and off for months to locate patterns and brute
force an answer. Each can be found at https://sites.google.com/a/mst.edu/
sethkitchen/.

4.1

Version 1:

The most basic was a 3-nested loop that incremented 3 variables by 1 and checks
if 1 is true. It checks every number combination, and involves no analysis.

4.2

Version 2:

The second version had the same loop structure as Version 1, but instead used the
new logarithmic equations 7.

4.3

Version 3:

The third version utilizes a single equation found by Ruslan Sharipov.[3] It is


stated as:
u4 a4 b4 + 6a4 u2 b4 c2 2u4 a4 b2 c2 2u4 a2 b4 c2 + 4u2 b4 a2 c4 + 4a4 u2 b2 c4
12u4 a2 b2 c4 + u4 a4 c4 + u4 b4 c4 + a4 b4 c4 + 6a4 u2 c6 + 6u2 b4 c6
8a2 b2 u2 c6 2u4 a2 c6 2u4 b2 c6 2a4 b2 c6 2b4 a2 c6 + u4 c8
+b4 c8 + a4 c8 + 4a2 u2 c8 + 4b2 u2 c8 12b2 a2 c8 + 6u2 c10 2a2 c10
2b2 c10 + c12 = 0
The program has 4 nested loops, and tries every possibility.

(8)

4.4

Version 4:

The fourth version checks the parametric equations found by Euler and Saunderson. The equations provide an infinite number of Euler Bricks, but not all. The
equations are:

a = x(4y 2 z 2 )
b = y(4x2 z 2 )
(9)

c = 4xyz
where x, y, z are Pythagorean triples. My program generates Pythagorean triples
through Euclids algorithm. Euclid shows:

x = m2 n2
y = 2mn
(10)

z = m2 + n2

4.5

Analysis:

Each of these programs was run for 2 108 milliseconds ( 5.5 hours) to test their
speed in relation to each other. Since the speed of the programs was unknown, the
programs were set to run 104 iterations. Because this took a lot longer than 6 hours
for some programs, some were dropped to 103 (Version 1 and Version 3)and to 102
(Version 2). The data was then truncated, if need be, to the bounds of the graphs.
Please note the iterations were treated as the independent variable because it is
impossible to stop at an exact time, and check the value of the variable. It is only
possible to check after every iteration. Since more than one iteration can happen
in a millisecond, we check time after every 20 iterations. We then switch time to
the x axis to make the data more understandable.

Individual Graphs

(11)

I1 (t) 1.8674t0.3811

(12)

I2 (t) 0.2181t0.3669

(13)

I3 (t) 5.4807t0.3005

(14)

I4 (t) 1.9932t0.4939

Comparison Graphs

(15)

(16)

First it is important to note the spacing of the data points. The points originally
start closer together meaning more iterations happen closer to the start. Because
the lesser variables only increment to highest variables, they do not increment as
much in the beginning. For example, take the most basic program. When c is 5,
10

a and b increment once before c is incremented, completing the iteration. When c


is 6, there are four incrementations. And this process continues, gradually slowing
the rate of iteration.
From Figure 12 we see Version 2 is actually the slowest. Because the natural
logarithm of each value was taken, it was assumed the stored values would be
smaller, and therefore speed up the algorithm. However, the stored value is not
smaller in memory because the decimal portion of the value is increased as much
as the integer portion is decreased. In addition, the algorithm takes a lot of time to
transition from the original value to the logarithmic value. In 100 years, according
to the trend equation, this version would only iterate up to 8404. Although the
logarithmic equations are not suited for computation, they can still be used for
further analysis.
To compare the other versions, see Figure 15. The second slowest was Version
3. Because Sharipovs equation used 4 variables instead of 3, the extra loop is the
reason the version is not as fast. I believe the purpose of Sharipovs equation was
to make the Euler brick problem easier to work with. One equation can easily be
modified, and three cannot. However, this equation is not yet efficient enough for
computation. In 100 years, this version would only iterate to 31242.
The second fastest is surprisingly the most basic program. However, according
to the trend equation it would take more than a lifetime to recheck every value up
to 1010 . Incrementing the largest variable from 1010 to 1010 + 1 takes more than a
day.
The fastest program is the parametric program set up by Saunderson. The
reason for the speed is it only has two loops. The efficiency is also the best. Unlike
the other programs, the value incremented is not the value checked. Every time
the program increments the variable, a new Euler brick is produced. Every value
given by this program is an Euler brick. It skips all numbers that are not Euler
bricks. In 100 years, the program would iterate 3 million times, and reach Euler
brick values around 4 1033 . This is definitely the best structure for computation.
There is still a problem with this program. Spohn proved in 1972 that there
are no perfect cuboids produced by Saundersons equations. This paper calls for
parametric equations that contain every Euler brick. It is clear the fastest and
most efficient form of computation will be parametric equations. This finding is
the main result of this research.

11

Nomenclature

approximately equal to; according to trend

is a member of; is contained in

the set of all integers

a, b, c the edge lengths of a cuboid


d, e, f the face diagonals of a cuboid
g

the space diagonal of a cuboid

In (t) The amount of iterations as a function of time. N is the version number


m, n Generators of a pythagorean triple where m is larger than n
t

time

u, a, b, c variables created by Sharipov which would lead to a Perfect Cuboid


x, y, z A pythagorean triple where a is the smallest, b is the middle, and c is the
largest (hypotenuse)

12

Acknowledgement
I would like to thank my adviser, Dr. Matt Insall for his weekly input on this
paper. Since I was not here to apply last year, I quickly sent emails to professors
around campus the first week of school. He kindly accepted my request to research.
The conversion to logarithms was an idea he created, and this was a foundation
to the paper.
I would like to thank Kory Tennant, Rachel McCormick, and Kenny Sherman
for their attendance to the weekly meetings. They each offered help with the
paper.
I would like to thank Ruslan Sharipov, John Ramsden, and Scott Hampton
for their work and communication through the process. Sharipov is the most
published researcher in the area. Ramsden is related to the computational part
of the problem, and he is currently working on finding new parametric equations.
Hampton is working on a proof on nonexistence.

References
[1] Butler, Bill. Durango Bills The Integer Brick Problem (The Euler Brick Problem). http://www.durangobill.com/IntegerBrick.html
[2] Curtis, Fred. Primitive Euler Bricks. F2.org. N.p. Feb. 2001. Web. 8 Sept. 2014.
http://f2.org/maths/peb.html#atleast1edgediv11
[3] Sharipov, R.A. Perfect cuboids and irreducible polynomials. Ufa Mathematical Journal, ISSN 2074-1863, Vol. 4, No. 1 (2012), P. 153-160; see also
arXiv:1108.5348.
[4] Euler Brick. Wikipedia. Wikimedia Foundation, n.d. Web. 08 Jan. 2015.
[5] Van Luijk, Ronald. On Perfect Cuboids. n.d. June 2000. Web, 8 Jan. 2015.
http://www.math.leidenuniv.nl/~rvl/ps/cuboids.pdf

13

Das könnte Ihnen auch gefallen